add custom dialer

- Validate resolved IPs immediately before connecting.
- Guard Shoutrrr requests during initialization and delivery.
- Require admins for services without HTTP client support.
- Test DNS rebinding, redirects, internal hosts, and authorization.

.dockerignore

@@ -1,6 +1,6 @@
 # Node.js dependencies
-node_modules/
-**/node_modules/
+node_modules
+internalsite/node_modules
 
 # Go build artifacts and binaries
 build

.github/workflows/docker-images.yml

@@ -29,7 +29,6 @@ jobs:
           # henrygd/beszel-agent:alpine
           - image: henrygd/beszel-agent
             dockerfile: ./internal/dockerfile_agent_alpine
-            flavor: latest=false
             registry: docker.io
             username_secret: DOCKERHUB_USERNAME
             password_secret: DOCKERHUB_TOKEN
@@ -56,7 +55,6 @@ jobs:
           # henrygd/beszel-agent-nvidia:slim
           - image: henrygd/beszel-agent-nvidia
             dockerfile: ./internal/dockerfile_agent_nvidia_slim
-            flavor: latest=false
             platforms: linux/amd64,linux/arm64
             registry: docker.io
             username_secret: DOCKERHUB_USERNAME
@@ -125,7 +123,6 @@ jobs:
           # ghcr.io/henrygd/beszel-agent-nvidia:slim
           - image: ghcr.io/${{ github.repository }}/beszel-agent-nvidia
             dockerfile: ./internal/dockerfile_agent_nvidia_slim
-            flavor: latest=false
             platforms: linux/amd64,linux/arm64
             registry: ghcr.io
             username: ${{ github.actor }}
@@ -153,7 +150,6 @@ jobs:
           # ghcr.io/henrygd/beszel-agent:alpine
           - image: ghcr.io/${{ github.repository }}/beszel-agent
             dockerfile: ./internal/dockerfile_agent_alpine
-            flavor: latest=false
             registry: ghcr.io
             username: ${{ github.actor }}
             password_secret: GITHUB_TOKEN
@@ -163,7 +159,7 @@ jobs:
               type=semver,pattern={{major}}.{{minor}}-alpine
               type=semver,pattern={{major}}-alpine
 
-          # henrygd/beszel-agent
+          # henrygd/beszel-agent (keep at bottom so it gets built after :alpine and gets the latest tag)
           - image: henrygd/beszel-agent
             dockerfile: ./internal/dockerfile_agent
             registry: docker.io
@@ -204,8 +200,6 @@ jobs:
         uses: docker/metadata-action@v6
         with:
           images: ${{ matrix.image }}
-          # Variant images must not overwrite the standard image's latest tag.
-          flavor: ${{ matrix.flavor || 'latest=auto' }}
           tags: ${{ matrix.tags }}
 
       # https://github.com/docker/login-action

agent/agent.go

@@ -48,9 +48,7 @@ type Agent struct {
 	keys                      []gossh.PublicKey                                     // SSH public keys
 	smartManager              *SmartManager                                         // Manages SMART data
 	systemdManager            *systemdManager                                       // Manages systemd services
-	monitorManager            *MonitorManager                                       // Manages network monitors
-	storagePoolManager        *StoragePoolManager                                   // Manages storage pool and dataset data
-	packageUpdates            *packageUpdatesManager                                // Checks for pending package updates
+	zfsManager                *ZfsManager                                           // Manages ZFS pool and dataset data
 }
 
 // NewAgent creates a new agent with the given data directory for persisting data.
@@ -124,15 +122,12 @@ func NewAgent(dataDir ...string) (agent *Agent, err error) {
 	// initialize handler registry
 	agent.handlerRegistry = NewHandlerRegistry()
 
-	// initialize monitor manager
-	agent.monitorManager = newMonitorManager()
+	agent.zfsManager = newZfsManager()
 
-	agent.storagePoolManager = newStoragePoolManager()
-
-	// Retain ZFS_INTERVAL for the shared storage pool detail refresh interval.
+	// ZFS_INTERVAL env var to update ZFS detail data at this interval
 	if zfsIntervalEnv, exists := utils.GetEnv("ZFS_INTERVAL"); exists {
 		if duration, err := time.ParseDuration(zfsIntervalEnv); err == nil && duration > 0 {
-			agent.storagePoolManager.detailInterval = duration
+			agent.zfsManager.detailInterval = duration
 			agent.systemDetails.ZfsInterval = duration
 			slog.Info("ZFS_INTERVAL", "duration", duration)
 		} else {
@@ -156,8 +151,6 @@ func NewAgent(dataDir ...string) (agent *Agent, err error) {
 		slog.Debug("SMART", "err", err)
 	}
 
-	agent.packageUpdates = newPackageUpdatesManager(agent.dataDir)
-
 	// initialize GPU manager
 	agent.gpuManager, err = NewGPUManager()
 	if err != nil {
@@ -199,11 +192,6 @@ func (a *Agent) gatherStats(options common.DataRequestOptions) *system.CombinedD
 		}
 	}
 
-	if a.monitorManager != nil {
-		data.Monitors = a.monitorManager.GetResults(cacheTimeMs)
-		slog.Debug("Monitors", "data", data.Monitors)
-	}
-
 	// skip updating systemd services if cache time is not the default 60sec interval
 	if a.systemdManager != nil && cacheTimeMs == defaultDataCacheTimeMs {
 		totalCount := uint16(a.systemdManager.getServiceStatsCount())
@@ -222,10 +210,6 @@ func (a *Agent) gatherStats(options common.DataRequestOptions) *system.CombinedD
 		}
 	}
 
-	if a.packageUpdates != nil {
-		data.Info.PackageUpdates = a.packageUpdates.get(time.Now())
-	}
-
 	data.Stats.ExtraFs = make(map[string]*system.FsStats)
 	data.Info.ExtraFsPct = make(map[string]float64)
 	for name, stats := range a.fsStats {
@@ -259,11 +243,7 @@ func (a *Agent) gatherStats(options common.DataRequestOptions) *system.CombinedD
 // Start initializes and starts the agent with optional WebSocket connection
 func (a *Agent) Start(serverOptions ServerOptions) error {
 	a.keys = serverOptions.Keys
-	err := a.connectionManager.Start(serverOptions)
-	if err != nil {
-		a.cleanupSensorShadow()
-	}
-	return err
+	return a.connectionManager.Start(serverOptions)
 }
 
 func (a *Agent) getFingerprint() string {

agent/btrfs/btrfs.go

@@ -1,26 +0,0 @@
-// Package btrfs reads btrfs filesystem state from sysfs.
-package btrfs
-
-// Filesystem is a mounted btrfs filesystem read from /sys/fs/btrfs/<uuid>.
-type Filesystem struct {
-	UUID     string // stable filesystem UUID from sysfs
-	MountID  string // kernel filesystem identity for matching monitored mounts
-	IODevice string // sole member block-device name, empty for multi-device/unknown pools
-	Name     string // label, else first mountpoint, else UUID
-	Size     uint64 // effective usable capacity, or raw member capacity when Raw
-	Raw      bool   // capacity and usage are physical bytes, unsuitable for disk alerts
-	Alloc    uint64 // raw bytes allocated to data, metadata and system chunks
-	Health   string // ONLINE, or DEGRADED when a device is missing
-	NRead    uint64 // cumulative bytes read across member devices
-	NWrite   uint64 // cumulative bytes written across member devices
-	Devices  []Device
-}
-
-// Device is one member device (devinfo/<devid>) with its error counters.
-type Device struct {
-	Name           string // "devid N"; sysfs does not expose the block device path
-	State          string // ONLINE or MISSING
-	ReadErrs       uint64
-	WriteErrs      uint64
-	CorruptionErrs uint64
-}

agent/btrfs/btrfs_linux.go

@@ -1,285 +0,0 @@
-//go:build linux
-
-package btrfs
-
-import (
-	"errors"
-	"fmt"
-	"os"
-	"path/filepath"
-	"strconv"
-	"strings"
-	"unsafe"
-
-	"github.com/henrygd/beszel/agent/utils"
-	"golang.org/x/sys/unix"
-)
-
-var (
-	sysfsPath       = "/sys/fs/btrfs"
-	mountsPath      = "/proc/self/mounts"
-	mountinfoPath   = "/proc/self/mountinfo"
-	mountUUID       = MountID
-	deviceSize      = ioctlDeviceSize
-	filesystemUsage = statfsUsage
-)
-
-// Filesystems returns all mounted btrfs filesystems, or nil when there are none.
-func Filesystems() ([]Filesystem, error) {
-	entries, err := os.ReadDir(sysfsPath)
-	if errors.Is(err, os.ErrNotExist) {
-		return nil, nil
-	}
-	if err != nil {
-		return nil, err
-	}
-	mounts := mountpointsByDevice()
-	var filesystems []Filesystem
-	for _, entry := range entries {
-		if !entry.IsDir() || entry.Name() == "features" {
-			continue
-		}
-		fs, err := readFilesystem(filepath.Join(sysfsPath, entry.Name()), mounts)
-		if err != nil {
-			return nil, fmt.Errorf("btrfs %s: %w", entry.Name(), err)
-		}
-		filesystems = append(filesystems, fs)
-	}
-	return filesystems, nil
-}
-
-func readFilesystem(dir string, mounts map[string]string) (Filesystem, error) {
-	fs := Filesystem{UUID: filepath.Base(dir), Name: utils.ReadStringFile(filepath.Join(dir, "label")), Health: "UNKNOWN"}
-	for _, kind := range []string{"data", "metadata", "system"} {
-		if value, ok := utils.ReadUintFile(filepath.Join(dir, "allocation", kind, "disk_used")); ok {
-			fs.Alloc += value
-		}
-	}
-	// devices/<name> links to the block device's sysfs directory.
-	devices, err := os.ReadDir(filepath.Join(dir, "devices"))
-	if err != nil && !errors.Is(err, os.ErrNotExist) {
-		return fs, err
-	}
-	mountpoint := mounts["uuid:"+fs.UUID]
-	if fs.Name == "" {
-		fs.Name = mountpoint
-	}
-	var backingSize uint64
-	for _, dev := range devices {
-		if mountpoint == "" {
-			mountpoint = mounts[dev.Name()]
-		}
-		if fs.Name == "" {
-			fs.Name = mountpoint
-		}
-		devDir := filepath.Join(dir, "devices", dev.Name())
-		if size, ok := utils.ReadUintFile(filepath.Join(devDir, "size")); ok {
-			backingSize += size * 512
-		}
-		if stat := strings.Fields(utils.ReadStringFile(filepath.Join(devDir, "stat"))); len(stat) >= 7 {
-			fs.NRead += parseUint(stat[2]) * 512
-			fs.NWrite += parseUint(stat[6]) * 512
-		}
-	}
-	devids, err := os.ReadDir(filepath.Join(dir, "devinfo"))
-	if err != nil && !errors.Is(err, os.ErrNotExist) {
-		return fs, err
-	}
-	capacityAvailable := len(devids) > 0
-	healthKnown := len(devids) > 0
-	for _, devid := range devids {
-		devDir := filepath.Join(dir, "devinfo", devid.Name())
-		// Replacement targets do not add filesystem capacity.
-		replaceTarget, _ := utils.ReadUintFile(filepath.Join(devDir, "replace_target"))
-		if replaceTarget != 1 {
-			devid, err := strconv.ParseUint(devid.Name(), 10, 64)
-			if err != nil {
-				return fs, err
-			}
-			size, err := deviceSize(mountpoint, devid)
-			if err != nil {
-				capacityAvailable = false
-			}
-			fs.Size += size
-		}
-		dev := Device{Name: "devid " + devid.Name(), State: "ONLINE"}
-		missing := utils.ReadStringFile(filepath.Join(devDir, "missing"))
-		if missing != "0" && missing != "1" {
-			healthKnown = false
-			dev.State = "UNKNOWN"
-		}
-		if missing == "1" {
-			dev.State = "MISSING"
-			fs.Health = "DEGRADED"
-		}
-		for line := range strings.Lines(utils.ReadStringFile(filepath.Join(devDir, "error_stats"))) {
-			if fields := strings.Fields(line); len(fields) == 2 {
-				switch fields[0] {
-				case "read_errs":
-					dev.ReadErrs = parseUint(fields[1])
-				case "write_errs":
-					dev.WriteErrs = parseUint(fields[1])
-				case "corruption_errs":
-					dev.CorruptionErrs = parseUint(fields[1])
-				}
-			}
-		}
-		fs.Devices = append(fs.Devices, dev)
-	}
-	// Use one capacity source for the whole filesystem: device IDs cannot be
-	// reliably matched to block-device names in sysfs. A partial ioctl result
-	// must not be added to the complete backing-device total.
-	if !capacityAvailable {
-		fs.Size = backingSize
-	}
-	if fs.Health != "DEGRADED" && healthKnown {
-		fs.Health = "ONLINE"
-	}
-	fs.MountID = mountUUID(mountpoint)
-	if len(devices) == 1 && len(devids) == 1 && fs.Health == "ONLINE" {
-		fs.IODevice = devices[0].Name()
-	}
-	fs.Raw = true
-	if used, available, err := filesystemUsage(mountpoint); err == nil {
-		// Effective capacity excludes reserved/unavailable space, so Size-Alloc
-		// is available to applications and the usage ratio matches df.
-		fs.Size, fs.Alloc, fs.Raw = used+available, used, false
-	}
-	if fs.Name == "" {
-		fs.Name = filepath.Base(dir)
-	}
-	return fs, nil
-}
-
-// mountpointsByDevice prefers UUID matches from mountinfo and retains source
-// device names as a fallback for environments where FS_INFO is unavailable.
-func mountpointsByDevice() map[string]string {
-	mounts := mountpointsByUUID(utils.ReadStringFile(mountinfoPath), mountUUID)
-	for line := range strings.Lines(utils.ReadStringFile(mountsPath)) {
-		fields := strings.Fields(line)
-		if len(fields) < 3 || fields[2] != "btrfs" {
-			continue
-		}
-		device := fields[0]
-		if resolved, err := filepath.EvalSymlinks(device); err == nil {
-			device = resolved
-		}
-		if _, seen := mounts[filepath.Base(device)]; !seen {
-			mounts[filepath.Base(device)] = unescapeMountPath(fields[1])
-		}
-	}
-	return mounts
-}
-
-func parseUint(s string) uint64 {
-	n, _ := strconv.ParseUint(s, 10, 64)
-	return n
-}
-
-// ioctlDeviceSize reads Btrfs's recorded device size, which can be smaller
-// than the block device after a filesystem resize. BTRFS_IOC_DEV_INFO is
-// _IOWR(0x94, 30, struct btrfs_ioctl_dev_info_args), a 4096-byte ABI structure.
-func ioctlDeviceSize(mountpoint string, devid uint64) (uint64, error) {
-	if mountpoint == "" {
-		return 0, errors.New("no accessible mountpoint")
-	}
-	f, err := os.Open(mountpoint)
-	if err != nil {
-		return 0, err
-	}
-	defer f.Close()
-	args := struct {
-		Devid      uint64
-		UUID       [16]byte
-		BytesUsed  uint64
-		TotalBytes uint64
-		Reserved   [4096 - 40]byte
-	}{Devid: devid}
-	_, _, errno := unix.Syscall(unix.SYS_IOCTL, f.Fd(), 0xd000941e, uintptr(unsafe.Pointer(&args)))
-	if errno != 0 {
-		return 0, errno
-	}
-	return args.TotalBytes, nil
-}
-
-// The filesystem magic is unsigned even when Statfs_t.Type is int32.
-func isBtrfs(stat *unix.Statfs_t) bool {
-	return uint32(stat.Type) == unix.BTRFS_SUPER_MAGIC
-}
-
-func statfsUsage(path string) (used, available uint64, err error) {
-	if path == "" {
-		return 0, 0, errors.New("no accessible mountpoint")
-	}
-	var stat unix.Statfs_t
-	if err = unix.Statfs(path, &stat); err != nil {
-		return
-	}
-	if !isBtrfs(&stat) {
-		return 0, 0, errors.New("mountpoint is not Btrfs")
-	}
-	blockSize := uint64(stat.Bsize)
-	return (stat.Blocks - min(stat.Blocks, stat.Bfree)) * blockSize, min(stat.Blocks, stat.Bavail) * blockSize, nil
-}
-
-// MountID returns the filesystem UUID via BTRFS_IOC_FS_INFO. Unlike statfs
-// f_fsid, this identity is shared by all subvolumes and bind mounts.
-func MountID(path string) string {
-	if path == "" {
-		return ""
-	}
-	var stat unix.Statfs_t
-	if unix.Statfs(path, &stat) != nil || !isBtrfs(&stat) {
-		return ""
-	}
-	f, err := os.Open(path)
-	if err != nil {
-		return ""
-	}
-	defer f.Close()
-	args := struct {
-		MaxID      uint64
-		NumDevices uint64
-		FSID       [16]byte
-		Reserved   [992]byte
-	}{}
-	// _IOR(0x94, 31, 1024). Reuse the platform's read-direction bits;
-	// MIPS/PowerPC use a different encoding than asm-generic.
-	request := uintptr(unix.FS_IOC_GETFLAGS&0xe0000000) | 0x0400941f
-	_, _, errno := unix.Syscall(unix.SYS_IOCTL, f.Fd(), request, uintptr(unsafe.Pointer(&args)))
-	if errno != 0 {
-		return ""
-	}
-	id := args.FSID
-	return fmt.Sprintf("%x-%x-%x-%x-%x", id[:4], id[4:6], id[6:8], id[8:10], id[10:])
-}
-
-// Btrfs mountinfo device numbers can be virtual (0:N), so query the UUID
-// through the mount instead of comparing those numbers with sysfs block devs.
-// Retry another path when a bind mount is inaccessible. Once resolved, reuse
-// the result for that mount device to avoid opening every Docker bind mount.
-func mountpointsByUUID(mountinfo string, identify func(string) string) map[string]string {
-	mounts := make(map[string]string)
-	resolved := make(map[string]bool)
-	for line := range strings.Lines(mountinfo) {
-		before, after, ok := strings.Cut(line, " - ")
-		fields, fs := strings.Fields(before), strings.Fields(after)
-		if !ok || len(fields) < 6 || len(fs) < 3 || fs[0] != "btrfs" || resolved[fields[2]] {
-			continue
-		}
-		path := unescapeMountPath(fields[4])
-		uuid := identify(path)
-		if uuid == "" {
-			continue
-		}
-		resolved[fields[2]] = true
-		if mounts["uuid:"+uuid] == "" {
-			mounts["uuid:"+uuid] = path
-		}
-	}
-	return mounts
-}
-
-func unescapeMountPath(path string) string {
-	return strings.NewReplacer(`\040`, " ", `\011`, "\t", `\012`, "\n", `\134`, `\`).Replace(path)
-}

agent/btrfs/btrfs_linux_test.go

@@ -1,274 +0,0 @@
-//go:build testing && linux
-
-package btrfs
-
-import (
-	"os"
-	"path/filepath"
-	"strconv"
-	"testing"
-
-	"github.com/henrygd/beszel/agent/utils"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-	"golang.org/x/sys/unix"
-)
-
-func TestFilesystems(t *testing.T) {
-	root := t.TempDir()
-	oldSysfs, oldMounts := sysfsPath, mountsPath
-	sysfsPath, mountsPath = root, filepath.Join(root, "mounts")
-	t.Cleanup(func() { sysfsPath, mountsPath = oldSysfs, oldMounts })
-
-	fsDir := filepath.Join(root, "1b2c3d4e-0000-0000-0000-000000000000")
-	write := func(rel, content string) {
-		path := filepath.Join(fsDir, rel)
-		require.NoError(t, os.MkdirAll(filepath.Dir(path), 0o755))
-		require.NoError(t, os.WriteFile(path, []byte(content), 0o644))
-	}
-	require.NoError(t, os.MkdirAll(filepath.Join(root, "features"), 0o755))
-	oldUsage := filesystemUsage
-	filesystemUsage = func(string) (uint64, uint64, error) { return 0, 0, os.ErrNotExist }
-	t.Cleanup(func() { filesystemUsage = oldUsage })
-	oldDeviceSize := deviceSize
-	t.Cleanup(func() { deviceSize = oldDeviceSize })
-	deviceSize = func(_ string, devid uint64) (uint64, error) {
-		value, _ := utils.ReadUintFile(filepath.Join(fsDir, "recorded-size", strconv.FormatUint(devid, 10)))
-		return value, nil
-	}
-	// Recorded member capacities differ from the unchanged backing devices.
-	write("recorded-size/1", "256000\n")
-	write("recorded-size/2", "128000\n")
-	write("label", "tank\n")
-	write("allocation/data/disk_used", "4096\n")
-	write("allocation/metadata/disk_used", "2048\n")
-	write("allocation/system/disk_used", "1024\n")
-	write("devices/sda/size", "1000\n")
-	write("devices/sda/stat", "10 0 200 0 20 0 400 0 0 0 0\n")
-	write("devices/sdb/size", "1000\n")
-	write("devices/sdb/stat", "10 0 100 0 20 0 100 0 0 0 0\n")
-	write("devinfo/1/missing", "0\n")
-	write("devinfo/1/error_stats", "write_errs 1\nread_errs 2\nflush_errs 0\ncorruption_errs 3\ngeneration_errs 0\n")
-	write("devinfo/2/missing", "1\n")
-
-	filesystems, err := Filesystems()
-	require.NoError(t, err)
-	require.Len(t, filesystems, 1)
-	assert.Equal(t, Filesystem{
-		UUID: "1b2c3d4e-0000-0000-0000-000000000000", Raw: true, Name: "tank", Size: 384000, Alloc: 7168, Health: "DEGRADED", NRead: 153600, NWrite: 256000,
-		Devices: []Device{
-			{Name: "devid 1", State: "ONLINE", ReadErrs: 2, WriteErrs: 1, CorruptionErrs: 3},
-			{Name: "devid 2", State: "MISSING"},
-		},
-	}, filesystems[0])
-
-	// Unlabeled filesystems fall back to the first mountpoint, then the UUID.
-	write("label", "\n")
-	require.NoError(t, os.WriteFile(mountsPath, []byte(
-		"/dev/sdz1 /other btrfs rw 0 0\n/dev/sdb /mnt/storage btrfs rw 0 0\n/dev/sdb /mnt/storage/sub btrfs rw,subvol=/sub 0 0\n",
-	), 0o644))
-	filesystems, err = Filesystems()
-	require.NoError(t, err)
-	assert.Equal(t, "/mnt/storage", filesystems[0].Name)
-
-	require.NoError(t, os.Remove(mountsPath))
-	filesystems, err = Filesystems()
-	require.NoError(t, err)
-	assert.Equal(t, "1b2c3d4e-0000-0000-0000-000000000000", filesystems[0].Name)
-	write("devinfo/3/replace_target", "1\n")
-	write("recorded-size/3", "512000\n")
-	filesystems, err = Filesystems()
-	require.NoError(t, err)
-	assert.Equal(t, uint64(384000), filesystems[0].Size, "replacement target must not inflate capacity")
-
-	deviceSize = func(string, uint64) (uint64, error) { return 0, os.ErrPermission }
-	filesystems, err = Filesystems()
-	require.NoError(t, err)
-	require.Len(t, filesystems, 1)
-	assert.Equal(t, uint64(1024000), filesystems[0].Size)
-	assert.Equal(t, "DEGRADED", filesystems[0].Health)
-	assert.Equal(t, uint64(153600), filesystems[0].NRead)
-
-	// A partial ioctl result must not be mixed with the backing-device total.
-	deviceSize = func(_ string, devid uint64) (uint64, error) {
-		if devid == 2 {
-			return 0, os.ErrPermission
-		}
-		return 256000, nil
-	}
-	filesystems, err = Filesystems()
-	require.NoError(t, err)
-	assert.Equal(t, uint64(1024000), filesystems[0].Size)
-
-	// With no mount visible (e.g. Docker), the real lookup falls back too.
-	deviceSize = ioctlDeviceSize
-	filesystems, err = Filesystems()
-	require.NoError(t, err)
-	require.Len(t, filesystems, 1)
-	assert.Equal(t, uint64(1024000), filesystems[0].Size)
-
-	filesystemUsage = func(string) (uint64, uint64, error) { return 100, 900, nil }
-	filesystems, err = Filesystems()
-	require.NoError(t, err)
-	assert.Equal(t, uint64(1000), filesystems[0].Size)
-	assert.Equal(t, uint64(100), filesystems[0].Alloc)
-	assert.False(t, filesystems[0].Raw)
-}
-
-func TestFilesystemsNoBtrfs(t *testing.T) {
-	oldPath := sysfsPath
-	sysfsPath = filepath.Join(t.TempDir(), "missing")
-	t.Cleanup(func() { sysfsPath = oldPath })
-
-	filesystems, err := Filesystems()
-	require.NoError(t, err)
-	assert.Nil(t, filesystems)
-}
-
-func TestIoctlDeviceSizeFailure(t *testing.T) {
-	_, err := ioctlDeviceSize("", 1)
-	require.Error(t, err)
-	_, err = ioctlDeviceSize(t.TempDir(), 1)
-	require.Error(t, err)
-	assert.ErrorIs(t, err, unix.ENOTTY)
-}
-
-func TestMountpointsDecodeEscapes(t *testing.T) {
-	oldMounts := mountsPath
-	mountsPath = filepath.Join(t.TempDir(), "mounts")
-	t.Cleanup(func() { mountsPath = oldMounts })
-	require.NoError(t, os.WriteFile(mountsPath, []byte("/dev/test-btrfs /mnt/my\\040data btrfs rw 0 0\n"), 0o644))
-	assert.Equal(t, "/mnt/my data", mountpointsByDevice()["test-btrfs"])
-}
-
-func TestFilesystemWithoutDevinfo(t *testing.T) {
-	root := t.TempDir()
-	require.NoError(t, os.MkdirAll(filepath.Join(root, "devices", "sda"), 0755))
-	require.NoError(t, os.WriteFile(filepath.Join(root, "devices", "sda", "size"), []byte("1000"), 0644))
-	fs, err := readFilesystem(root, nil)
-	require.NoError(t, err)
-	assert.Equal(t, uint64(512000), fs.Size)
-	assert.True(t, fs.Raw)
-	assert.Equal(t, "UNKNOWN", fs.Health)
-	assert.Empty(t, fs.Devices)
-
-	require.NoError(t, os.MkdirAll(filepath.Join(root, "devinfo", "1"), 0755))
-	fs, err = readFilesystem(root, nil)
-	require.NoError(t, err)
-	assert.Equal(t, "UNKNOWN", fs.Health)
-	require.Len(t, fs.Devices, 1)
-	assert.Equal(t, "UNKNOWN", fs.Devices[0].State)
-
-	// Some older interfaces lack the devices directory too.
-	fs, err = readFilesystem(t.TempDir(), nil)
-	require.NoError(t, err)
-	assert.Equal(t, "UNKNOWN", fs.Health)
-}
-
-func TestLocalBtrfsUsage(t *testing.T) {
-	path := os.Getenv("BESZEL_TEST_BTRFS_MOUNT")
-	if path == "" {
-		t.Skip("set BESZEL_TEST_BTRFS_MOUNT for read-only live validation")
-	}
-	used, available, err := statfsUsage(path)
-	require.NoError(t, err)
-	filesystems, err := Filesystems()
-	require.NoError(t, err)
-	for _, fs := range filesystems {
-		if !fs.Raw && fs.Alloc == used && fs.Size == used+available {
-			t.Logf("pool=%s used=%d available=%d effective_capacity=%d", fs.Name, used, available, fs.Size)
-			return
-		}
-	}
-	t.Fatal("collector did not report the mounted filesystem's usable capacity")
-}
-
-func TestMountID(t *testing.T) {
-	assert.Empty(t, MountID(""))
-	assert.Empty(t, MountID(filepath.Join(t.TempDir(), "missing")))
-	path := os.Getenv("BESZEL_TEST_BTRFS_MOUNT")
-	if path == "" {
-		t.Skip("set BESZEL_TEST_BTRFS_MOUNT for live identity validation")
-	}
-	id := MountID(path)
-	require.NotEmpty(t, id)
-	assert.Equal(t, id, MountID(filepath.Join(path, ".")))
-}
-
-func TestMountinfoUUIDLookup(t *testing.T) {
-	info := `1 0 0:40 /@ /inaccessible ro shared:1 - btrfs /dev/mapper/unavailable rw
-2 0 0:40 /@/docker/hosts /etc/hosts ro - btrfs /dev/mapper/unavailable rw
-3 0 0:40 /@/docker/hostname /etc/hostname ro - btrfs /dev/mapper/unavailable rw
-4 0 0:41 /subvol /extra-filesystems/my\040disk ro master:2 - btrfs /dev/missing rw
-5 0 0:42 / /ext4 ro - ext4 /dev/mapper/unavailable rw
-malformed
-6 0 0:43 / /bad ro - btrfs
-`
-	var calls []string
-	mounts := mountpointsByUUID(info, func(path string) string {
-		calls = append(calls, path)
-		switch path {
-		case "/etc/hosts":
-			return "root-uuid"
-		case "/extra-filesystems/my disk":
-			return "extra-uuid"
-		}
-		return ""
-	})
-	assert.Equal(t, map[string]string{"uuid:root-uuid": "/etc/hosts", "uuid:extra-uuid": "/extra-filesystems/my disk"}, mounts)
-	assert.Equal(t, []string{"/inaccessible", "/etc/hosts", "/extra-filesystems/my disk"}, calls)
-}
-
-func TestDockerFilesystemWithoutDeviceNodes(t *testing.T) {
-	root := t.TempDir()
-	oldSysfs, oldMounts, oldInfo, oldUUID, oldUsage := sysfsPath, mountsPath, mountinfoPath, mountUUID, filesystemUsage
-	t.Cleanup(func() {
-		sysfsPath, mountsPath, mountinfoPath, mountUUID, filesystemUsage = oldSysfs, oldMounts, oldInfo, oldUUID, oldUsage
-	})
-	sysfsPath = filepath.Join(root, "sysfs")
-	mountsPath = filepath.Join(root, "missing-mounts")
-	mountinfoPath = filepath.Join(root, "mountinfo")
-	uuid := "11111111-1111-4111-8111-111111111111"
-	dir := filepath.Join(sysfsPath, uuid)
-	for path, content := range map[string]string{"devices/dm-0/size": "1000", "devinfo/1/missing": "0"} {
-		target := filepath.Join(dir, path)
-		require.NoError(t, os.MkdirAll(filepath.Dir(target), 0755))
-		require.NoError(t, os.WriteFile(target, []byte(content), 0644))
-	}
-	require.NoError(t, os.WriteFile(mountinfoPath, []byte("2 1 0:40 /@/docker/hosts /etc/hosts ro - btrfs /dev/mapper/not-in-container rw\n"), 0644))
-	mountUUID = func(path string) string {
-		if path == "/etc/hosts" {
-			return uuid
-		}
-		return ""
-	}
-	filesystemUsage = func(path string) (uint64, uint64, error) { require.Equal(t, "/etc/hosts", path); return 100, 900, nil }
-	fs, err := Filesystems()
-	require.NoError(t, err)
-	require.Len(t, fs, 1)
-	assert.Equal(t, uuid, fs[0].MountID)
-	assert.Equal(t, "dm-0", fs[0].IODevice)
-	assert.False(t, fs[0].Raw)
-	assert.Equal(t, uint64(1000), fs[0].Size)
-}
-
-func TestLivePoolMountIdentity(t *testing.T) {
-	path := os.Getenv("BESZEL_TEST_BTRFS_MOUNT")
-	if path == "" {
-		t.Skip("set BESZEL_TEST_BTRFS_MOUNT for live validation")
-	}
-	id := MountID(path)
-	require.NotEmpty(t, id)
-	pools, err := Filesystems()
-	require.NoError(t, err)
-	for _, pool := range pools {
-		if pool.UUID != id {
-			continue
-		}
-		assert.Equal(t, id, pool.MountID)
-		assert.False(t, pool.Raw)
-		t.Logf("uuid=%s mount_identity=%s io_device=%s raw=%v", pool.UUID, pool.MountID, pool.IODevice, pool.Raw)
-		return
-	}
-	t.Fatal("mounted Btrfs filesystem was not discovered")
-}

agent/btrfs/btrfs_unsupported.go

@@ -1,11 +0,0 @@
-//go:build !linux
-
-package btrfs
-
-import "errors"
-
-func Filesystems() ([]Filesystem, error) {
-	return nil, errors.ErrUnsupported
-}
-
-func MountID(string) string { return "" }

agent/client.go

@@ -30,11 +30,6 @@ const (
 	wsDeadline = 120 * time.Second
 )
 
-// errNoHubURL is returned when HUB_URL is unset. This is not a failure
-// condition: an agent configured with only a public key runs in SSH-only mode,
-// where the hub dials the agent and no outbound WebSocket client is expected.
-var errNoHubURL = errors.New("HUB_URL environment variable not set")
-
 type caCertFileError struct {
 	err error
 }
@@ -68,7 +63,7 @@ type WebSocketClient struct {
 func newWebSocketClient(agent *Agent) (client *WebSocketClient, err error) {
 	hubURLStr, exists := utils.GetEnv("HUB_URL")
 	if !exists {
-		return nil, errNoHubURL
+		return nil, errors.New("HUB_URL environment variable not set")
 	}
 
 	client = &WebSocketClient{}

agent/client_test.go

@@ -32,28 +32,6 @@ import (
 	"golang.org/x/crypto/ssh"
 )
 
-// TestNewWebSocketClientNoHubURL verifies that an unset HUB_URL returns the
-// errNoHubURL sentinel rather than an opaque error. Callers rely on this to
-// distinguish SSH-only mode -- a supported configuration in which the hub dials
-// the agent -- from an actual misconfiguration.
-func TestNewWebSocketClientNoHubURL(t *testing.T) {
-	agent := createTestAgent(t)
-
-	// t.Setenv registers restoration of the original value; unset afterwards so
-	// GetEnv's LookupEnv reports the variable as absent rather than empty.
-	t.Setenv("BESZEL_AGENT_HUB_URL", "")
-	os.Unsetenv("BESZEL_AGENT_HUB_URL")
-	t.Setenv("HUB_URL", "")
-	os.Unsetenv("HUB_URL")
-	t.Setenv("BESZEL_AGENT_TOKEN", "test-token")
-
-	client, err := newWebSocketClient(agent)
-
-	require.Error(t, err)
-	assert.Nil(t, client)
-	assert.ErrorIs(t, err, errNoHubURL)
-}
-
 // TestNewWebSocketClient tests WebSocket client creation
 func TestNewWebSocketClient(t *testing.T) {
 	agent := createTestAgent(t)

agent/connection_manager.go

@@ -91,15 +91,7 @@ func (c *ConnectionManager) Start(serverOptions ServerOptions) error {
 		if errors.As(err, &caCertErr) {
 			return err
 		}
-		disableSSH, _ := utils.GetEnv("DISABLE_SSH")
-		if errors.Is(err, errNoHubURL) && disableSSH != "true" {
-			// SSH-only mode: the hub dials the agent, so there is nothing to warn
-			// about. With SSH also disabled there is no connection method at all,
-			// so that case still warns.
-			slog.Debug("WebSocket client not configured", "err", err)
-		} else {
-			slog.Warn("Error creating WebSocket client", "err", err)
-		}
+		slog.Warn("Error creating WebSocket client", "err", err)
 	}
 	c.wsClient = wsClient
 
@@ -153,9 +145,7 @@ func (c *ConnectionManager) Start(serverOptions ServerOptions) error {
 //	}
 func (c *ConnectionManager) stop() error {
 	_ = c.agent.StopServer()
-	c.agent.monitorManager.Stop()
 	c.closeWebSocket()
-	c.agent.cleanupSensorShadow()
 	return health.CleanUp()
 }
 

agent/disk.go

@@ -3,7 +3,6 @@ package agent
 import (
 	"context"
 	"log/slog"
-	"math"
 	"os"
 	"path/filepath"
 	"runtime"
@@ -19,11 +18,11 @@ import (
 // fsRegistrationContext holds the shared lookup state needed to resolve a
 // filesystem into the tracked fsStats key and metadata.
 type fsRegistrationContext struct {
-	filesystem     string // device part of optional FILESYSTEM env var
-	filesystemName string // optional custom name from FILESYSTEM=device__name
-	isWindows      bool
-	efPath         string // path to extra filesystems (default "/extra-filesystems")
-	diskIoCounters map[string]disk.IOCountersStat
+	filesystem         string // device part of optional FILESYSTEM env var
+	filesystemName     string // optional custom name from FILESYSTEM=device__name
+	isWindows          bool
+	efPath             string // path to extra filesystems (default "/extra-filesystems")
+	diskIoCounters     map[string]disk.IOCountersStat
 }
 
 // diskDiscovery groups the transient state for a single initializeDiskInfo run so
@@ -326,11 +325,11 @@ func (a *Agent) initializeDiskInfo() {
 	}
 	slog.Debug("Disk I/O", "diskstats", diskIoCounters)
 	ctx := fsRegistrationContext{
-		filesystem:     filesystem,
-		filesystemName: filesystemName,
-		isWindows:      isWindows,
-		diskIoCounters: diskIoCounters,
-		efPath:         "/extra-filesystems",
+		filesystem:         filesystem,
+		filesystemName:     filesystemName,
+		isWindows:          isWindows,
+		diskIoCounters:     diskIoCounters,
+		efPath:             "/extra-filesystems",
 	}
 
 	// Get the appropriate root mount point for this system
@@ -541,8 +540,8 @@ func (a *Agent) initializeDiskIoStats(diskIoCounters map[string]disk.IOCountersS
 	// ZFS datasets have no /proc/diskstats entry, so they are excluded from
 	// I/O tracking instead of warning about a missing device (#1541).
 	var zfsMountpoints map[string]bool
-	if a.storagePoolManager != nil {
-		zfsMountpoints = a.storagePoolManager.ZfsMountpoints()
+	if a.zfsManager != nil {
+		zfsMountpoints = a.zfsManager.ZfsMountpoints()
 	}
 	for device, stats := range a.fsStats {
 		if zfsMountpoints[stats.Mountpoint] {
@@ -575,8 +574,8 @@ func (a *Agent) updateDiskUsage(systemStats *system.Stats) {
 	// ZFS dataset mountpoints use `zfs list` values because statfs(2) reports
 	// dataset-level usage that excludes child datasets (#1541).
 	var zfsUsage map[string]zfsDatasetUsage
-	if a.storagePoolManager != nil {
-		zfsUsage = a.storagePoolManager.DatasetUsage()
+	if a.zfsManager != nil {
+		zfsUsage = a.zfsManager.DatasetUsage()
 	}
 
 	// disk usage
@@ -687,27 +686,25 @@ func (a *Agent) updateDiskIo(cacheTimeMs uint16, systemStats *system.Stats) {
 			// This is the total number of milliseconds spent by all reads (as
 			// measured from __make_request() to end_that_request_last()).
 			// https://www.kernel.org/doc/Documentation/iostats.txt (fields 4, 8)
-			deltaReadTime := ioTimeDelta(d.ReadTime, prev.readTime)
-			deltaWriteTime := ioTimeDelta(d.WriteTime, prev.writeTime)
-			diskReadTime := utils.TwoDecimals(float64(deltaReadTime) / float64(msElapsed) * 100)
-			diskWriteTime := utils.TwoDecimals(float64(deltaWriteTime) / float64(msElapsed) * 100)
+			diskReadTime := utils.TwoDecimals(float64(d.ReadTime-prev.readTime) / float64(msElapsed) * 100)
+			diskWriteTime := utils.TwoDecimals(float64(d.WriteTime-prev.writeTime) / float64(msElapsed) * 100)
 
 			// I/O utilization %: fraction of wall time the device had any I/O in progress (0-100).
-			diskIoUtilPct := utils.TwoDecimals(float64(ioTimeDelta(d.IoTime, prev.ioTime)) / float64(msElapsed) * 100)
+			diskIoUtilPct := utils.TwoDecimals(float64(d.IoTime-prev.ioTime) / float64(msElapsed) * 100)
 
 			// Weighted I/O: queue-depth weighted I/O time, normalized to interval (can exceed 100%).
 			// Linux kernel field 11: incremented by iops_in_progress × ms_since_last_update.
 			// Used to display queue depth. Multipled by 100 to increase accuracy of digit truncation (divided by 100 in UI).
-			diskWeightedIO := utils.TwoDecimals(float64(ioTimeDelta(d.WeightedIO, prev.weightedIO)) / float64(msElapsed) * 100)
+			diskWeightedIO := utils.TwoDecimals(float64(d.WeightedIO-prev.weightedIO) / float64(msElapsed) * 100)
 
 			// r_await / w_await: average time per read/write operation in milliseconds.
 			// Equivalent to r_await and w_await in iostat.
 			var rAwait, wAwait float64
 			if deltaReadCount := d.ReadCount - prev.readCount; deltaReadCount > 0 {
-				rAwait = utils.TwoDecimals(float64(deltaReadTime) / float64(deltaReadCount))
+				rAwait = utils.TwoDecimals(float64(d.ReadTime-prev.readTime) / float64(deltaReadCount))
 			}
 			if deltaWriteCount := d.WriteCount - prev.writeCount; deltaWriteCount > 0 {
-				wAwait = utils.TwoDecimals(float64(deltaWriteTime) / float64(deltaWriteCount))
+				wAwait = utils.TwoDecimals(float64(d.WriteTime-prev.writeTime) / float64(deltaWriteCount))
 			}
 
 			// Update global fsStats baseline for cross-interval correctness
@@ -743,21 +740,6 @@ func (a *Agent) updateDiskIo(cacheTimeMs uint16, systemStats *system.Stats) {
 	}
 }
 
-// ioTimeDelta returns the increase of a cumulative millisecond counter from
-// the disk I/O stats. Linux prints these fields of /proc/diskstats as 32-bit
-// unsigned ints, so they wrap to zero at 2^32. A busy disk reaches that in
-// days for the weighted I/O time. Other platforms report 64-bit counters,
-// so a lower value there is a reset.
-func ioTimeDelta(current, previous uint64) uint64 {
-	if current >= previous {
-		return current - previous
-	}
-	if runtime.GOOS == "linux" && previous <= math.MaxUint32 {
-		return current + (math.MaxUint32 + 1 - previous)
-	}
-	return 0
-}
-
 // getRootMountPoint returns the appropriate root mount point for the system.
 // On Windows it returns the system drive (e.g. "C:").
 // For immutable systems like Fedora Silverblue, it returns /sysroot instead of /

agent/disk_io_linux_test.go

@@ -1,75 +0,0 @@
-//go:build linux
-
-package agent
-
-import (
-	"fmt"
-	"os"
-	"path/filepath"
-	"testing"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/system"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-// Linux prints four millisecond fields of /proc/diskstats as 32-bit unsigned ints:
-// read time, write time, io time and weighted io time. They wrap to zero at 2^32.
-func TestUpdateDiskIoTimeCounterWrap(t *testing.T) {
-	const wrap = uint64(1) << 32
-
-	tests := []struct {
-		name string
-		base uint64 // added to every previous time counter
-	}{
-		{"no wrap", 0},
-		{"32-bit wrap", wrap - 1000},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			// Deltas over 60s: read 300ms / 10 ops, write 400ms / 20 ops,
-			// io time 1200ms, weighted io 3000ms.
-			prev := prevDisk{
-				readBytes:  20000 * 512,
-				writeBytes: 10000 * 512,
-				readTime:   tt.base + 900,
-				writeTime:  tt.base + 700,
-				ioTime:     tt.base + 400,
-				weightedIO: tt.base,
-				readCount:  1000,
-				writeCount: 500,
-				at:         time.Now().Add(-60 * time.Second),
-			}
-			cur := func(v uint64) uint64 { return v % wrap }
-			line := fmt.Sprintf("   8       0 sda %d 0 %d %d %d 0 %d %d 0 %d %d\n",
-				1010, 21200, cur(prev.readTime+300),
-				520, 10400, cur(prev.writeTime+400),
-				cur(prev.ioTime+1200), cur(prev.weightedIO+3000))
-
-			dir := t.TempDir()
-			require.NoError(t, os.WriteFile(filepath.Join(dir, "diskstats"), []byte(line), 0o644))
-			t.Setenv("HOST_PROC", dir)
-			t.Setenv("HOST_SYS", dir)
-			t.Setenv("HOST_DEV", dir)
-			t.Setenv("HOST_RUN", dir)
-
-			fs := &system.FsStats{Root: true}
-			a := &Agent{
-				fsNames:  []string{"sda"},
-				fsStats:  map[string]*system.FsStats{"sda": fs},
-				diskPrev: map[uint16]map[string]prevDisk{60000: {"sda": prev}},
-			}
-			var stats system.Stats
-			a.updateDiskIo(60000, &stats)
-
-			// Same order as DiskIoStats in system.FsStats.
-			want := [6]float64{0.5, 0.67, 2, 30, 20, 5}
-			for i := range want {
-				assert.InDelta(t, want[i], fs.DiskIoStats[i], 0.01, "DiskIoStats[%d]", i)
-				assert.InDelta(t, want[i], stats.DiskIoStats[i], 0.01, "system DiskIoStats[%d]", i)
-			}
-		})
-	}
-}

agent/disk_test.go

@@ -3,9 +3,7 @@
 package agent
 
 import (
-	"math"
 	"os"
-	"runtime"
 	"strings"
 	"testing"
 	"time"
@@ -1043,17 +1041,3 @@ func TestInitializeDiskIoStatsResetsTrackedDevices(t *testing.T) {
 	assert.Equal(t, uint64(50), agent.fsStats["sdb"].TotalRead)
 	assert.Equal(t, uint64(60), agent.fsStats["sdb"].TotalWrite)
 }
-
-func TestIoTimeDelta(t *testing.T) {
-	assert.Equal(t, uint64(300), ioTimeDelta(1200, 900))
-
-	// A lower value is a 32-bit wrap only on Linux. Other platforms
-	// report 64-bit counters, so there it is a reset.
-	var want uint64
-	if runtime.GOOS == "linux" {
-		want = 1200
-	}
-	assert.Equal(t, want, ioTimeDelta(200, math.MaxUint32+1-1000))
-
-	assert.Equal(t, uint64(0), ioTimeDelta(200, math.MaxUint32+1000))
-}

agent/disk_zfs_test.go

@@ -4,7 +4,6 @@ package agent
 
 import (
 	"testing"
-	"time"
 
 	"github.com/henrygd/beszel/agent/zfs"
 	"github.com/henrygd/beszel/internal/entities/system"
@@ -17,8 +16,8 @@ import (
 // is a ZFS dataset reports `zfs list` usage (which includes child datasets)
 // instead of the dataset-scoped statfs values (#1541).
 func TestUpdateDiskUsageZfsMountpoint(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) {
+	zm := &ZfsManager{}
+	zm.datasetsFn = func() ([]zfs.Dataset, error) {
 		return []zfs.Dataset{
 			{Name: "tank", Used: 12000000000000, Avail: 11999000000000, Mountpoint: "/tank"},
 		}, nil
@@ -27,7 +26,7 @@ func TestUpdateDiskUsageZfsMountpoint(t *testing.T) {
 		fsStats: map[string]*system.FsStats{
 			"tank": {Root: false, Mountpoint: "/tank"},
 		},
-		storagePoolManager: zm,
+		zfsManager: zm,
 	}
 
 	var stats system.Stats
@@ -44,8 +43,8 @@ func TestUpdateDiskUsageZfsMountpoint(t *testing.T) {
 // TestUpdateDiskUsageZfsRootPopulatesSystemStats verifies the root disk values
 // are derived from ZFS usage when the root mountpoint is a ZFS dataset.
 func TestUpdateDiskUsageZfsRootPopulatesSystemStats(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) {
+	zm := &ZfsManager{}
+	zm.datasetsFn = func() ([]zfs.Dataset, error) {
 		return []zfs.Dataset{
 			{Name: "rpool/ROOT/pve-1", Used: 900000000000, Avail: 300000000000, Mountpoint: "/"},
 		}, nil
@@ -54,7 +53,7 @@ func TestUpdateDiskUsageZfsRootPopulatesSystemStats(t *testing.T) {
 		fsStats: map[string]*system.FsStats{
 			"rpool/ROOT/pve-1": {Root: true, Mountpoint: "/"},
 		},
-		storagePoolManager: zm,
+		zfsManager: zm,
 	}
 
 	var stats system.Stats
@@ -86,8 +85,8 @@ func TestUpdateDiskUsageWithoutZfsManager(t *testing.T) {
 // TestInitializeDiskIoStatsSkipsZfsMountpoints verifies ZFS filesystems are
 // excluded from diskstats I/O tracking instead of warning about a missing device.
 func TestInitializeDiskIoStatsSkipsZfsMountpoints(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) {
+	zm := &ZfsManager{}
+	zm.datasetsFn = func() ([]zfs.Dataset, error) {
 		return []zfs.Dataset{{Name: "tank", Mountpoint: "/tank"}}, nil
 	}
 	agent := &Agent{
@@ -95,8 +94,8 @@ func TestInitializeDiskIoStatsSkipsZfsMountpoints(t *testing.T) {
 			"tank": {Root: false, Mountpoint: "/tank"},
 			"sda1": {Root: false, Mountpoint: "/mnt/data"},
 		},
-		storagePoolManager: zm,
-		diskPrev:           make(map[uint16]map[string]prevDisk),
+		zfsManager: zm,
+		diskPrev:   make(map[uint16]map[string]prevDisk),
 	}
 
 	agent.initializeDiskIoStats(map[string]disk.IOCountersStat{

agent/docker.go

@@ -65,15 +65,10 @@ type dockerManager struct {
 	dockerVersionChecked bool                        // Whether a version probe has completed successfully
 	isWindows            bool                        // Whether the Docker Engine API is running on Windows
 	buf                  *bytes.Buffer               // Buffer to store and read response bodies
+	apiStats             *container.ApiStats         // Reusable API stats object
 	excludeContainers    []string                    // Patterns to exclude containers by name
 	usingPodman          bool                        // Whether the Docker Engine API is running on Podman
 
-	registryClient       *http.Client                  // Client for registry requests; nil uses a client with a 10-second timeout
-	imageUpdatesDisabled bool                          // Whether image update checks are disabled by configuration
-	imageUpdatesMutex    sync.RWMutex                  // Protects imageUpdates, its entries, and imageUpdatesRunning
-	imageUpdates         map[string]*imageUpdateStatus // Shared update status keyed by normalized image reference
-	imageUpdatesRunning  bool                          // Whether a background image-update batch is in progress
-
 	// Cache-time-aware tracking for CPU stats (similar to cpu.go)
 	// Maps cache time intervals to container-specific CPU usage tracking
 	lastCpuContainer map[uint16]map[string]uint64    // cacheTimeMs -> containerId -> last cpu container usage
@@ -166,9 +161,6 @@ func (dm *dockerManager) getDockerStats(cacheTimeMs uint16) ([]*container.Stats,
 		clear(dm.validIds)
 	}
 
-	// Only schedule auxiliary work here; metrics never wait for image discovery.
-	dm.refreshImageUpdates(dm.apiContainerList, time.Now())
-
 	var failedContainers []*container.ApiInfo
 
 	for _, ctr := range dm.apiContainerList {
@@ -514,17 +506,6 @@ func (dm *dockerManager) updateContainerStats(ctr *container.ApiInfo, cacheTimeM
 		}
 	}
 
-	// Read and decode the response before locking shared stats to avoid blocking
-	defer resp.Body.Close()
-	if resp.StatusCode != http.StatusOK {
-		return fmt.Errorf("container stats request failed: %s", resp.Status)
-	}
-	res := &container.ApiStats{}
-	if err := json.NewDecoder(resp.Body).Decode(res); err != nil {
-		return err
-	}
-	updateAvailable := dm.cachedImageUpdate(ctr.Image)
-
 	dm.containerStatsMutex.Lock()
 	defer dm.containerStatsMutex.Unlock()
 
@@ -539,9 +520,6 @@ func (dm *dockerManager) updateContainerStats(ctr *container.ApiInfo, cacheTimeM
 	stats.Status = statusText
 	stats.Health = health
 
-	stats.Image = ctr.Image
-	stats.UpdateAvailable = updateAvailable
-
 	if len(ctr.Ports) > 0 {
 		stats.Ports = convertContainerPortsToString(ctr)
 	}
@@ -554,6 +532,12 @@ func (dm *dockerManager) updateContainerStats(ctr *container.ApiInfo, cacheTimeM
 	stats.NetworkSent = 0
 	stats.NetworkRecv = 0
 
+	res := dm.apiStats
+	res.Networks = nil
+	if err := dm.decode(resp, res); err != nil {
+		return err
+	}
+
 	// Initialize CPU tracking for this cache time interval
 	dm.initializeCpuTracking(cacheTimeMs)
 
@@ -689,8 +673,6 @@ func newDockerManager(agent *Agent) *dockerManager {
 		userAgent: "Docker-Client/",
 	}
 
-	dockerImageCheck, _ := utils.GetEnv("DOCKER_IMAGE_CHECK")
-
 	// Read container exclusion patterns from environment variable
 	var excludeContainers []string
 	if excludeStr, set := utils.GetEnv("EXCLUDE_CONTAINERS"); set && excludeStr != "" {
@@ -710,11 +692,11 @@ func newDockerManager(agent *Agent) *dockerManager {
 			Timeout:   timeout,
 			Transport: userAgentTransport,
 		},
-		containerStatsMap:    make(map[string]*container.Stats),
-		sem:                  make(chan struct{}, 5),
-		apiContainerList:     []*container.ApiInfo{},
-		excludeContainers:    excludeContainers,
-		imageUpdatesDisabled: dockerImageCheck == "false",
+		containerStatsMap: make(map[string]*container.Stats),
+		sem:               make(chan struct{}, 5),
+		apiContainerList:  []*container.ApiInfo{},
+		apiStats:          &container.ApiStats{},
+		excludeContainers: excludeContainers,
 
 		// Initialize cache-time-aware tracking structures
 		lastCpuContainer:    make(map[uint16]map[string]uint64),

agent/docker_image_updates.go

@@ -1,108 +0,0 @@
-package agent
-
-import (
-	"log/slog"
-	"sync"
-	"time"
-
-	"github.com/distribution/reference"
-	"github.com/henrygd/beszel/internal/entities/container"
-)
-
-const imageUpdateInterval = time.Hour
-
-type imageUpdateStatus struct {
-	available bool
-	checkedAt time.Time
-}
-
-func normalizedImageReference(image string) string {
-	named, err := reference.ParseNormalizedNamed(image)
-	if err != nil {
-		return ""
-	}
-	// Digest-pinned references cannot move to a new version.
-	if _, pinned := named.(reference.Digested); pinned {
-		return ""
-	}
-	return reference.TagNameOnly(named).String()
-}
-
-// refreshImageUpdates starts at most one background batch. Neither its network
-// work nor its completion is part of the container metrics wait group.
-func (dm *dockerManager) refreshImageUpdates(containers []*container.ApiInfo, now time.Time) {
-	if dm.imageUpdatesDisabled {
-		return
-	}
-	dm.imageUpdatesMutex.Lock()
-	defer dm.imageUpdatesMutex.Unlock()
-	if dm.imageUpdatesRunning {
-		return
-	}
-	if dm.imageUpdates == nil {
-		dm.imageUpdates = make(map[string]*imageUpdateStatus)
-	}
-	active := make(map[string]struct{}, len(containers))
-	pending := make(map[string]*imageUpdateStatus)
-	for _, ctr := range containers {
-		if len(ctr.Names) > 0 && dm.shouldExcludeContainer(ctr.Names[0][1:]) {
-			continue
-		}
-		key := normalizedImageReference(ctr.Image)
-		if key == "" {
-			continue
-		}
-		active[key] = struct{}{}
-		entry := dm.imageUpdates[key]
-		if entry == nil {
-			entry = &imageUpdateStatus{}
-			dm.imageUpdates[key] = entry
-		}
-		if entry.checkedAt.IsZero() || now.Sub(entry.checkedAt) >= imageUpdateInterval {
-			pending[key] = entry
-		}
-	}
-	for key := range dm.imageUpdates {
-		if _, ok := active[key]; !ok {
-			delete(dm.imageUpdates, key)
-		}
-	}
-	if len(pending) == 0 {
-		return
-	}
-	dm.imageUpdatesRunning = true
-	go func() {
-		// Limit auxiliary requests even on hosts running many different images.
-		sem := make(chan struct{}, 2)
-		var wg sync.WaitGroup
-		for key, entry := range pending {
-			sem <- struct{}{}
-			wg.Add(1)
-			go func() {
-				defer wg.Done()
-				defer func() { <-sem }()
-				available, err := dm.checkImageUpdate(key)
-				if err != nil {
-					available = false
-					slog.Debug("Image update check failed", "image", key, "err", err)
-				}
-				dm.imageUpdatesMutex.Lock()
-				entry.available = available
-				entry.checkedAt = time.Now()
-				dm.imageUpdatesMutex.Unlock()
-			}()
-		}
-		wg.Wait()
-		dm.imageUpdatesMutex.Lock()
-		dm.imageUpdatesRunning = false
-		dm.imageUpdatesMutex.Unlock()
-	}()
-}
-
-func (dm *dockerManager) cachedImageUpdate(image string) bool {
-	key := normalizedImageReference(image)
-	dm.imageUpdatesMutex.RLock()
-	defer dm.imageUpdatesMutex.RUnlock()
-	entry := dm.imageUpdates[key]
-	return entry != nil && entry.available
-}

agent/docker_image_updates_test.go

@@ -1,248 +0,0 @@
-//go:build testing
-
-package agent
-
-import (
-	"encoding/json"
-	"fmt"
-	"github.com/fxamacker/cbor/v2"
-	"io"
-	"net/http"
-	"net/http/httptest"
-	"strings"
-	"sync/atomic"
-	"testing"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/container"
-	"github.com/stretchr/testify/require"
-)
-
-func waitForImageUpdates(t *testing.T, dm *dockerManager) {
-	t.Helper()
-	require.Eventually(t, func() bool {
-		dm.imageUpdatesMutex.RLock()
-		defer dm.imageUpdatesMutex.RUnlock()
-		return !dm.imageUpdatesRunning
-	}, time.Second*3, time.Millisecond)
-}
-
-func TestDisableDockerImageUpdateCheck(t *testing.T) {
-	t.Setenv("BESZEL_AGENT_DOCKER_IMAGE_CHECK", "false")
-	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if r.URL.Path == "/version" {
-			fmt.Fprint(w, `{"Version":"25.0.0"}`)
-			return
-		}
-		http.NotFound(w, r)
-	}))
-	defer server.Close()
-	t.Setenv("BESZEL_AGENT_DOCKER_HOST", server.URL)
-
-	dm := newDockerManager(nil)
-	require.True(t, dm.imageUpdatesDisabled)
-	dm.registryClient = &http.Client{Transport: roundTripFunc(func(*http.Request) (*http.Response, error) {
-		t.Fatal("disabled image update check made a registry request")
-		return nil, nil
-	})}
-	dm.refreshImageUpdates([]*container.ApiInfo{{Image: "nginx", Names: []string{"/nginx"}}}, time.Now())
-	require.False(t, dm.imageUpdatesRunning)
-	require.Nil(t, dm.imageUpdates)
-}
-
-func TestImageUpdateCacheAndStats(t *testing.T) {
-	local := "sha256:" + strings.Repeat("a", 64)
-	remote := "sha256:" + strings.Repeat("b", 64)
-	var inspections, lookups atomic.Int32
-	var fail atomic.Bool
-	var upToDate atomic.Bool
-	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		switch {
-		case strings.HasPrefix(r.URL.Path, "/images/"):
-			inspections.Add(1)
-			fmt.Fprintf(w, `{"RepoDigests":["docker.io/library/nginx@%s"]}`, local)
-		case r.URL.Path == "/containers/json":
-			fmt.Fprint(w, `[{"Id":"aaaaaaaaaaaa","Names":["/one"],"Image":"nginx","Status":"Up 2 hours"},{"Id":"bbbbbbbbbbbb","Names":["/two"],"Image":"docker.io/library/nginx:latest","Status":"Up 2 hours"}]`)
-		case strings.Contains(r.URL.Path, "/stats"):
-			fmt.Fprint(w, `{"memory_stats":{"usage":1048576},"cpu_stats":{},"networks":{}}`)
-		default:
-			http.NotFound(w, r)
-		}
-	}))
-	defer server.Close()
-	dm := newDockerManagerForVersionTest(server)
-	dm.dockerVersionChecked = true
-	dm.registryClient = &http.Client{Timeout: time.Second, Transport: roundTripFunc(func(r *http.Request) (*http.Response, error) {
-		if fail.Load() {
-			return nil, fmt.Errorf("registry unavailable")
-		}
-		response := &http.Response{StatusCode: 200, Header: make(http.Header), Body: io.NopCloser(strings.NewReader(`{"token":"test"}`))}
-		if r.Method == http.MethodHead {
-			lookups.Add(1)
-			digest := remote
-			if upToDate.Load() {
-				digest = local
-			}
-			response.Header.Set("Docker-Content-Digest", digest)
-		}
-		return response, nil
-	})}
-	stats, err := dm.getDockerStats(defaultCacheTimeMs)
-	require.NoError(t, err)
-	require.Len(t, stats, 2)
-	waitForImageUpdates(t, dm)
-	require.EqualValues(t, 1, lookups.Load())
-	require.EqualValues(t, 1, inspections.Load())
-	stats, err = dm.getDockerStats(defaultCacheTimeMs)
-	require.NoError(t, err)
-	for _, stat := range stats {
-		require.True(t, stat.UpdateAvailable)
-		if stat.Id == "aaaaaaaaaaaa" {
-			require.Equal(t, "nginx", stat.Image)
-		} else {
-			require.Equal(t, "docker.io/library/nginx:latest", stat.Image)
-		}
-	}
-	require.EqualValues(t, 1, lookups.Load())
-
-	expire := func() {
-		dm.imageUpdatesMutex.Lock()
-		dm.imageUpdates["docker.io/library/nginx:latest"].checkedAt = time.Now().Add(-imageUpdateInterval)
-		dm.imageUpdatesMutex.Unlock()
-	}
-	upToDate.Store(true)
-	expire()
-	_, err = dm.getDockerStats(defaultCacheTimeMs)
-	require.NoError(t, err)
-	waitForImageUpdates(t, dm)
-	require.EqualValues(t, 2, lookups.Load())
-	require.False(t, dm.cachedImageUpdate("nginx:latest"))
-
-	// An expired positive result is cleared on failure, and the failure itself
-	// is cached so realtime stats do not retry a broken registry every second.
-	dm.imageUpdatesMutex.Lock()
-	dm.imageUpdates["docker.io/library/nginx:latest"].available = true
-	dm.imageUpdatesMutex.Unlock()
-	fail.Store(true)
-	expire()
-	_, err = dm.getDockerStats(defaultCacheTimeMs)
-	require.NoError(t, err)
-	waitForImageUpdates(t, dm)
-	failedInspections := inspections.Load()
-	stats, err = dm.getDockerStats(defaultCacheTimeMs)
-	require.NoError(t, err)
-	require.Len(t, stats, 2)
-	require.Equal(t, failedInspections, inspections.Load())
-	for _, stat := range stats {
-		require.False(t, stat.UpdateAvailable)
-		require.Equal(t, 1.0, stat.Mem)
-	}
-}
-
-func TestImageDiscoveryDoesNotBlockStats(t *testing.T) {
-	started := make(chan struct{}, 1)
-	release := make(chan struct{})
-	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if strings.HasPrefix(r.URL.Path, "/images/") {
-			fmt.Fprintf(w, `{"RepoDigests":["example.com/app@sha256:%s"]}`, strings.Repeat("a", 64))
-		} else {
-			fmt.Fprint(w, `{"memory_stats":{"usage":1048576}}`)
-		}
-	}))
-	defer server.Close()
-	dm := newDockerManagerForVersionTest(server)
-	defer func() { close(release); waitForImageUpdates(t, dm) }()
-	dm.registryClient = &http.Client{Transport: roundTripFunc(func(r *http.Request) (*http.Response, error) {
-		started <- struct{}{}
-		<-release
-		return nil, fmt.Errorf("timeout")
-	})}
-	ctr := &container.ApiInfo{IdShort: "aaaaaaaaaaaa", Image: "example.com/app", Names: []string{"/one"}}
-	dm.refreshImageUpdates([]*container.ApiInfo{ctr}, time.Now())
-	select {
-	case <-started:
-	case <-time.After(3 * time.Second):
-		t.Fatal("check did not start")
-	}
-	done := make(chan error, 1)
-	go func() { done <- dm.updateContainerStats(ctr, defaultCacheTimeMs) }()
-	select {
-	case err := <-done:
-		require.NoError(t, err)
-	case <-time.After(time.Second):
-		t.Fatal("registry blocked stats")
-	}
-	dm.imageUpdatesMutex.RLock()
-	require.True(t, dm.imageUpdatesRunning)
-	dm.imageUpdatesMutex.RUnlock()
-}
-
-func TestNormalizeImageUpdateReferences(t *testing.T) {
-	require.Equal(t, normalizedImageReference("nginx"), normalizedImageReference("docker.io/library/nginx:latest"))
-	require.Empty(t, normalizedImageReference("bad reference"))
-	require.Empty(t, normalizedImageReference("nginx@sha256:"+strings.Repeat("a", 64)))
-}
-
-// A stats request can return headers promptly and then stall while reading its
-// body. The stats-map mutex must remain available during that read.
-func TestStatsResponseBodyDoesNotHoldStatsLock(t *testing.T) {
-	started := make(chan struct{})
-	release := make(chan struct{})
-	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		w.WriteHeader(http.StatusOK)
-		w.(http.Flusher).Flush()
-		close(started)
-		<-release
-		fmt.Fprint(w, `{"memory_stats":{"usage":1048576}}`)
-	}))
-	defer server.Close()
-	dm := newDockerManagerForVersionTest(server)
-	done := make(chan error, 1)
-	go func() {
-		done <- dm.updateContainerStats(&container.ApiInfo{IdShort: "aaaaaaaaaaaa", Names: []string{"/one"}, Image: "nginx"}, defaultCacheTimeMs)
-	}()
-	<-started
-	locked := make(chan struct{})
-	go func() { dm.containerStatsMutex.Lock(); dm.containerStatsMutex.Unlock(); close(locked) }()
-	select {
-	case <-locked:
-	case <-time.After(time.Second):
-		close(release)
-		<-done
-		t.Fatal("Docker response body held the stats mutex")
-	}
-	close(release)
-	require.NoError(t, <-done)
-}
-
-func TestImageUpdateStatsEncoding(t *testing.T) {
-	original := container.Stats{Image: "nginx:latest", UpdateAvailable: true}
-	encoded, err := cbor.Marshal(original)
-	require.NoError(t, err)
-	var fields map[int]any
-	require.NoError(t, cbor.Unmarshal(encoded, &fields))
-	require.Equal(t, true, fields[11])
-	require.Equal(t, "nginx:latest", fields[8])
-	var decoded container.Stats
-	require.NoError(t, cbor.Unmarshal(encoded, &decoded))
-	require.True(t, decoded.UpdateAvailable)
-	require.Equal(t, original.Image, decoded.Image)
-	encoded, err = json.Marshal(original)
-	require.NoError(t, err)
-	require.Contains(t, string(encoded), `"u":true`)
-}
-
-func TestImageUpdateCacheExpiryBoundaryAndPruning(t *testing.T) {
-	now := time.Now()
-	key := normalizedImageReference("nginx")
-	dm := &dockerManager{imageUpdates: map[string]*imageUpdateStatus{
-		key:                           {available: true, checkedAt: now},
-		"unused.example/image:latest": {checkedAt: now},
-	}}
-	dm.refreshImageUpdates([]*container.ApiInfo{{Image: "nginx"}}, now.Add(imageUpdateInterval-time.Nanosecond))
-	require.False(t, dm.imageUpdatesRunning)
-	require.Len(t, dm.imageUpdates, 1)
-	require.True(t, dm.cachedImageUpdate("nginx:latest"))
-	dm.refreshImageUpdates(nil, now)
-	require.Empty(t, dm.imageUpdates)
-}

agent/docker_registry.go

@@ -1,224 +0,0 @@
-package agent
-
-import (
-	_ "crypto/sha256"
-	"encoding/json"
-	"fmt"
-	"net/http"
-	"net/url"
-	"slices"
-	"strings"
-	"time"
-
-	"github.com/distribution/reference"
-	"github.com/opencontainers/go-digest"
-)
-
-const imageRegistryTimeout = 10 * time.Second
-
-const imageManifestAccept = "application/vnd.docker.distribution.manifest.list.v2+json, " +
-	"application/vnd.docker.distribution.manifest.v2+json, " +
-	"application/vnd.oci.image.manifest.v1+json, " +
-	"application/vnd.oci.image.index.v1+json"
-
-// checkImageUpdate compares the digest recorded by Docker for image with the
-// digest currently advertised by its registry. A digest-pinned reference is
-// immutable and therefore never has an update available.
-func (dm *dockerManager) checkImageUpdate(image string) (bool, error) {
-	named, err := reference.ParseNormalizedNamed(image)
-	if err != nil {
-		return false, fmt.Errorf("parse image reference %q: %w", image, err)
-	}
-	if _, pinned := named.(reference.Digested); pinned {
-		return false, nil
-	}
-	named = reference.TagNameOnly(named)
-
-	registry := reference.Domain(named)
-	repository := reference.Path(named)
-	tag := named.(reference.Tagged).Tag()
-
-	localDigests, err := dm.inspectImageDigests(image, registry, repository)
-	if err != nil {
-		return false, err
-	}
-
-	remoteDigest, err := dm.registryImageDigest(registry, repository, tag)
-	if err != nil {
-		return false, err
-	}
-
-	return !slices.Contains(localDigests, remoteDigest), nil
-}
-
-// inspectImageDigests reads Docker's image metadata without using dm.decode.
-// The checker runs in the image-discovery goroutine, so it must not hold any
-// of the container statistics locks while waiting on the Docker API.
-func (dm *dockerManager) inspectImageDigests(image, registry, repository string) ([]string, error) {
-	if dm.client == nil {
-		return nil, fmt.Errorf("inspect image %q: Docker client is unavailable", image)
-	}
-
-	endpoint := "http://localhost/images/" + url.PathEscape(image) + "/json"
-	resp, err := dm.client.Get(endpoint)
-	if err != nil {
-		return nil, fmt.Errorf("inspect image %q: %w", image, err)
-	}
-	defer resp.Body.Close()
-	if resp.StatusCode != http.StatusOK {
-		return nil, fmt.Errorf("inspect image %q failed: %s", image, responseStatus(resp))
-	}
-
-	var inspect struct {
-		RepoDigests []string `json:"RepoDigests"`
-	}
-	if err := json.NewDecoder(resp.Body).Decode(&inspect); err != nil {
-		return nil, fmt.Errorf("decode image inspect %q: %w", image, err)
-	}
-	if len(inspect.RepoDigests) == 0 {
-		return nil, fmt.Errorf("inspect image %q returned no repository digests", image)
-	}
-
-	localDigests := matchingRepositoryDigests(inspect.RepoDigests, registry, repository)
-	if len(localDigests) == 0 {
-		return nil, fmt.Errorf("inspect image %q returned no valid digest for %s/%s", image, registry, repository)
-	}
-	return localDigests, nil
-}
-
-// matchingRepositoryDigests returns all valid digests belonging to the requested
-// repository. Container engines can return both index and platform manifest digests for one
-// local image, in either order.
-func matchingRepositoryDigests(repoDigests []string, registry, repository string) []string {
-	var digests []string
-	for _, repoDigest := range repoDigests {
-		repoDigest = strings.TrimSpace(repoDigest)
-		at := strings.LastIndexByte(repoDigest, '@')
-		if at <= 0 || at == len(repoDigest)-1 || strings.Contains(repoDigest[:at], "@") {
-			continue
-		}
-
-		repoRef, err := reference.ParseNormalizedNamed(repoDigest[:at])
-		if err != nil || reference.Path(repoRef) != repository || !sameRegistry(reference.Domain(repoRef), registry) {
-			continue
-		}
-		if _, hasTag := repoRef.(reference.Tagged); hasTag {
-			continue
-		}
-
-		d, err := digest.Parse(repoDigest[at+1:])
-		if err != nil {
-			continue
-		}
-		digests = append(digests, d.String())
-	}
-	return digests
-}
-
-func sameRegistry(left, right string) bool {
-	left = canonicalRegistry(left)
-	right = canonicalRegistry(right)
-	return left == right ||
-		(left == "ghcr.io" && right == "lscr.io") ||
-		(left == "lscr.io" && right == "ghcr.io")
-}
-
-func canonicalRegistry(registry string) string {
-	if registry == "index.docker.io" {
-		return "docker.io"
-	}
-	return registry
-}
-
-func (dm *dockerManager) registryImageDigest(registry, repository, tag string) (string, error) {
-	client := dm.registryClient
-	if client == nil {
-		client = &http.Client{Timeout: imageRegistryTimeout}
-	}
-
-	token, err := dm.registryToken(client, registry, repository)
-	if err != nil {
-		return "", err
-	}
-
-	host := registry
-	if registry == "docker.io" {
-		host = "registry-1.docker.io"
-	}
-	manifestURL := "https://" + host + "/v2/" + repository + "/manifests/" + url.PathEscape(tag)
-	req, err := http.NewRequest(http.MethodHead, manifestURL, nil)
-	if err != nil {
-		return "", fmt.Errorf("create manifest request: %w", err)
-	}
-	req.Header.Set("Accept", imageManifestAccept)
-	if token != "" {
-		req.Header.Set("Authorization", "Bearer "+token)
-	}
-
-	resp, err := client.Do(req)
-	if err != nil {
-		return "", fmt.Errorf("fetch manifest %s:%s: %w", registry, repository, err)
-	}
-	defer resp.Body.Close()
-	if resp.StatusCode != http.StatusOK {
-		return "", fmt.Errorf("manifest request for %s:%s failed: %s", repository, tag, responseStatus(resp))
-	}
-
-	remote := strings.TrimSpace(resp.Header.Get("Docker-Content-Digest"))
-	d, err := digest.Parse(remote)
-	if err != nil {
-		return "", fmt.Errorf("manifest request for %s:%s returned invalid digest: %w", repository, tag, err)
-	}
-	return d.String(), nil
-}
-
-func (dm *dockerManager) registryToken(client *http.Client, registry, repository string) (string, error) {
-	var authURL string
-	switch registry {
-	case "docker.io":
-		authURL = "https://auth.docker.io/token?service=registry.docker.io&scope=" + url.QueryEscape("repository:"+repository+":pull")
-	case "ghcr.io", "lscr.io":
-		// lscr.io is the LinuxServer alias for its GHCR-backed images.
-		authURL = "https://ghcr.io/token?service=ghcr.io&scope=" + url.QueryEscape("repository:"+repository+":pull")
-	default:
-		// Anonymous registries remain supported, as they were before the
-		// authenticated Docker Hub and GHCR paths were added.
-		return "", nil
-	}
-
-	req, err := http.NewRequest(http.MethodGet, authURL, nil)
-	if err != nil {
-		return "", fmt.Errorf("create registry auth request: %w", err)
-	}
-	resp, err := client.Do(req)
-	if err != nil {
-		return "", fmt.Errorf("fetch registry auth token for %s: %w", repository, err)
-	}
-	defer resp.Body.Close()
-	if resp.StatusCode != http.StatusOK {
-		return "", fmt.Errorf("registry auth request for %s failed: %s", repository, responseStatus(resp))
-	}
-
-	var tokenResponse struct {
-		Token       string `json:"token"`
-		AccessToken string `json:"access_token"`
-	}
-	if err := json.NewDecoder(resp.Body).Decode(&tokenResponse); err != nil {
-		return "", fmt.Errorf("decode registry auth response for %s: %w", repository, err)
-	}
-	token := strings.TrimSpace(tokenResponse.Token)
-	if token == "" {
-		token = strings.TrimSpace(tokenResponse.AccessToken)
-	}
-	if token == "" {
-		return "", fmt.Errorf("registry auth response for %s contained no token", repository)
-	}
-	return token, nil
-}
-
-func responseStatus(resp *http.Response) string {
-	if resp.Status != "" {
-		return resp.Status
-	}
-	return http.StatusText(resp.StatusCode)
-}

agent/docker_registry_test.go

@@ -1,243 +0,0 @@
-//go:build testing
-
-package agent
-
-import (
-	"fmt"
-	"io"
-	"net/http"
-	"net/http/httptest"
-	"strings"
-	"sync/atomic"
-	"testing"
-
-	"github.com/stretchr/testify/require"
-)
-
-type registryTransportFunc func(*http.Request) (*http.Response, error)
-
-func (fn registryTransportFunc) RoundTrip(req *http.Request) (*http.Response, error) {
-	return fn(req)
-}
-
-func registryResponse(status int, body string) *http.Response {
-	return &http.Response{
-		StatusCode: status,
-		Status:     fmt.Sprintf("%d %s", status, http.StatusText(status)),
-		Header:     make(http.Header),
-		Body:       io.NopCloser(strings.NewReader(body)),
-	}
-}
-
-func registryDigest(fill byte) string {
-	return "sha256:" + strings.Repeat(string(fill), 64)
-}
-
-func newRegistryChecker(t *testing.T, inspectBody string, transport http.RoundTripper) *dockerManager {
-	t.Helper()
-	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if strings.HasPrefix(r.URL.Path, "/images/") {
-			w.Header().Set("Content-Type", "application/json")
-			_, _ = io.WriteString(w, inspectBody)
-			return
-		}
-		http.NotFound(w, r)
-	}))
-	t.Cleanup(server.Close)
-
-	return &dockerManager{
-		client:         newDockerManagerForVersionTest(server).client,
-		registryClient: &http.Client{Transport: transport},
-	}
-}
-
-func TestCheckImageUpdateUsesInspectAndManifestDigests(t *testing.T) {
-	local := registryDigest('a')
-	remote := registryDigest('b')
-	var authCalls, manifestCalls atomic.Int32
-	dm := newRegistryChecker(t, fmt.Sprintf(`{"RepoDigests":["docker.io/library/alpine@%s"]}`, local), registryTransportFunc(func(req *http.Request) (*http.Response, error) {
-		switch {
-		case req.Method == http.MethodGet && req.URL.Host == "auth.docker.io":
-			authCalls.Add(1)
-			require.Equal(t, "/token", req.URL.Path)
-			return registryResponse(http.StatusOK, `{"token":"test-token"}`), nil
-		case req.Method == http.MethodHead && req.URL.Host == "registry-1.docker.io":
-			manifestCalls.Add(1)
-			require.Equal(t, "/v2/library/alpine/manifests/latest", req.URL.Path)
-			require.Equal(t, "Bearer test-token", req.Header.Get("Authorization"))
-			resp := registryResponse(http.StatusOK, "")
-			resp.Header.Set("Docker-Content-Digest", remote)
-			return resp, nil
-		default:
-			return registryResponse(http.StatusNotFound, ""), nil
-		}
-	}))
-
-	available, err := dm.checkImageUpdate("alpine")
-	require.NoError(t, err)
-	require.True(t, available)
-	require.EqualValues(t, 1, authCalls.Load())
-	require.EqualValues(t, 1, manifestCalls.Load())
-}
-
-func TestCheckImageUpdateMatchesAnyRepositoryDigest(t *testing.T) {
-	platform := registryDigest('a')
-	index := registryDigest('b')
-	other := registryDigest('c')
-
-	for _, test := range []struct {
-		name      string
-		digests   []string
-		remote    string
-		available bool
-	}{
-		{name: "platform then index, remote index", digests: []string{platform, index}, remote: index},
-		{name: "index then platform, remote index", digests: []string{index, platform}, remote: index},
-		{name: "platform then index, remote platform", digests: []string{platform, index}, remote: platform},
-		{name: "index then platform, remote platform", digests: []string{index, platform}, remote: platform},
-		{name: "neither matches", digests: []string{platform, index}, remote: other, available: true},
-	} {
-		t.Run(test.name, func(t *testing.T) {
-			inspect := fmt.Sprintf(`{"RepoDigests":["docker.io/library/busybox@%s","docker.io/library/alpine@%s","docker.io/library/alpine@sha256:invalid","docker.io/library/alpine@%s"]}`, test.remote, test.digests[0], test.digests[1])
-			var manifestCalls atomic.Int32
-			dm := newRegistryChecker(t, inspect, registryTransportFunc(func(req *http.Request) (*http.Response, error) {
-				if req.Method == http.MethodGet {
-					return registryResponse(http.StatusOK, `{"token":"test"}`), nil
-				}
-				manifestCalls.Add(1)
-				require.Equal(t, http.MethodHead, req.Method)
-				resp := registryResponse(http.StatusOK, "")
-				resp.Header.Set("Docker-Content-Digest", test.remote)
-				return resp, nil
-			}))
-
-			available, err := dm.checkImageUpdate("alpine")
-			require.NoError(t, err)
-			require.Equal(t, test.available, available)
-			require.EqualValues(t, 1, manifestCalls.Load())
-		})
-	}
-}
-
-func TestCheckImageUpdateReportsUnknownInspectState(t *testing.T) {
-	for _, test := range []struct {
-		name string
-		body string
-	}{
-		{name: "missing field", body: `{}`},
-		{name: "empty field", body: `{"RepoDigests":[]}`},
-		{name: "malformed reference", body: `{"RepoDigests":["not-a-repo-digest"]}`},
-		{name: "wrong repository", body: `{"RepoDigests":["docker.io/library/busybox@` + registryDigest('a') + `"]}`},
-		{name: "malformed digest", body: `{"RepoDigests":["docker.io/library/alpine@sha256:not-a-digest"]}`},
-	} {
-		t.Run(test.name, func(t *testing.T) {
-			var registryCalls atomic.Int32
-			dm := newRegistryChecker(t, test.body, registryTransportFunc(func(req *http.Request) (*http.Response, error) {
-				registryCalls.Add(1)
-				return registryResponse(http.StatusOK, `{"token":"unexpected"}`), nil
-			}))
-
-			available, err := dm.checkImageUpdate("alpine")
-			require.Error(t, err)
-			require.False(t, available)
-			require.EqualValues(t, 0, registryCalls.Load(), "invalid local state must not query a registry")
-		})
-	}
-}
-
-func TestCheckImageUpdateChecksInspectAuthAndManifestStatuses(t *testing.T) {
-	local := registryDigest('a')
-	validInspect := fmt.Sprintf(`{"RepoDigests":["docker.io/library/alpine@%s"]}`, local)
-
-	tests := []struct {
-		name         string
-		inspectCode  int
-		authCode     int
-		manifestCode int
-		remote       string
-		want         string
-	}{
-		{name: "inspect status", inspectCode: http.StatusNotFound, want: "inspect image"},
-		{name: "auth status", inspectCode: http.StatusOK, authCode: http.StatusUnauthorized, want: "registry auth"},
-		{name: "manifest status", inspectCode: http.StatusOK, authCode: http.StatusOK, manifestCode: http.StatusNotFound, remote: local, want: "manifest request"},
-		{name: "missing digest", inspectCode: http.StatusOK, authCode: http.StatusOK, manifestCode: http.StatusOK, want: "invalid digest"},
-	}
-	for _, test := range tests {
-		t.Run(test.name, func(t *testing.T) {
-			server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-				if test.inspectCode != http.StatusOK && strings.HasPrefix(r.URL.Path, "/images/") {
-					w.WriteHeader(test.inspectCode)
-					return
-				}
-				_, _ = io.WriteString(w, validInspect)
-			}))
-			t.Cleanup(server.Close)
-
-			calls := 0
-			dm := &dockerManager{client: newDockerManagerForVersionTest(server).client, registryClient: &http.Client{Transport: registryTransportFunc(func(req *http.Request) (*http.Response, error) {
-				calls++
-				if req.Method == http.MethodGet {
-					return registryResponse(test.authCode, `{"token":"test"}`), nil
-				}
-				response := registryResponse(test.manifestCode, "")
-				response.Header.Set("Docker-Content-Digest", test.remote)
-				return response, nil
-			})}}
-
-			_, err := dm.checkImageUpdate("alpine")
-			require.Error(t, err)
-			require.Contains(t, err.Error(), test.want)
-			if test.inspectCode != http.StatusOK {
-				require.Zero(t, calls)
-			}
-		})
-	}
-}
-
-func TestCheckImageUpdateSupportsAnonymousAndLSCRRegistries(t *testing.T) {
-	t.Run("anonymous registry", func(t *testing.T) {
-		local := registryDigest('a')
-		var calls atomic.Int32
-		dm := newRegistryChecker(t, fmt.Sprintf(`{"RepoDigests":["example.com/app@%s"]}`, local), registryTransportFunc(func(req *http.Request) (*http.Response, error) {
-			calls.Add(1)
-			require.Equal(t, http.MethodHead, req.Method)
-			require.Equal(t, "example.com", req.URL.Host)
-			resp := registryResponse(http.StatusOK, "")
-			resp.Header.Set("Docker-Content-Digest", local)
-			return resp, nil
-		}))
-		available, err := dm.checkImageUpdate("example.com/app")
-		require.NoError(t, err)
-		require.False(t, available)
-		require.EqualValues(t, 1, calls.Load())
-	})
-
-	t.Run("lscr ghcr alias", func(t *testing.T) {
-		local := registryDigest('a')
-		var authCalls, manifestCalls atomic.Int32
-		dm := newRegistryChecker(t, fmt.Sprintf(`{"RepoDigests":["ghcr.io/linuxserver/app@%s"]}`, local), registryTransportFunc(func(req *http.Request) (*http.Response, error) {
-			if req.Method == http.MethodGet {
-				authCalls.Add(1)
-				return registryResponse(http.StatusOK, `{"token":"test"}`), nil
-			}
-			manifestCalls.Add(1)
-			require.Equal(t, "lscr.io", req.URL.Host)
-			resp := registryResponse(http.StatusOK, "")
-			resp.Header.Set("Docker-Content-Digest", local)
-			return resp, nil
-		}))
-		available, err := dm.checkImageUpdate("lscr.io/linuxserver/app")
-		require.NoError(t, err)
-		require.False(t, available)
-		require.EqualValues(t, 1, authCalls.Load())
-		require.EqualValues(t, 1, manifestCalls.Load())
-	})
-}
-
-func TestCheckImageUpdateSkipsPinnedDigest(t *testing.T) {
-	image := "docker.io/library/alpine@" + registryDigest('a')
-	dm := &dockerManager{}
-	available, err := dm.checkImageUpdate(image)
-	require.NoError(t, err)
-	require.False(t, available)
-}

agent/docker_test.go

@@ -1184,6 +1184,7 @@ func TestUpdateContainerStatsPodmanCpuCalculation(t *testing.T) {
 			}
 		})},
 		containerStatsMap: make(map[string]*container.Stats),
+		apiStats:          &container.ApiStats{},
 		usingPodman:       true,
 		lastCpuContainer: map[uint16]map[string]uint64{
 			defaultCacheTimeMs: {"0123456789ab": prevCpuUsage},
@@ -1675,6 +1676,7 @@ func TestUpdateContainerStatsUsesPodmanInspectHealthFallback(t *testing.T) {
 			}
 		})},
 		containerStatsMap:   make(map[string]*container.Stats),
+		apiStats:            &container.ApiStats{},
 		usingPodman:         true,
 		lastCpuContainer:    make(map[uint16]map[string]uint64),
 		lastCpuSystem:       make(map[uint16]map[string]uint64),

agent/fans.go

@@ -12,7 +12,7 @@ import (
 )
 
 type fanSensor struct {
-	key, path, chip string
+	key, path string
 }
 
 var getFanSensors = newFanSensorCache(hwmonRoot)
@@ -34,10 +34,6 @@ func (a *Agent) updateFans(systemStats *system.Stats) {
 		slog.Debug("Error reading fans", "err", err)
 		return
 	}
-	// Filter before reading fan*_input: each read can wake an idle GPU.
-	if a.sensorConfig != nil && a.sensorConfig.skipGPU {
-		sensors = filterGpuFans(sensors)
-	}
 	fans := readFanSensors(sensors)
 	if len(fans) == 0 {
 		return
@@ -104,7 +100,7 @@ func discoverHwmonFans(root string) ([]fanSensor, error) {
 			if label != "" {
 				key = chipName + "_" + label
 			}
-			sensors = append(sensors, fanSensor{key, inputPath, chipName})
+			sensors = append(sensors, fanSensor{key, inputPath})
 		}
 	}
 	return sensors, nil
@@ -119,15 +115,3 @@ func readFanSensors(sensors []fanSensor) map[string]uint16 {
 	}
 	return fans
 }
-
-// filterGpuFans drops GPU chips without touching the shared cache backing array.
-func filterGpuFans(sensors []fanSensor) []fanSensor {
-	kept := make([]fanSensor, 0, len(sensors))
-	for _, sensor := range sensors {
-		if isGpuChipName(sensor.chip) {
-			continue
-		}
-		kept = append(kept, sensor)
-	}
-	return kept
-}

agent/fans_test.go

@@ -103,20 +103,3 @@ func TestFanDiscoveryCache(t *testing.T) {
 	fans = readFanSensors(sensors)
 	assert.Equal(t, map[string]uint16{"chip_fan1": 1200}, fans)
 }
-
-func TestFilterGpuFans(t *testing.T) {
-	root := t.TempDir()
-	writeFile(t, filepath.Join(root, "hwmon0", "name"), "xe\n")
-	writeFile(t, filepath.Join(root, "hwmon0", "fan1_input"), "1200\n")
-	writeFile(t, filepath.Join(root, "hwmon1", "name"), "nct6798\n")
-	writeFile(t, filepath.Join(root, "hwmon1", "fan1_input"), "800\n")
-
-	discovered, err := discoverHwmonFans(root)
-	require.NoError(t, err)
-	require.Len(t, discovered, 2)
-
-	filtered := filterGpuFans(discovered)
-	require.Len(t, filtered, 1)
-	assert.Equal(t, "nct6798_fan1", filtered[0].key)
-	assert.Len(t, discovered, 2)
-}

agent/gpu.go

@@ -454,8 +454,8 @@ func (gm *GPUManager) storeSnapshot(id string, gpu *system.GPUData, cacheKey uin
 // It only reports capability presence and does not apply policy decisions.
 func (gm *GPUManager) discoverGpuCapabilities() gpuCapabilities {
 	caps := gpuCapabilities{
-		hasAmdSysfs:   gm.hasAmdSysfs(),
-		hasXe:         gm.hasXe(),
+		hasAmdSysfs: gm.hasAmdSysfs(),
+		hasXe:       gm.hasXe(),
 		hasIntelSysfs: gm.hasIntelSysfs(),
 	}
 	if _, err := exec.LookPath(nvidiaSmiCmd); err == nil {
@@ -750,36 +750,9 @@ func (gm *GPUManager) resolveLegacyCollectorPriority(caps gpuCapabilities) []col
 	return priorities
 }
 
-// gpuHwmonChips are hwmon chip names belonging to GPUs. Sensor reads on some
-// of these drivers (notably Intel Xe, where each read is a runtime PM resume)
-// wake the card, so SKIP_GPU must avoid touching them, not just hide them.
-var gpuHwmonChips = []string{"xe", "i915", "amdgpu", "radeon", "nvidia", "nouveau"}
-
-func isGpuChipName(name string) bool {
-	name = strings.ToLower(strings.TrimSpace(name))
-	for _, chip := range gpuHwmonChips {
-		if name == chip {
-			return true
-		}
-	}
-	return false
-}
-
-// SensorKeys are "<chip>" or "<chip>_<label>".
-func isGpuSensorKey(key string) bool {
-	key = strings.ToLower(strings.TrimSpace(key))
-	for _, chip := range gpuHwmonChips {
-		if key == chip || strings.HasPrefix(key, chip+"_") {
-			return true
-		}
-	}
-	return false
-}
-
 // NewGPUManager creates and initializes a new GPUManager
 func NewGPUManager() (*GPUManager, error) {
 	if skipGPU, _ := utils.GetEnv("SKIP_GPU"); skipGPU == "true" {
-		slog.Info("SKIP_GPU enabled, skipping GPU monitoring (collectors, temperatures, and fans)")
 		return nil, nil
 	}
 	var gm GPUManager

agent/gpu_test.go

@@ -1119,6 +1119,7 @@ func TestCalculateGPUAverage(t *testing.T) {
 }
 
 func TestGPUCapabilitiesAndLegacyPriority(t *testing.T) {
+	// Save original PATH
 	hasAmdSysfs := (&GPUManager{}).hasAmdSysfs()
 
 	tests := []struct {
@@ -1212,7 +1213,7 @@ echo "[]"`
 		{
 			name: "no gpu tools available",
 			setupCommands: func(_ string) error {
-				// The subtest already restricts PATH to its empty temporary directory.
+				t.Setenv("PATH", "")
 				return nil
 			},
 			wantErr: true,

agent/handlers.go

@@ -7,7 +7,6 @@ import (
 
 	"github.com/fxamacker/cbor/v2"
 	"github.com/henrygd/beszel/internal/common"
-	"github.com/henrygd/beszel/internal/entities/monitor"
 	"github.com/henrygd/beszel/internal/entities/smart"
 
 	"log/slog"
@@ -52,7 +51,6 @@ func NewHandlerRegistry() *HandlerRegistry {
 	registry.Register(common.GetContainerInfo, &GetContainerInfoHandler{})
 	registry.Register(common.GetSmartData, &GetSmartDataHandler{})
 	registry.Register(common.GetSystemdInfo, &GetSystemdInfoHandler{})
-	registry.Register(common.SyncNetworkMonitors, &SyncNetworkMonitorsHandler{})
 	registry.Register(common.GetZfsData, &GetZfsDataHandler{})
 
 	return registry
@@ -188,14 +186,14 @@ func (h *GetSmartDataHandler) Handle(hctx *HandlerContext) error {
 type GetZfsDataHandler struct{}
 
 func (h *GetZfsDataHandler) Handle(hctx *HandlerContext) error {
-	if hctx.Agent.storagePoolManager == nil {
+	if hctx.Agent.zfsManager == nil {
 		return hctx.SendResponse(nil, hctx.RequestID)
 	}
 	var req common.ZfsDataRequest
 	if err := cbor.Unmarshal(hctx.Request.Data, &req); err != nil {
 		return err
 	}
-	return hctx.SendResponse(hctx.Agent.storagePoolManager.GetDetail(req.Force), hctx.RequestID)
+	return hctx.SendResponse(hctx.Agent.zfsManager.GetDetail(req.Force), hctx.RequestID)
 }
 
 ////////////////////////////////////////////////////////////////////////////
@@ -225,21 +223,3 @@ func (h *GetSystemdInfoHandler) Handle(hctx *HandlerContext) error {
 
 	return hctx.SendResponse(details, hctx.RequestID)
 }
-
-////////////////////////////////////////////////////////////////////////////
-////////////////////////////////////////////////////////////////////////////
-
-// SyncNetworkMonitorsHandler handles monitor configuration sync from hub
-type SyncNetworkMonitorsHandler struct{}
-
-func (h *SyncNetworkMonitorsHandler) Handle(hctx *HandlerContext) error {
-	var req monitor.SyncRequest
-	if err := cbor.Unmarshal(hctx.Request.Data, &req); err != nil {
-		return err
-	}
-	resp, err := hctx.Agent.monitorManager.HandleSyncRequest(req)
-	if err != nil {
-		return err
-	}
-	return hctx.SendResponse(resp, hctx.RequestID)
-}

agent/handlers_test.go

@@ -34,19 +34,19 @@ func TestNewAgentResponseSmartData(t *testing.T) {
 
 func TestGetZfsDataHandlerForceRefresh(t *testing.T) {
 	poolCalls := 0
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) {
+	zm := &ZfsManager{detailInterval: time.Hour}
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) {
 		poolCalls++
 		return []zfs.PoolStat{{Name: "tank", Alloc: uint64(poolCalls)}}, nil
 	}
-	zm.backends[0].poolStatusesFn = func() ([]zfs.PoolStatus, error) { return nil, nil }
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
+	zm.poolStatusesFn = func() ([]zfs.PoolStatus, error) { return nil, nil }
+	zm.datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
 	zm.GetDetail(false)
 
 	requestData, err := cbor.Marshal(common.ZfsDataRequest{Force: true})
 	assert.NoError(t, err)
 	ctx := &HandlerContext{
-		Agent: &Agent{storagePoolManager: zm},
+		Agent: &Agent{zfsManager: zm},
 		Request: &common.HubRequest[cbor.RawMessage]{
 			Action: common.GetZfsData,
 			Data:   requestData,

agent/mdraid_linux.go

@@ -201,9 +201,12 @@ func mdraidSmartStatus(health mdraidHealth) string {
 	if health.mismatchCnt > 0 {
 		return "WARNING"
 	}
-	// "check" and "repair" are requested consistency scans, not evidence of
-	// array failure. With no health issues above, keep scrubbing green while
-	// reporting the sync action and progress attributes.
+	// "check" scans for consistency problems without repairing mismatches.
+	// With no mismatches, keep it green while reporting progress attributes.
+	switch syncAction {
+	case "repair":
+		return "WARNING"
+	}
 	switch state {
 	case "clean", "active", "active-idle", "write-pending", "read-auto", "readonly":
 		return "PASSED"

agent/mdraid_linux_test.go

@@ -174,25 +174,8 @@ func TestMdraidSmartStatus(t *testing.T) {
 	if got := mdraidSmartStatus(mdraidHealth{arrayState: "clean", mismatchCnt: 1}); got != "WARNING" {
 		t.Fatalf("mdraidSmartStatus(clean+mismatch) = %q, want WARNING", got)
 	}
-	for _, tc := range []struct {
-		name   string
-		health mdraidHealth
-		want   string
-	}{
-		{"clean", mdraidHealth{arrayState: "clean"}, "PASSED"},
-		{"active", mdraidHealth{arrayState: "active"}, "PASSED"},
-		{"mismatch", mdraidHealth{arrayState: "active", mismatchCnt: 1}, "WARNING"},
-		{"degraded", mdraidHealth{arrayState: "active", degraded: 1}, "FAILED"},
-		{"faulty member", mdraidHealth{arrayState: "active", faultyDisks: 1}, "FAILED"},
-		{"inactive", mdraidHealth{arrayState: "inactive"}, "FAILED"},
-		{"unknown", mdraidHealth{arrayState: "unknown"}, "UNKNOWN"},
-	} {
-		t.Run("repair/"+tc.name, func(t *testing.T) {
-			tc.health.syncAction = "repair"
-			if got := mdraidSmartStatus(tc.health); got != tc.want {
-				t.Fatalf("mdraidSmartStatus(%+v) = %q, want %s", tc.health, got, tc.want)
-			}
-		})
+	if got := mdraidSmartStatus(mdraidHealth{arrayState: "clean", syncAction: "repair"}); got != "WARNING" {
+		t.Fatalf("mdraidSmartStatus(repair) = %q, want WARNING", got)
 	}
 	if got := mdraidSmartStatus(mdraidHealth{arrayState: "clean"}); got != "PASSED" {
 		t.Fatalf("mdraidSmartStatus(clean) = %q, want PASSED", got)

agent/network_monitor.go

@@ -1,195 +0,0 @@
-package agent
-
-import (
-	"errors"
-	"fmt"
-	"net/http"
-	"sync"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/monitor"
-)
-
-// MonitorManager manages network monitor configurations and task lifetimes.
-type MonitorManager struct {
-	mu          sync.RWMutex
-	monitors    map[string]*monitorTask // keyed by monitor ID
-	probe       monitorProbe
-	certCheck   certChecker
-	resumeGuard monitorResumeGuard
-}
-
-func newMonitorManager() *MonitorManager {
-	return newMonitorManagerWithProbe(networkMonitorProbe(&http.Client{Timeout: monitor.MaxProbeTimeout}))
-}
-
-func newMonitorManagerWithProbe(probe monitorProbe) *MonitorManager {
-	return &MonitorManager{monitors: make(map[string]*monitorTask), probe: probe, certCheck: checkCert}
-}
-
-// SyncMonitors replaces all monitor tasks with the given configs.
-func (pm *MonitorManager) SyncMonitors(configs []monitor.Config) {
-	pm.mu.Lock()
-	defer pm.mu.Unlock()
-
-	// Build set of new keys
-	newKeys := make(map[string]monitor.Config, len(configs))
-	for _, cfg := range configs {
-		if cfg.ID == "" {
-			continue
-		}
-		newKeys[cfg.ID] = cfg
-	}
-
-	// Stop removed monitors
-	for key, task := range pm.monitors {
-		if _, exists := newKeys[key]; !exists {
-			task.cancel()
-			delete(pm.monitors, key)
-		}
-	}
-
-	// Start new monitors and restart tasks whose config changed.
-	for key, cfg := range newKeys {
-		task, exists := pm.monitors[key]
-		if exists && task.config == cfg {
-			continue
-		}
-		if exists {
-			task.cancel()
-		}
-		task = newMonitorTaskFromExisting(cfg, task)
-		task.resumeGuard = &pm.resumeGuard
-		pm.resumeGuard.start()
-		pm.monitors[key] = task
-		pm.startMonitor(task)
-	}
-	if len(pm.monitors) == 0 {
-		pm.resumeGuard.shutdown()
-	}
-}
-
-// HandleSyncRequest applies a full or incremental monitor sync request.
-func (pm *MonitorManager) HandleSyncRequest(req monitor.SyncRequest) (monitor.SyncResponse, error) {
-	switch req.Action {
-	case monitor.SyncActionReplace:
-		pm.SyncMonitors(req.Configs)
-		return monitor.SyncResponse{}, nil
-	case monitor.SyncActionUpsert:
-		result, err := pm.UpsertMonitor(req.Config, req.RunNow)
-		if err != nil {
-			return monitor.SyncResponse{}, err
-		}
-		if result == nil {
-			return monitor.SyncResponse{}, nil
-		}
-		return monitor.SyncResponse{Result: *result}, nil
-	case monitor.SyncActionDelete:
-		if req.Config.ID == "" {
-			return monitor.SyncResponse{}, errors.New("missing monitor ID for delete")
-		}
-		pm.DeleteMonitor(req.Config.ID)
-		return monitor.SyncResponse{}, nil
-	default:
-		return monitor.SyncResponse{}, fmt.Errorf("unknown monitor sync action: %d", req.Action)
-	}
-}
-
-// UpsertMonitor creates or replaces a single monitor task.
-func (pm *MonitorManager) UpsertMonitor(config monitor.Config, runNow bool) (*monitor.Result, error) {
-	if config.ID == "" {
-		return nil, errors.New("missing monitor ID")
-	}
-
-	pm.mu.Lock()
-	task, exists := pm.monitors[config.ID]
-	if exists && task.config == config {
-		pm.mu.Unlock()
-		if !runNow {
-			return nil, nil
-		}
-		return pm.runNow(task), nil
-	}
-	if exists {
-		task.cancel()
-	}
-	task = newMonitorTaskFromExisting(config, task)
-	task.resumeGuard = &pm.resumeGuard
-	pm.resumeGuard.start()
-	pm.monitors[config.ID] = task
-	pm.mu.Unlock()
-
-	if runNow {
-		result := pm.runNow(task)
-		pm.startMonitor(task)
-		return result, nil
-	}
-	pm.startMonitor(task)
-	return nil, nil
-}
-
-// runNow runs a probe and any due certificate check concurrently, so the
-// response fits within the hub's single probe timeout budget.
-func (pm *MonitorManager) runNow(task *monitorTask) *monitor.Result {
-	var wg sync.WaitGroup
-	wg.Go(func() { task.refreshCert(pm.certCheck) })
-	result := task.runProbe(pm.probe)
-	wg.Wait()
-	if result != nil {
-		result.Cert = task.certInfo()
-	}
-	return result
-}
-
-// DeleteMonitor stops and removes a single monitor task.
-func (pm *MonitorManager) DeleteMonitor(id string) {
-	if id == "" {
-		return
-	}
-	pm.mu.Lock()
-	defer pm.mu.Unlock()
-	if task, exists := pm.monitors[id]; exists {
-		task.cancel()
-		delete(pm.monitors, id)
-	}
-	if len(pm.monitors) == 0 {
-		pm.resumeGuard.shutdown()
-	}
-}
-
-// GetResults returns aggregated results for all monitors over the last supplied duration in ms.
-func (pm *MonitorManager) GetResults(durationMs uint16) map[string]monitor.Result {
-	pm.mu.RLock()
-	defer pm.mu.RUnlock()
-
-	results := make(map[string]monitor.Result, len(pm.monitors))
-	now := time.Now()
-	duration := time.Duration(durationMs) * time.Millisecond
-
-	for _, task := range pm.monitors {
-		result, ok := task.history.result(duration, now)
-
-		if !ok {
-			continue
-		}
-		// Only the default interval updates monitor records on the hub, so
-		// realtime requests must not consume the unsent certificate.
-		if durationMs == defaultDataCacheTimeMs {
-			result.Cert = task.takeUnsentCert()
-		}
-		results[task.config.ID] = result
-	}
-
-	return results
-}
-
-// Stop stops all monitor tasks.
-func (pm *MonitorManager) Stop() {
-	pm.mu.Lock()
-	defer pm.mu.Unlock()
-	for key, task := range pm.monitors {
-		task.cancel()
-		delete(pm.monitors, key)
-	}
-	pm.resumeGuard.shutdown()
-}

agent/network_monitor_cert.go

@@ -1,74 +0,0 @@
-package agent
-
-import (
-	"context"
-	"crypto/tls"
-	"errors"
-	"fmt"
-	"net"
-	"net/url"
-	"strings"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/monitor"
-)
-
-const (
-	certCheckInterval      = 24 * time.Hour
-	certCheckRetryInterval = time.Hour
-)
-
-// certChecker fetches the leaf certificate for an HTTPS target.
-type certChecker func(context.Context, string) (monitor.CertInfo, error)
-
-// certCheckEnabled reports whether a monitor's certificate is checked, which is
-// the case for every HTTP monitor with an https target.
-func certCheckEnabled(config monitor.Config) bool {
-	return config.Protocol == "http" && len(config.Target) > 8 && strings.EqualFold(config.Target[:8], "https://")
-}
-
-// checkCert reads the leaf certificate presented by an HTTPS target. The chain is
-// not verified, so expired or self-signed certificates are still reported.
-func checkCert(ctx context.Context, target string) (monitor.CertInfo, error) {
-	address, host, err := certAddress(target)
-	if err != nil {
-		return monitor.CertInfo{}, err
-	}
-	ctx, cancel := context.WithTimeout(ctx, monitor.MaxProbeTimeout)
-	defer cancel()
-	dialer := tls.Dialer{Config: &tls.Config{ServerName: host, InsecureSkipVerify: true}}
-	conn, err := dialer.DialContext(ctx, "tcp", address)
-	if err != nil {
-		return monitor.CertInfo{}, err
-	}
-	defer conn.Close()
-	certs := conn.(*tls.Conn).ConnectionState().PeerCertificates
-	if len(certs) == 0 {
-		return monitor.CertInfo{}, errors.New("no peer certificates")
-	}
-	leaf := certs[0]
-	return monitor.CertInfo{
-		Expires: leaf.NotAfter.UnixMilli(),
-		Issuer:  leaf.Issuer.CommonName,
-	}, nil
-}
-
-// certAddress returns the dial address and server name for an HTTPS URL.
-func certAddress(target string) (address, host string, err error) {
-	u, err := url.Parse(target)
-	if err != nil {
-		return "", "", err
-	}
-	if !strings.EqualFold(u.Scheme, "https") {
-		return "", "", fmt.Errorf("certificate check requires an https target: %s", target)
-	}
-	host = u.Hostname()
-	if host == "" {
-		return "", "", fmt.Errorf("missing host in target: %s", target)
-	}
-	port := u.Port()
-	if port == "" {
-		port = "443"
-	}
-	return net.JoinHostPort(host, port), host, nil
-}

agent/network_monitor_cert_test.go

@@ -1,184 +0,0 @@
-//go:build testing
-
-package agent
-
-import (
-	"context"
-	"errors"
-	"net/http"
-	"net/http/httptest"
-	"testing"
-	"testing/synctest"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/monitor"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestCheckCertReadsUnverifiedLeaf(t *testing.T) {
-	server := httptest.NewTLSServer(http.HandlerFunc(func(http.ResponseWriter, *http.Request) {}))
-	defer server.Close()
-
-	// httptest uses a self-signed certificate, which must still be reported.
-	info, err := checkCert(context.Background(), server.URL)
-	require.NoError(t, err)
-	leaf := server.Certificate()
-	assert.Equal(t, leaf.NotAfter.UnixMilli(), info.Expires)
-	assert.Equal(t, leaf.Issuer.CommonName, info.Issuer)
-}
-
-func TestCertAddress(t *testing.T) {
-	tests := []struct {
-		target, address, host string
-		wantErr               bool
-	}{
-		{target: "https://example.com", address: "example.com:443", host: "example.com"},
-		{target: "https://example.com:8443/path?q=1", address: "example.com:8443", host: "example.com"},
-		{target: "HTTPS://[::1]:9443", address: "[::1]:9443", host: "::1"},
-		{target: "http://example.com", wantErr: true},
-		{target: "https://", wantErr: true},
-	}
-	for _, tt := range tests {
-		address, host, err := certAddress(tt.target)
-		if tt.wantErr {
-			assert.Error(t, err, tt.target)
-			continue
-		}
-		require.NoError(t, err, tt.target)
-		assert.Equal(t, tt.address, address)
-		assert.Equal(t, tt.host, host)
-	}
-}
-
-func TestRefreshCertCadence(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		task := newMonitorTask(monitor.Config{ID: "test", Target: "https://example.test", Protocol: "http"})
-		defer task.cancel()
-		var calls int
-		var fail error
-		// Far enough out that the regular interval applies for the whole test.
-		expires := time.Now().Add(365 * 24 * time.Hour).UnixMilli()
-		check := func(context.Context, string) (monitor.CertInfo, error) {
-			calls++
-			if fail != nil {
-				return monitor.CertInfo{}, fail
-			}
-			return monitor.CertInfo{Expires: expires + int64(calls)}, nil
-		}
-
-		task.refreshCert(check)
-		require.NotNil(t, task.certInfo())
-		assert.Equal(t, expires+1, task.certInfo().Expires)
-
-		// Not due again until the check interval passes.
-		time.Sleep(certCheckInterval - time.Second)
-		task.refreshCert(check)
-		assert.Equal(t, 1, calls)
-		time.Sleep(time.Second)
-		task.refreshCert(check)
-		assert.Equal(t, 2, calls)
-
-		// Failures keep the last known certificate and retry sooner.
-		fail = errors.New("connection refused")
-		time.Sleep(certCheckInterval)
-		task.refreshCert(check)
-		assert.Equal(t, 3, calls)
-		assert.Equal(t, expires+2, task.certInfo().Expires)
-		time.Sleep(certCheckRetryInterval)
-		fail = nil
-		task.refreshCert(check)
-		assert.Equal(t, 4, calls)
-		assert.Equal(t, expires+4, task.certInfo().Expires)
-	})
-}
-
-func TestRefreshCertRetriesSoonerNearExpiry(t *testing.T) {
-	for _, tc := range []struct {
-		name     string
-		expires  time.Duration // relative to the check
-		interval time.Duration
-	}{
-		{"expired", -time.Hour, certCheckRetryInterval},
-		{"expires before next regular check", certCheckInterval - time.Minute, certCheckRetryInterval},
-		{"expires after next regular check", certCheckInterval + time.Minute, certCheckInterval},
-	} {
-		t.Run(tc.name, func(t *testing.T) {
-			synctest.Test(t, func(t *testing.T) {
-				task := newMonitorTask(monitor.Config{ID: "test", Target: "https://example.test", Protocol: "http"})
-				defer task.cancel()
-				var calls int
-				check := func(context.Context, string) (monitor.CertInfo, error) {
-					calls++
-					return monitor.CertInfo{Expires: time.Now().Add(tc.expires).UnixMilli()}, nil
-				}
-				task.refreshCert(check)
-				time.Sleep(tc.interval - time.Second)
-				task.refreshCert(check)
-				assert.Equal(t, 1, calls)
-				time.Sleep(time.Second)
-				task.refreshCert(check)
-				assert.Equal(t, 2, calls)
-			})
-		})
-	}
-}
-
-func TestCertCheckEnabled(t *testing.T) {
-	tests := []struct {
-		protocol, target string
-		want             bool
-	}{
-		{"http", "https://example.com", true},
-		{"http", "HTTPS://example.com:8443/path", true},
-		{"http", "http://example.com", false},
-		{"http", "https://", false},
-		{"tcp", "https://example.com", false},
-		{"icmp", "example.com", false},
-	}
-	for _, tt := range tests {
-		assert.Equal(t, tt.want, certCheckEnabled(monitor.Config{Protocol: tt.protocol, Target: tt.target}), tt.protocol+" "+tt.target)
-	}
-}
-
-func TestRefreshCertSkipsNonHTTPS(t *testing.T) {
-	task := newMonitorTask(monitor.Config{ID: "test", Target: "http://example.test", Protocol: "http"})
-	defer task.cancel()
-	task.refreshCert(func(context.Context, string) (monitor.CertInfo, error) {
-		t.Fatal("certificate check must not run for non-https targets")
-		return monitor.CertInfo{}, nil
-	})
-	assert.Nil(t, task.certInfo())
-}
-
-func TestUpsertMonitorRunNowIncludesCert(t *testing.T) {
-	server := httptest.NewTLSServer(http.HandlerFunc(func(http.ResponseWriter, *http.Request) {}))
-	defer server.Close()
-
-	pm := newMonitorManagerWithProbe(func(context.Context, monitor.Config) (int64, error) { return 100, nil })
-	defer pm.Stop()
-	config := monitor.Config{ID: "cert", Target: server.URL, Protocol: "http", Interval: 60}
-	result, err := pm.UpsertMonitor(config, true)
-	require.NoError(t, err)
-	require.NotNil(t, result)
-	require.NotNil(t, result.Cert)
-	assert.Equal(t, server.Certificate().NotAfter.UnixMilli(), result.Cert.Expires)
-
-	// Realtime results never carry the certificate, and the default interval
-	// sends it only once per check.
-	assert.Nil(t, pm.GetResults(1000)["cert"].Cert)
-	results := pm.GetResults(defaultDataCacheTimeMs)
-	require.NotNil(t, results["cert"].Cert)
-	assert.Equal(t, result.Cert.Expires, results["cert"].Cert.Expires)
-	assert.Nil(t, pm.GetResults(defaultDataCacheTimeMs)["cert"].Cert)
-
-	// Changing the interval keeps the known certificate without resending it.
-	config.Interval = 30
-	_, err = pm.UpsertMonitor(config, false)
-	require.NoError(t, err)
-	pm.mu.RLock()
-	task := pm.monitors["cert"]
-	pm.mu.RUnlock()
-	assert.NotNil(t, task.certInfo())
-	assert.Nil(t, pm.GetResults(defaultDataCacheTimeMs)["cert"].Cert)
-}

agent/network_monitor_history.go

@@ -1,274 +0,0 @@
-package agent
-
-import (
-	"math"
-	"sync"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/monitor"
-)
-
-// Monitors run at user-defined intervals (e.g., every 10s).
-// To keep memory usage low and constant, data is stored in two layers:
-// 1. Raw samples: The most recent individual results (kept for monitorRawRetention).
-// 2. Minute buckets: A ring buffer of 61 buckets, each representing one
-//    wall-clock minute. Samples collected within the same minute are aggregated
-//    (sum, min, max, count) into a single bucket.
-//
-// Short-term requests (<= 61s) use raw samples.
-// Long-term requests (up to 1h) use the minute buckets to avoid storing thousands
-// of individual data points.
-
-const (
-	// monitorRawRetention is the duration to keep individual samples
-	monitorRawRetention = 61 * time.Second
-	// monitorMinuteBucketLen is the number of 1-minute buckets to keep (1 hour + 1 for partials)
-	monitorMinuteBucketLen int32 = 61
-)
-
-// monitorHistory owns retention and aggregation, independently of probe execution.
-type monitorHistory struct {
-	mu          sync.Mutex
-	sampleCount int64
-	samples     []monitorSample
-	buckets     [monitorMinuteBucketLen]monitorBucket
-}
-
-func newMonitorHistory() *monitorHistory {
-	// Start small for typical intervals; append grows the buffer for faster probes.
-	return &monitorHistory{samples: make([]monitorSample, 0, 4)}
-}
-
-func (h *monitorHistory) clone() *monitorHistory {
-	h.mu.Lock()
-	defer h.mu.Unlock()
-	cloned := newMonitorHistory()
-	cloned.samples = append(cloned.samples, h.samples...)
-	cloned.buckets = h.buckets
-	cloned.sampleCount = h.sampleCount
-	return cloned
-}
-
-func (h *monitorHistory) result(duration time.Duration, now time.Time) (monitor.Result, bool) {
-	h.mu.Lock()
-	defer h.mu.Unlock()
-	return h.resultLocked(duration, now)
-}
-
-func (h *monitorHistory) record(sample monitorSample) monitor.Result {
-	h.mu.Lock()
-	defer h.mu.Unlock()
-	h.addSampleLocked(sample)
-	result, _ := h.resultLocked(time.Minute, sample.timestamp)
-	return result
-}
-
-// monitorSample stores one monitor attempt and its collection time.
-type monitorSample struct {
-	responseUs int64 // -1 means loss
-	timestamp  time.Time
-}
-
-// monitorBucket stores one minute of aggregated monitor data.
-type monitorBucket struct {
-	minute int32
-	filled bool
-	stats  monitorAggregate
-}
-
-// monitorAggregate accumulates successful response stats and total sample counts.
-type monitorAggregate struct {
-	sumUs        int64
-	minUs        int64
-	maxUs        int64
-	totalCount   int64
-	successCount int64
-}
-
-// newMonitorAggregate initializes an aggregate with an unset minimum value.
-func newMonitorAggregate() monitorAggregate {
-	return monitorAggregate{minUs: math.MaxInt64}
-}
-
-// addResponse folds a single monitor sample into the aggregate.
-func (agg *monitorAggregate) addResponse(responseUs int64) {
-	agg.totalCount++
-	if responseUs < 0 {
-		return
-	}
-	agg.successCount++
-	agg.sumUs += responseUs
-	if responseUs < agg.minUs {
-		agg.minUs = responseUs
-	}
-	if responseUs > agg.maxUs {
-		agg.maxUs = responseUs
-	}
-}
-
-// addAggregate merges another aggregate into this one.
-func (agg *monitorAggregate) addAggregate(other monitorAggregate) {
-	if other.totalCount == 0 {
-		return
-	}
-	agg.totalCount += other.totalCount
-	agg.successCount += other.successCount
-	agg.sumUs += other.sumUs
-	if other.successCount == 0 {
-		return
-	}
-	if agg.minUs == math.MaxInt64 || other.minUs < agg.minUs {
-		agg.minUs = other.minUs
-	}
-	if other.maxUs > agg.maxUs {
-		agg.maxUs = other.maxUs
-	}
-}
-
-// hasData reports whether the aggregate contains any samples.
-func (agg monitorAggregate) hasData() bool {
-	return agg.totalCount > 0
-}
-
-// result converts the aggregate into the monitor result format.
-func (agg monitorAggregate) result() monitor.Result {
-	avg := agg.avgResponse()
-	result := monitor.Result{
-		AvgResponse:  avg,
-		MinResponse:  agg.minUs,
-		MaxResponse:  agg.maxUs,
-		PacketLoss:   agg.lossPercentage(),
-		TotalCount:   agg.totalCount,
-		SuccessCount: agg.successCount,
-		ResponseSum:  agg.sumUs,
-	}
-	if agg.successCount == 0 {
-		result.MinResponse, result.MaxResponse = 0, 0
-	}
-	return result
-}
-
-// avgResponse returns the rounded average of successful samples.
-func (agg monitorAggregate) avgResponse() int64 {
-	if agg.successCount == 0 {
-		return 0
-	}
-	return agg.sumUs / agg.successCount
-
-}
-
-// lossPercentage returns the rounded failure rate for the aggregate.
-func (agg monitorAggregate) lossPercentage() float64 {
-	if agg.totalCount == 0 {
-		return 0
-	}
-	return math.Round(float64(agg.totalCount-agg.successCount)/float64(agg.totalCount)*10000) / 100
-}
-
-// resultLocked returns the aggregated monitor result for the requested duration along with a bool indicating whether any data was available.
-func (h *monitorHistory) resultLocked(duration time.Duration, now time.Time) (monitor.Result, bool) {
-	agg := h.aggregateLocked(duration, now)
-	if !agg.hasData() {
-		// short realtime windows (e.g. the 1s window used for 1m/realtime charts) often fall
-		// between monitor samples since monitors run at longer, user-defined intervals; fall back to
-		// the most recent sample so realtime requests still report current status.
-		agg = h.latestSampleAggregateLocked()
-	}
-	hourAgg := h.aggregateLocked(time.Hour, now)
-	if !agg.hasData() {
-		return monitor.Result{}, false
-	}
-
-	result := agg.result()
-	if len(h.samples) > 0 {
-		result.LastProbeAt = h.samples[len(h.samples)-1].timestamp.UnixMilli()
-	}
-
-	result.AvgResponse1h = hourAgg.avgResponse()
-	result.MinResponse1h = hourAgg.minUs
-	result.MaxResponse1h = hourAgg.maxUs
-	result.PacketLoss1h = hourAgg.lossPercentage()
-	result.SampleCount = h.sampleCount
-
-	if hourAgg.successCount == 0 {
-		result.MinResponse1h, result.MaxResponse1h = 0, 0
-	}
-	return result, true
-}
-
-// latestSampleAggregateLocked returns an aggregate containing only the most recent sample, if any.
-func (h *monitorHistory) latestSampleAggregateLocked() monitorAggregate {
-	agg := newMonitorAggregate()
-	if len(h.samples) == 0 {
-		return agg
-	}
-	agg.addResponse(h.samples[len(h.samples)-1].responseUs)
-	return agg
-}
-
-// aggregateLocked collects monitor data for the requested time window.
-func (h *monitorHistory) aggregateLocked(duration time.Duration, now time.Time) monitorAggregate {
-	cutoff := now.Add(-duration)
-	// Keep short windows exact; longer windows read from minute buckets to avoid raw-sample retention.
-	if duration <= monitorRawRetention {
-		return aggregateSamplesSince(h.samples, cutoff)
-	}
-	return aggregateBucketsSince(h.buckets[:], cutoff, now)
-}
-
-// aggregateSamplesSince aggregates raw samples newer than the cutoff.
-func aggregateSamplesSince(samples []monitorSample, cutoff time.Time) monitorAggregate {
-	agg := newMonitorAggregate()
-	for _, sample := range samples {
-		if sample.timestamp.Before(cutoff) {
-			continue
-		}
-		agg.addResponse(sample.responseUs)
-	}
-	return agg
-}
-
-// aggregateBucketsSince aggregates minute buckets overlapping the requested window.
-func aggregateBucketsSince(buckets []monitorBucket, cutoff, now time.Time) monitorAggregate {
-	agg := newMonitorAggregate()
-	startMinute := int32(cutoff.Unix() / 60)
-	endMinute := int32(now.Unix() / 60)
-	for _, bucket := range buckets {
-		if !bucket.filled || bucket.minute < startMinute || bucket.minute > endMinute {
-			continue
-		}
-		agg.addAggregate(bucket.stats)
-	}
-	return agg
-}
-
-// addSampleLocked stores a fresh sample in both raw and per-minute retention buffers.
-func (h *monitorHistory) addSampleLocked(sample monitorSample) {
-	h.sampleCount++
-	cutoff := sample.timestamp.Add(-monitorRawRetention)
-	start := 0
-	for i := range h.samples {
-		if !h.samples[i].timestamp.Before(cutoff) {
-			start = i
-			break
-		}
-		if i == len(h.samples)-1 {
-			start = len(h.samples)
-		}
-	}
-	if start > 0 {
-		size := copy(h.samples, h.samples[start:])
-		h.samples = h.samples[:size]
-	}
-	h.samples = append(h.samples, sample)
-
-	minute := int32(sample.timestamp.Unix() / 60)
-	// Each slot stores one wall-clock minute, so the ring stays fixed-size at ~1h per monitor.
-	bucket := &h.buckets[minute%monitorMinuteBucketLen]
-	if !bucket.filled || bucket.minute != minute {
-		bucket.minute = minute
-		bucket.filled = true
-		bucket.stats = newMonitorAggregate()
-	}
-	bucket.stats.addResponse(sample.responseUs)
-}

agent/network_monitor_history_test.go

@@ -1,154 +0,0 @@
-package agent
-
-import (
-	"testing"
-	"time"
-
-	"github.com/fxamacker/cbor/v2"
-	"github.com/henrygd/beszel/internal/entities/monitor"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestMonitorHistoryWindowCounts(t *testing.T) {
-	history := newMonitorHistory()
-	now := time.Now()
-	// This older success counts toward lifetime warm-up, but not this window.
-	history.record(monitorSample{responseUs: 1000, timestamp: now.Add(-2 * time.Minute)})
-	history.record(monitorSample{responseUs: 10, timestamp: now.Add(-30 * time.Second)})
-	history.record(monitorSample{responseUs: 21, timestamp: now.Add(-20 * time.Second)})
-	history.record(monitorSample{responseUs: -1, timestamp: now.Add(-10 * time.Second)})
-	result, ok := history.result(time.Minute, now)
-	require.True(t, ok)
-	assert.EqualValues(t, 4, result.SampleCount)
-	assert.EqualValues(t, 3, result.TotalCount)
-	assert.EqualValues(t, 2, result.SuccessCount)
-	assert.EqualValues(t, 31, result.ResponseSum, "preserve the sum before average rounding")
-	assert.EqualValues(t, 15, result.AvgResponse)
-	assert.Equal(t, 33.33, result.PacketLoss)
-
-	encoded, err := cbor.Marshal(result)
-	require.NoError(t, err)
-	var decoded monitor.Result
-	require.NoError(t, cbor.Unmarshal(encoded, &decoded))
-	assert.Equal(t, result, decoded)
-	stats := monitor.Stats{}.FromResult(decoded)
-	assert.Equal(t, result.TotalCount, stats.TotalCount)
-	assert.Equal(t, result.SuccessCount, stats.SuccessCount)
-	assert.Equal(t, result.ResponseSum, stats.ResponseSum)
-
-	// Reads do not consume samples. A short window's latest-sample fallback
-	// carries the count for that single failure, not the minute or lifetime count.
-	repeated, _ := history.result(time.Minute, now)
-	assert.Equal(t, result, repeated)
-	fallback, ok := history.result(time.Second, now)
-	require.True(t, ok)
-	assert.EqualValues(t, 1, fallback.TotalCount)
-	assert.Zero(t, fallback.SuccessCount)
-	assert.Zero(t, fallback.ResponseSum)
-	assert.Equal(t, 100.0, fallback.PacketLoss)
-	assert.EqualValues(t, 4, fallback.SampleCount)
-}
-
-func TestMonitorHistoryAggregateLockedUsesRawSamplesForShortWindows(t *testing.T) {
-	now := time.Date(2026, time.April, 21, 12, 0, 0, 0, time.UTC)
-	history := newMonitorHistory()
-
-	history.addSampleLocked(monitorSample{responseUs: 10, timestamp: now.Add(-90 * time.Second)})
-	history.addSampleLocked(monitorSample{responseUs: 20, timestamp: now.Add(-30 * time.Second)})
-	history.addSampleLocked(monitorSample{responseUs: -1, timestamp: now.Add(-10 * time.Second)})
-
-	agg := history.aggregateLocked(time.Minute, now)
-	require.True(t, agg.hasData())
-	assert.Equal(t, int64(2), agg.totalCount)
-	assert.Equal(t, int64(1), agg.successCount)
-	result := agg.result()
-	assert.Equal(t, int64(20), result.AvgResponse)
-	assert.Equal(t, int64(20), result.MinResponse)
-	assert.Equal(t, int64(20), result.MaxResponse)
-	assert.Equal(t, 50.0, result.PacketLoss)
-}
-
-func TestMonitorHistoryAggregateLockedUsesMinuteBucketsForLongWindows(t *testing.T) {
-	now := time.Date(2026, time.April, 21, 12, 0, 30, 0, time.UTC)
-	history := newMonitorHistory()
-
-	history.addSampleLocked(monitorSample{responseUs: 10, timestamp: now.Add(-11 * time.Minute)})
-	history.addSampleLocked(monitorSample{responseUs: 20, timestamp: now.Add(-9 * time.Minute)})
-	history.addSampleLocked(monitorSample{responseUs: 40, timestamp: now.Add(-5 * time.Minute)})
-	history.addSampleLocked(monitorSample{responseUs: -1, timestamp: now.Add(-90 * time.Second)})
-	history.addSampleLocked(monitorSample{responseUs: 30, timestamp: now.Add(-30 * time.Second)})
-
-	agg := history.aggregateLocked(10*time.Minute, now)
-	require.True(t, agg.hasData())
-	assert.Equal(t, int64(4), agg.totalCount)
-	assert.Equal(t, int64(3), agg.successCount)
-	result := agg.result()
-	assert.Equal(t, int64(30), result.AvgResponse)
-	assert.Equal(t, int64(20), result.MinResponse)
-	assert.Equal(t, int64(40), result.MaxResponse)
-	assert.Equal(t, 25.0, result.PacketLoss)
-}
-
-func TestMonitorHistoryAddSampleLockedTrimsRawSamplesButKeepsBucketHistory(t *testing.T) {
-	now := time.Date(2026, time.April, 21, 12, 0, 0, 0, time.UTC)
-	history := newMonitorHistory()
-
-	history.addSampleLocked(monitorSample{responseUs: 10, timestamp: now.Add(-10 * time.Minute)})
-	history.addSampleLocked(monitorSample{responseUs: 20, timestamp: now})
-
-	require.Len(t, history.samples, 1)
-	assert.Equal(t, int64(20), history.samples[0].responseUs)
-
-	agg := history.aggregateLocked(10*time.Minute, now)
-	require.True(t, agg.hasData())
-	assert.Equal(t, int64(2), agg.totalCount)
-	assert.Equal(t, int64(2), agg.successCount)
-	result := agg.result()
-	assert.Equal(t, int64(15), result.AvgResponse)
-	assert.Equal(t, int64(10), result.MinResponse)
-	assert.Equal(t, int64(20), result.MaxResponse)
-	assert.Equal(t, 0.0, result.PacketLoss)
-}
-
-func TestMonitorHistoryProbeTimestamp(t *testing.T) {
-	history := newMonitorHistory()
-	start := time.Date(2026, time.September, 14, 12, 0, 0, 0, time.UTC)
-	_, ok := history.result(time.Minute, start)
-	require.False(t, ok)
-	first := history.record(monitorSample{responseUs: 20, timestamp: start})
-	assert.Equal(t, start.UnixMilli(), first.LastProbeAt)
-	for minute := 0; minute < 5; minute++ {
-		now := start.Add(time.Duration(minute)*time.Minute + time.Second)
-		// Realtime reads must not consume freshness for the persistence request.
-		for _, window := range []time.Duration{time.Second, time.Minute} {
-			result, ok := history.result(window, now)
-			require.True(t, ok)
-			assert.Equal(t, first.LastProbeAt, result.LastProbeAt)
-			assert.Equal(t, int64(20), result.AvgResponse)
-		}
-	}
-	next := start.Add(5 * time.Minute)
-	failed := history.record(monitorSample{responseUs: -1, timestamp: next})
-	assert.Equal(t, next.UnixMilli(), failed.LastProbeAt)
-	assert.Equal(t, float64(100), failed.PacketLoss)
-	repeated, ok := history.result(time.Minute, next.Add(2*time.Minute))
-	require.True(t, ok)
-	assert.Equal(t, failed.LastProbeAt, repeated.LastProbeAt)
-	assert.Equal(t, float64(100), repeated.PacketLoss)
-}
-
-func TestMonitorHistorySampleCount(t *testing.T) {
-	history := newMonitorHistory()
-	now := time.Now()
-	// Both failed and successful probes count, including older samples so
-	// monitors with hourly intervals can finish warming up.
-	history.record(monitorSample{responseUs: -1, timestamp: now.Add(-2 * time.Hour)})
-	for i, response := range []int64{10, -1, 20} {
-		result := history.record(monitorSample{responseUs: response, timestamp: now.Add(time.Duration(i) * time.Second)})
-		assert.EqualValues(t, i+2, result.SampleCount)
-	}
-	result, ok := history.clone().result(time.Minute, now.Add(3*time.Second))
-	require.True(t, ok)
-	assert.EqualValues(t, 4, result.SampleCount)
-}

agent/network_monitor_ping.go

@@ -1,312 +0,0 @@
-package agent
-
-import (
-	"bytes"
-	"context"
-	"crypto/rand"
-	"errors"
-	"fmt"
-	"math"
-	"net"
-	"os"
-	"os/exec"
-	"regexp"
-	"runtime"
-	"strconv"
-	"strings"
-	"sync"
-	"sync/atomic"
-	"time"
-
-	"golang.org/x/net/icmp"
-	"golang.org/x/net/ipv4"
-	"golang.org/x/net/ipv6"
-
-	"log/slog"
-)
-
-// Match the numeric RTT independently of the localized label used by Windows.
-var pingTimeRegex = regexp.MustCompile(`(?i)[=<]\s*([0-9]+(?:[.,][0-9]+)?)\s*ms\b`)
-
-var icmpSequence atomic.Uint32
-
-type icmpPacketConn interface {
-	Close() error
-}
-
-// icmpMethod tracks which ICMP approach to use. Once a method succeeds or
-// all native methods fail, the choice is cached so subsequent monitors skip
-// the trial-and-error overhead.
-type icmpMethod uint8
-
-const (
-	icmpUntried      icmpMethod = iota // haven't tried yet
-	icmpRaw                            // privileged raw socket
-	icmpDatagram                       // unprivileged datagram socket
-	icmpExecFallback                   // shell out to system ping command
-)
-
-// icmpFamily holds the network parameters and cached detection result for one address family.
-type icmpFamily struct {
-	rawNetwork   string    // e.g. "ip4:icmp" or "ip6:ipv6-icmp"
-	dgramNetwork string    // e.g. "udp4" or "udp6"
-	listenAddr   string    // "0.0.0.0" or "::"
-	echoType     icmp.Type // outgoing echo request type
-	replyType    icmp.Type // expected echo reply type
-	proto        int       // IANA protocol number for parsing replies
-	isIPv6       bool
-	mode         icmpMethod // cached detection result (guarded by icmpModeMu)
-}
-
-var (
-	icmpV4 = icmpFamily{
-		rawNetwork:   "ip4:icmp",
-		dgramNetwork: "udp4",
-		listenAddr:   "0.0.0.0",
-		echoType:     ipv4.ICMPTypeEcho,
-		replyType:    ipv4.ICMPTypeEchoReply,
-		proto:        1,
-	}
-	icmpV6 = icmpFamily{
-		rawNetwork:   "ip6:ipv6-icmp",
-		dgramNetwork: "udp6",
-		listenAddr:   "::",
-		echoType:     ipv6.ICMPTypeEchoRequest,
-		replyType:    ipv6.ICMPTypeEchoReply,
-		proto:        58,
-		isIPv6:       true,
-	}
-	icmpModeMu sync.Mutex
-	icmpListen = func(network, listenAddr string) (icmpPacketConn, error) {
-		return icmp.ListenPacket(network, listenAddr)
-	}
-)
-
-// monitorICMP sends an ICMP echo request and measures round-trip response.
-// Supports both IPv4 and IPv6 targets. The ICMP method (raw socket,
-// unprivileged datagram, or exec fallback) is detected once per address
-// family and cached for subsequent monitors.
-// Returns response in microseconds, or -1 and an error on failure.
-func monitorICMP(ctx context.Context, target string) (int64, error) {
-	ctx, cancel := context.WithTimeout(ctx, 3*time.Second)
-	defer cancel()
-
-	family, ip, err := resolveICMPTarget(ctx, target)
-	if err != nil {
-		return -1, err
-	}
-
-	icmpModeMu.Lock()
-	if family.mode == icmpUntried {
-		family.mode = detectICMPMode(family, icmpListen)
-	}
-	mode := family.mode
-	icmpModeMu.Unlock()
-
-	switch mode {
-	case icmpRaw:
-		return monitorICMPNative(ctx, family.rawNetwork, family, &net.IPAddr{IP: ip})
-	case icmpDatagram:
-		return monitorICMPNative(ctx, family.dgramNetwork, family, &net.UDPAddr{IP: ip})
-	case icmpExecFallback:
-		return monitorICMPExec(ctx, ip.String(), family.isIPv6)
-	default:
-		return -1, errors.New("unsupported ICMP mode")
-	}
-}
-
-// resolveICMPTarget resolves a target hostname or IP to determine the address
-// family and concrete IP address. Prefers IPv4 for dual-stack hostnames.
-func resolveICMPTarget(ctx context.Context, target string) (*icmpFamily, net.IP, error) {
-	if ip := net.ParseIP(target); ip != nil {
-		if ip.To4() != nil {
-			return &icmpV4, ip.To4(), nil
-		}
-		return &icmpV6, ip, nil
-	}
-
-	ips, err := net.DefaultResolver.LookupIP(ctx, "ip", target)
-	if err != nil || len(ips) == 0 {
-		return nil, nil, err
-	}
-	for _, ip := range ips {
-		if v4 := ip.To4(); v4 != nil {
-			return &icmpV4, v4, nil
-		}
-	}
-	return &icmpV6, ips[0], nil
-}
-
-func detectICMPMode(family *icmpFamily, listen func(network, listenAddr string) (icmpPacketConn, error)) icmpMethod {
-	label := "IPv4"
-	if family.isIPv6 {
-		label = "IPv6"
-	}
-
-	conn, err := listen(family.rawNetwork, family.listenAddr)
-	slog.Debug("ICMP raw socket test", "family", label, "err", err)
-	if err == nil {
-		conn.Close()
-		return icmpRaw
-	}
-
-	conn, err = listen(family.dgramNetwork, family.listenAddr)
-	slog.Debug("ICMP datagram socket test", "family", label, "err", err)
-	if err == nil {
-		conn.Close()
-		return icmpDatagram
-	}
-
-	return icmpExecFallback
-}
-
-// monitorICMPNative sends an ICMP echo request using Go's x/net/icmp package.
-func monitorICMPNative(ctx context.Context, network string, family *icmpFamily, dst net.Addr) (int64, error) {
-	conn, err := icmp.ListenPacket(network, family.listenAddr)
-	if err != nil {
-		return -1, err
-	}
-	defer conn.Close()
-
-	return monitorICMPPacket(ctx, conn, family, dst)
-}
-
-func monitorICMPPacket(ctx context.Context, conn net.PacketConn, family *icmpFamily, dst net.Addr) (int64, error) {
-	if err := ctx.Err(); err != nil {
-		return -1, err
-	}
-	// Closing the socket interrupts both reads and writes on cancellation.
-	stop := context.AfterFunc(ctx, func() { _ = conn.Close() })
-	defer stop()
-
-	// Prepare correlation data before starting the round-trip timer. The token
-	// also distinguishes delayed replies after the 16-bit sequence wraps.
-	token := make([]byte, 16)
-	if _, err := rand.Read(token); err != nil {
-		return -1, err
-	}
-	echo := &icmp.Echo{
-		ID:   os.Getpid() & 0xffff,
-		Seq:  int(icmpSequence.Add(1) & 0xffff),
-		Data: token,
-	}
-	// Linux ping sockets replace the Echo ID with their bound port. Darwin
-	// datagram sockets and raw sockets preserve the supplied ID.
-	if local, ok := conn.LocalAddr().(*net.UDPAddr); ok && runtime.GOOS == "linux" {
-		echo.ID = local.Port
-	}
-	targetIP := icmpAddrIP(dst)
-	msg := &icmp.Message{
-		Type: family.echoType,
-		Code: 0,
-		Body: echo,
-	}
-	msgBytes, err := msg.Marshal(nil)
-	if err != nil {
-		return -1, err
-	}
-
-	// Set deadline before sending
-	if err := conn.SetDeadline(time.Now().Add(3 * time.Second)); err != nil {
-		return -1, err
-	}
-
-	buf := make([]byte, 1500)
-	start := time.Now()
-	if _, err := conn.WriteTo(msgBytes, dst); err != nil {
-		return -1, err
-	}
-
-	// Read reply
-	for {
-		n, peer, err := conn.ReadFrom(buf)
-		received := time.Now()
-		if err != nil {
-			return -1, err
-		}
-		if !targetIP.Equal(icmpAddrIP(peer)) {
-			continue
-		}
-
-		reply, err := icmp.ParseMessage(family.proto, buf[:n])
-		if err != nil || reply.Type != family.replyType || reply.Code != 0 {
-			continue
-		}
-
-		body, ok := reply.Body.(*icmp.Echo)
-		if ok && body.ID == echo.ID && body.Seq == echo.Seq && bytes.Equal(body.Data, echo.Data) {
-			return received.Sub(start).Microseconds(), nil
-		}
-		// Keep waiting for our reply without extending the original deadline.
-	}
-}
-
-func icmpAddrIP(addr net.Addr) net.IP {
-	switch addr := addr.(type) {
-	case *net.IPAddr:
-		return addr.IP
-	case *net.UDPAddr:
-		return addr.IP
-	default:
-		return nil
-	}
-}
-
-// pingCommand selects the executable and arguments for the supported agent platforms.
-// The context deadline enforces the timeout: -W has incompatible meanings across
-// Linux, BSD IPv4 ping, and macOS ping6.
-func pingCommand(goos, target string, isIPv6 bool) (string, []string, error) {
-	family := "-4"
-	if isIPv6 {
-		family = "-6"
-	}
-	switch goos {
-	case "windows":
-		return "ping", []string{family, "-n", "1", "-w", "3000", target}, nil
-	case "linux":
-		return "ping", []string{family, "-n", "-c", "1", target}, nil
-	case "darwin", "freebsd", "openbsd":
-		command := "ping"
-		if isIPv6 {
-			command = "ping6"
-		}
-		return command, []string{"-n", "-c", "1", target}, nil
-	default:
-		return "", nil, fmt.Errorf("ping fallback is unsupported on %s", goos)
-	}
-}
-
-// monitorICMPExec falls back to the system ping command. Returns -1 and an error on failure.
-func monitorICMPExec(ctx context.Context, target string, isIPv6 bool) (int64, error) {
-	ctx, cancel := context.WithTimeout(ctx, 3*time.Second)
-	defer cancel()
-	name, args, err := pingCommand(runtime.GOOS, target, isIPv6)
-	if err != nil {
-		return -1, err
-	}
-	cmd := exec.CommandContext(ctx, name, args...)
-	// Keep Unix output and decimal formatting stable. Windows ignores LC_ALL.
-	cmd.Env = append(os.Environ(), "LC_ALL=C")
-	output, err := cmd.Output()
-	if ctx.Err() != nil {
-		return -1, ctx.Err()
-	}
-	if err != nil {
-		return -1, fmt.Errorf("%s failed: %w", name, err)
-	}
-	return parsePingResponse(output)
-}
-
-// parsePingResponse returns the reported RTT, never subprocess execution time.
-// For a bounded value such as Windows' time<1ms, retain the reported upper bound.
-func parsePingResponse(output []byte) (int64, error) {
-	matches := pingTimeRegex.FindSubmatch(output)
-	if len(matches) < 2 {
-		return -1, errors.New("ping output contains no round-trip time")
-	}
-	ms, err := strconv.ParseFloat(strings.ReplaceAll(string(matches[1]), ",", "."), 64)
-	if err != nil || math.IsInf(ms, 0) || ms >= float64(math.MaxInt64)/1000 {
-		return -1, errors.New("invalid round-trip time in ping output")
-	}
-	return int64(math.Round(ms * 1000)), nil
-}

agent/network_monitor_ping_test.go

@@ -1,433 +0,0 @@
-//go:build testing
-
-package agent
-
-import (
-	"context"
-	"errors"
-	"fmt"
-	"net"
-	"os"
-	"path/filepath"
-	"runtime"
-	"testing"
-	"time"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-	"golang.org/x/net/icmp"
-)
-
-type testICMPPacketConn struct{}
-
-func (testICMPPacketConn) Close() error { return nil }
-
-type blockingICMPConn struct {
-	net.PacketConn
-	reading chan struct{}
-}
-
-func (c *blockingICMPConn) WriteTo(p []byte, addr net.Addr) (int, error) {
-	return len(p), nil
-}
-
-func (c *blockingICMPConn) ReadFrom(p []byte) (int, net.Addr, error) {
-	close(c.reading)
-	return c.PacketConn.ReadFrom(p)
-}
-
-func TestMonitorICMPPacketCancellation(t *testing.T) {
-	conn, err := net.ListenPacket("udp4", "127.0.0.1:0")
-	require.NoError(t, err)
-	defer conn.Close()
-	blocking := &blockingICMPConn{PacketConn: conn, reading: make(chan struct{})}
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-	done := make(chan error, 1)
-	go func() {
-		_, err := monitorICMPPacket(ctx, blocking, &icmpV4, conn.LocalAddr())
-		done <- err
-	}()
-	select {
-	case <-blocking.reading:
-	case <-time.After(time.Second):
-		t.Fatal("probe did not begin reading")
-	}
-	cancel()
-	select {
-	case err := <-done:
-		require.Error(t, err)
-	case <-time.After(time.Second):
-		t.Fatal("cancellation did not interrupt the socket read")
-	}
-}
-
-func TestMonitorICMPExecCancellation(t *testing.T) {
-	if runtime.GOOS == "windows" {
-		t.Skip("test uses a POSIX shell stub for ping")
-	}
-	dir := t.TempDir()
-	require.NoError(t, os.WriteFile(filepath.Join(dir, "ping"), []byte("#!/bin/sh\nexec sleep 30\n"), 0o755))
-	t.Setenv("PATH", dir+string(os.PathListSeparator)+os.Getenv("PATH"))
-	ctx, cancel := context.WithTimeout(context.Background(), 100*time.Millisecond)
-	defer cancel()
-	done := make(chan error, 1)
-	go func() {
-		_, err := monitorICMPExec(ctx, "127.0.0.1", false)
-		done <- err
-	}()
-	select {
-	case err := <-done:
-		require.ErrorIs(t, err, context.DeadlineExceeded)
-	case <-time.After(time.Second):
-		t.Fatal("cancellation did not terminate ping")
-	}
-}
-
-func TestPingCommand(t *testing.T) {
-	for _, goos := range []string{"linux", "windows", "darwin", "freebsd", "openbsd"} {
-		for _, ipv6 := range []bool{false, true} {
-			t.Run(fmt.Sprintf("%s/ipv6=%t", goos, ipv6), func(t *testing.T) {
-				target, family := "192.0.2.1", "-4"
-				if ipv6 {
-					target, family = "2001:db8::1", "-6"
-				}
-				name, args, err := pingCommand(goos, target, ipv6)
-				require.NoError(t, err)
-				wantName := "ping"
-				wantArgs := []string{"-n", "-c", "1", target}
-				switch goos {
-				case "windows":
-					wantArgs = []string{family, "-n", "1", "-w", "3000", target}
-				case "linux":
-					wantArgs = append([]string{family}, wantArgs...)
-				default:
-					if ipv6 {
-						wantName = "ping6"
-					}
-				}
-				assert.Equal(t, wantName, name)
-				assert.Equal(t, wantArgs, args)
-			})
-		}
-	}
-	_, _, err := pingCommand("unsupported", "192.0.2.1", false)
-	require.Error(t, err)
-}
-
-func TestParsePingResponse(t *testing.T) {
-	for _, tc := range []struct {
-		name   string
-		output string
-		wantUs int64
-	}{
-		{"linux", "64 bytes from 192.0.2.1: icmp_seq=1 ttl=64 time=12.345 ms", 12345},
-		{"bsd", "64 bytes from 192.0.2.1: icmp_seq=0 ttl=64 time=0.023 ms", 23},
-		{"ipv6", "64 bytes from 2001:db8::1: icmp_seq=0 hlim=64 time=1.234 ms", 1234},
-		{"windows", "Reply from 192.0.2.1: bytes=32 time=12ms TTL=128", 12000},
-		{"windows submillisecond", "Reply from ::1: time<1ms", 1000},
-		{"localized windows", "Antwort von 192.0.2.1: Bytes=32 Zeit=12ms TTL=128", 12000},
-		{"decimal comma", "64 bytes from 192.0.2.1: time=1,234 ms", 1234},
-		{"rounding", "time=0.1236 ms", 124},
-		{"empty", "", -1},
-		{"timeout", "Request timed out.", -1},
-		{"unreachable", "Reply from 192.0.2.1: Destination host unreachable.", -1},
-		{"malformed", "time=oops ms", -1},
-		{"negative", "time=-1 ms", -1},
-		{"overflow", "time=999999999999999999999 ms", -1},
-	} {
-		t.Run(tc.name, func(t *testing.T) {
-			responseUs, err := parsePingResponse([]byte(tc.output))
-			if tc.wantUs < 0 {
-				require.Error(t, err)
-			} else {
-				require.NoError(t, err)
-			}
-			assert.Equal(t, tc.wantUs, responseUs)
-		})
-	}
-}
-
-func TestMonitorICMPExecOutput(t *testing.T) {
-	if runtime.GOOS == "windows" {
-		t.Skip("test uses a POSIX shell stub for ping")
-	}
-	for _, tc := range []struct {
-		name   string
-		output string
-		exit   int
-		wantUs int64
-	}{
-		{"success", "time=1.234 ms", 0, 1234},
-		{"missing RTT", "unrecognized output", 0, -1},
-		{"failed command with RTT", "time=1.234 ms", 1, -1},
-	} {
-		t.Run(tc.name, func(t *testing.T) {
-			dir := t.TempDir()
-			// Also verify an inherited locale cannot override the C locale.
-			script := fmt.Sprintf("#!/bin/sh\n[ \"$LC_ALL\" = C ] || exit 2\nprintf '%%s\\n' '%s'\nexit %d\n", tc.output, tc.exit)
-			require.NoError(t, os.WriteFile(filepath.Join(dir, "ping"), []byte(script), 0o755))
-			t.Setenv("PATH", dir+string(os.PathListSeparator)+os.Getenv("PATH"))
-			t.Setenv("LC_ALL", "de_DE.UTF-8")
-			responseUs, err := monitorICMPExec(t.Context(), "127.0.0.1", false)
-			if tc.wantUs < 0 {
-				require.Error(t, err)
-			} else {
-				require.NoError(t, err)
-			}
-			assert.Equal(t, tc.wantUs, responseUs)
-		})
-	}
-}
-
-type icmpTestReply struct {
-	data []byte
-	peer net.Addr
-}
-
-type scriptedICMPConn struct {
-	net.PacketConn
-	local        net.Addr
-	onWrite      func([]byte, net.Addr)
-	replies      []icmpTestReply
-	reads        int
-	deadlineSets int
-}
-
-func (c *scriptedICMPConn) LocalAddr() net.Addr { return c.local }
-
-func (c *scriptedICMPConn) SetDeadline(deadline time.Time) error {
-	c.deadlineSets++
-	return nil
-}
-
-func (c *scriptedICMPConn) WriteTo(data []byte, dst net.Addr) (int, error) {
-	c.onWrite(data, dst)
-	return len(data), nil
-}
-
-func (c *scriptedICMPConn) ReadFrom(buf []byte) (int, net.Addr, error) {
-	c.reads++
-	if len(c.replies) == 0 {
-		return 0, nil, os.ErrDeadlineExceeded
-	}
-	reply := c.replies[0]
-	c.replies = c.replies[1:]
-	return copy(buf, reply.data), reply.peer, nil
-}
-
-func TestMonitorICMPReplyCorrelation(t *testing.T) {
-	for _, family := range []*icmpFamily{&icmpV4, &icmpV6} {
-		for _, datagram := range []bool{false, true} {
-			network := family.rawNetwork
-			ip, other := net.ParseIP("192.0.2.1"), net.ParseIP("192.0.2.2")
-			if family.isIPv6 {
-				ip, other = net.ParseIP("2001:db8::1"), net.ParseIP("2001:db8::2")
-			}
-			var dst net.Addr = &net.IPAddr{IP: ip}
-			var wrongPeer net.Addr = &net.IPAddr{IP: other}
-			if datagram {
-				network = family.dgramNetwork
-				dst = &net.UDPAddr{IP: ip}
-				wrongPeer = &net.UDPAddr{IP: other}
-			}
-			for _, mismatch := range []string{"source", "id", "sequence", "payload", "type", "code", "malformed"} {
-				for _, eventuallyMatches := range []bool{false, true} {
-					ending := "timeout"
-					if eventuallyMatches {
-						ending = "success"
-					}
-					t.Run(network+"/"+mismatch+"/"+ending, func(t *testing.T) {
-						conn := &scriptedICMPConn{local: &net.IPAddr{IP: net.IPv4zero}}
-						if datagram {
-							conn.local = &net.UDPAddr{Port: 12345}
-							if runtime.GOOS == "linux" {
-								// Deliberately differ from the process ID.
-								conn.local = &net.UDPAddr{Port: (os.Getpid() % 65534) + 1}
-							}
-						}
-						conn.onWrite = func(data []byte, target net.Addr) {
-							require.Equal(t, dst, target)
-							request, err := icmp.ParseMessage(family.proto, data)
-							require.NoError(t, err)
-							echo := request.Body.(*icmp.Echo)
-							expectedID := os.Getpid() & 0xffff
-							if datagram && runtime.GOOS == "linux" {
-								expectedID = conn.local.(*net.UDPAddr).Port
-							}
-							require.Equal(t, expectedID, echo.ID)
-							reply := &icmp.Message{Type: family.replyType, Body: echo}
-							valid, err := reply.Marshal(nil)
-							require.NoError(t, err)
-							peer := dst
-							switch mismatch {
-							case "source":
-								peer = wrongPeer
-							case "id":
-								echo.ID ^= 1
-							case "sequence":
-								echo.Seq ^= 1
-							case "payload":
-								echo.Data[0] ^= 1
-							case "type":
-								reply.Type = family.echoType
-							case "code":
-								reply.Code = 1
-							}
-							invalid, err := reply.Marshal(nil)
-							require.NoError(t, err)
-							if mismatch == "malformed" {
-								invalid = invalid[:2]
-							}
-							conn.replies = []icmpTestReply{{invalid, peer}}
-							if eventuallyMatches {
-								conn.replies = append(conn.replies, icmpTestReply{valid, dst})
-							}
-						}
-						elapsed, err := monitorICMPPacket(context.Background(), conn, family, dst)
-						if eventuallyMatches {
-							require.NoError(t, err)
-							assert.GreaterOrEqual(t, elapsed, int64(0))
-						} else {
-							require.ErrorIs(t, err, os.ErrDeadlineExceeded)
-							assert.Equal(t, int64(-1), elapsed)
-						}
-						assert.Equal(t, 2, conn.reads)
-						assert.Equal(t, 1, conn.deadlineSets)
-					})
-				}
-			}
-		}
-	}
-}
-
-func TestMonitorICMPLoopback(t *testing.T) {
-	for _, family := range []*icmpFamily{&icmpV4, &icmpV6} {
-		for _, network := range []string{family.rawNetwork, family.dgramNetwork} {
-			t.Run(network, func(t *testing.T) {
-				conn, err := icmp.ListenPacket(network, family.listenAddr)
-				if err != nil {
-					t.Skipf("ICMP socket unavailable: %v", err)
-				}
-				defer conn.Close()
-				ip := net.ParseIP("127.0.0.1")
-				if family.isIPv6 {
-					ip = net.ParseIP("::1")
-				}
-				var dst net.Addr = &net.IPAddr{IP: ip}
-				if network == family.dgramNetwork {
-					dst = &net.UDPAddr{IP: ip}
-				}
-				elapsed, err := monitorICMPPacket(context.Background(), conn, family, dst)
-				require.NoError(t, err)
-				assert.GreaterOrEqual(t, elapsed, int64(0))
-			})
-		}
-	}
-}
-
-func TestDetectICMPMode(t *testing.T) {
-	tests := []struct {
-		name         string
-		family       *icmpFamily
-		rawErr       error
-		udpErr       error
-		want         icmpMethod
-		wantNetworks []string
-	}{
-		{
-			name:         "IPv4 prefers raw socket when available",
-			family:       &icmpV4,
-			want:         icmpRaw,
-			wantNetworks: []string{"ip4:icmp"},
-		},
-		{
-			name:         "IPv4 uses datagram when raw unavailable",
-			family:       &icmpV4,
-			rawErr:       errors.New("operation not permitted"),
-			want:         icmpDatagram,
-			wantNetworks: []string{"ip4:icmp", "udp4"},
-		},
-		{
-			name:         "IPv4 falls back to exec when both unavailable",
-			family:       &icmpV4,
-			rawErr:       errors.New("operation not permitted"),
-			udpErr:       errors.New("protocol not supported"),
-			want:         icmpExecFallback,
-			wantNetworks: []string{"ip4:icmp", "udp4"},
-		},
-		{
-			name:         "IPv6 prefers raw socket when available",
-			family:       &icmpV6,
-			want:         icmpRaw,
-			wantNetworks: []string{"ip6:ipv6-icmp"},
-		},
-		{
-			name:         "IPv6 uses datagram when raw unavailable",
-			family:       &icmpV6,
-			rawErr:       errors.New("operation not permitted"),
-			want:         icmpDatagram,
-			wantNetworks: []string{"ip6:ipv6-icmp", "udp6"},
-		},
-		{
-			name:         "IPv6 falls back to exec when both unavailable",
-			family:       &icmpV6,
-			rawErr:       errors.New("operation not permitted"),
-			udpErr:       errors.New("protocol not supported"),
-			want:         icmpExecFallback,
-			wantNetworks: []string{"ip6:ipv6-icmp", "udp6"},
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			calls := make([]string, 0, 2)
-			listen := func(network, listenAddr string) (icmpPacketConn, error) {
-				require.Equal(t, tt.family.listenAddr, listenAddr)
-				calls = append(calls, network)
-				switch network {
-				case tt.family.rawNetwork:
-					if tt.rawErr != nil {
-						return nil, tt.rawErr
-					}
-				case tt.family.dgramNetwork:
-					if tt.udpErr != nil {
-						return nil, tt.udpErr
-					}
-				default:
-					t.Fatalf("unexpected network %q", network)
-				}
-				return testICMPPacketConn{}, nil
-			}
-
-			assert.Equal(t, tt.want, detectICMPMode(tt.family, listen))
-			assert.Equal(t, tt.wantNetworks, calls)
-		})
-	}
-}
-
-func TestResolveICMPTarget(t *testing.T) {
-	t.Run("IPv4 literal", func(t *testing.T) {
-		family, ip, err := resolveICMPTarget(context.Background(), "127.0.0.1")
-		require.NoError(t, err)
-		require.NotNil(t, family)
-		assert.False(t, family.isIPv6)
-		assert.Equal(t, "127.0.0.1", ip.String())
-	})
-
-	t.Run("IPv6 literal", func(t *testing.T) {
-		family, ip, err := resolveICMPTarget(context.Background(), "::1")
-		require.NoError(t, err)
-		require.NotNil(t, family)
-		assert.True(t, family.isIPv6)
-		assert.Equal(t, "::1", ip.String())
-	})
-
-	t.Run("IPv4-mapped IPv6 resolves as IPv4", func(t *testing.T) {
-		family, ip, err := resolveICMPTarget(context.Background(), "::ffff:127.0.0.1")
-		require.NoError(t, err)
-		require.NotNil(t, family)
-		assert.False(t, family.isIPv6)
-		assert.Equal(t, "127.0.0.1", ip.String())
-	})
-}

agent/network_monitor_probe.go

@@ -1,133 +0,0 @@
-package agent
-
-import (
-	"context"
-	"errors"
-	"fmt"
-	"net"
-	"net/http"
-	"time"
-
-	"github.com/henrygd/beszel"
-	"github.com/henrygd/beszel/internal/entities/monitor"
-)
-
-const networkMonitorUserAgent = "Beszel-Agent/" + beszel.Version + " (+https://beszel.dev)"
-
-// monitorProbe performs one check. Errors are recorded as loss by the task runner.
-// Implementations must honor cancellation and bound their execution time.
-type monitorProbe func(context.Context, monitor.Config) (int64, error)
-
-func networkMonitorProbe(client *http.Client) monitorProbe {
-	return func(ctx context.Context, config monitor.Config) (int64, error) {
-		switch config.Protocol {
-		case "icmp":
-			return monitorICMP(ctx, config.Target)
-		case "tcp":
-			return monitorTCP(ctx, config.Target, config.Port)
-		case "http":
-			return monitorHTTP(ctx, client, config.Target)
-		case "dns":
-			return monitorDNS(ctx, config.Target, config.Server)
-		default:
-			return -1, fmt.Errorf("unknown monitor protocol: %s", config.Protocol)
-		}
-	}
-}
-
-// monitorTCP measures connection establishment time, including address fallback
-// but excluding DNS resolution.
-// Returns -1 and an error on failure.
-func monitorTCP(ctx context.Context, target string, port uint16) (int64, error) {
-	ctx, cancel := context.WithTimeout(ctx, 3*time.Second)
-	defer cancel()
-
-	// Resolve DNS first, outside the timing window but within the probe deadline.
-	ips, err := net.DefaultResolver.LookupHost(ctx, target)
-	if err != nil {
-		return -1, err
-	}
-	if len(ips) == 0 {
-		return -1, errors.New("no addresses resolved for TCP monitor")
-	}
-	portString := fmt.Sprintf("%d", port)
-	deadline, _ := ctx.Deadline()
-
-	// Share the remaining probe budget across addresses so an unresponsive
-	// first address cannot consume all the time available for alternatives.
-	start := time.Now()
-	for i, ip := range ips {
-		if err := ctx.Err(); err != nil {
-			return -1, err
-		}
-		dialer := net.Dialer{Timeout: time.Until(deadline) / time.Duration(len(ips)-i)}
-		var conn net.Conn
-		conn, err = dialer.DialContext(ctx, "tcp", net.JoinHostPort(ip, portString))
-		if err != nil {
-			continue
-		}
-		responseUs := time.Since(start).Microseconds()
-		conn.Close()
-		return responseUs, nil
-	}
-	return -1, err
-}
-
-// monitorDNS measures DNS resolution response time in microseconds. If server is
-// non-empty, the lookup is sent to that DNS server (host or host:port, default
-// port 53) instead of the system resolver. Returns -1 and an error on failure.
-func monitorDNS(ctx context.Context, target, server string) (int64, error) {
-	ctx, cancel := context.WithTimeout(ctx, 3*time.Second)
-	defer cancel()
-
-	resolver := net.DefaultResolver
-	if server != "" {
-		resolver = dnsResolverForServer(server)
-	}
-
-	start := time.Now()
-	ips, err := resolver.LookupHost(ctx, target)
-	if err != nil || len(ips) == 0 {
-		return -1, err
-	}
-	return time.Since(start).Microseconds(), nil
-}
-
-// dnsResolverForServer builds a resolver that sends lookups to the given DNS
-// server address instead of the system resolver. server may be a bare host or
-// host:port; when no port is given, the standard DNS port 53 is used.
-func dnsResolverForServer(server string) *net.Resolver {
-	address := server
-	if _, _, err := net.SplitHostPort(server); err != nil {
-		address = net.JoinHostPort(server, "53")
-	}
-	return &net.Resolver{
-		PreferGo: true,
-		Dial: func(ctx context.Context, network, _ string) (net.Conn, error) {
-			var dialer net.Dialer
-			return dialer.DialContext(ctx, network, address)
-		},
-	}
-}
-
-// monitorHTTP measures HTTP GET request response in microseconds. Returns -1 and an error on failure.
-func monitorHTTP(ctx context.Context, client *http.Client, url string) (int64, error) {
-	if client == nil {
-		client = http.DefaultClient
-	}
-	start := time.Now()
-	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
-	if err != nil {
-		return -1, err
-	}
-	req.Header.Set("User-Agent", networkMonitorUserAgent)
-	resp, err := client.Do(req)
-	if err != nil {
-		return -1, err
-	}
-	resp.Body.Close()
-	if resp.StatusCode >= 400 {
-		return -1, fmt.Errorf("HTTP error: %s", resp.Status)
-	}
-	return time.Since(start).Microseconds(), nil
-}

agent/network_monitor_resume.go

@@ -1,88 +0,0 @@
-package agent
-
-import (
-	"sync"
-	"time"
-)
-
-const (
-	monitorResumeHeartbeat = 10 * time.Second
-	// Allow scheduling jitter without mistaking an ordinary tick for resume.
-	monitorResumeGap   = 2 * monitorResumeHeartbeat
-	monitorResumePause = 10 * time.Second
-)
-
-// monitorResumeGuard detects likely suspend/resume using wall time. A long
-// process stall or forward clock adjustment can also trigger the bounded pause.
-// One heartbeat is shared by all configured monitors.
-type monitorResumeGuard struct {
-	mu         sync.Mutex
-	stop       chan struct{}
-	lastTick   time.Time
-	pauseUntil time.Time
-	generation uint32
-}
-
-func (g *monitorResumeGuard) start() {
-	g.mu.Lock()
-	defer g.mu.Unlock()
-	if g.stop != nil {
-		return
-	}
-	stop := make(chan struct{})
-	g.stop = stop
-	g.lastTick = time.Now().Round(0)
-	g.pauseUntil = time.Time{}
-	go func() {
-		ticker := time.NewTicker(monitorResumeHeartbeat)
-		defer ticker.Stop()
-		for {
-			select {
-			case <-stop:
-				return
-			case <-ticker.C:
-				g.mu.Lock()
-				if g.stop == stop {
-					g.observe(time.Now())
-				}
-				g.mu.Unlock()
-			}
-		}
-	}()
-}
-
-func (g *monitorResumeGuard) shutdown() {
-	g.mu.Lock()
-	defer g.mu.Unlock()
-	if g.stop != nil {
-		close(g.stop)
-		g.stop = nil
-		g.generation++
-	}
-}
-
-// observe requires mu. Strip the monotonic component because it can stop during
-// suspend. Read the current time rather than the ticker's queued timestamp.
-func (g *monitorResumeGuard) observe(now time.Time) {
-	now = now.Round(0)
-	if now.Sub(g.lastTick) > monitorResumeGap {
-		g.pauseUntil = now.Add(monitorResumePause)
-		g.generation++
-	}
-	g.lastTick = now
-}
-
-// snapshot also observes time so a probe waking before the heartbeat detects
-// resume itself. A changed generation invalidates probes spanning suspend.
-func (g *monitorResumeGuard) snapshot() (generation uint32, allowed bool) {
-	if g == nil {
-		return 0, true
-	}
-	g.mu.Lock()
-	defer g.mu.Unlock()
-	if g.stop == nil {
-		return g.generation, true
-	}
-	g.observe(time.Now())
-	return g.generation, !g.lastTick.Before(g.pauseUntil)
-}

agent/network_monitor_resume_test.go

@@ -1,121 +0,0 @@
-//go:build testing
-
-package agent
-
-import (
-	"context"
-	"errors"
-	"sync/atomic"
-	"testing"
-	"testing/synctest"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/monitor"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func simulateMonitorSleep(g *monitorResumeGuard) {
-	g.mu.Lock()
-	g.lastTick = time.Now().Add(-time.Hour).Round(0)
-	g.mu.Unlock()
-}
-
-func TestMonitorResumePause(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		var g monitorResumeGuard
-		g.start()
-		defer g.shutdown()
-		generation, allowed := g.snapshot()
-		require.True(t, allowed)
-		// Heartbeats alone must keep the guard current between infrequent probes.
-		time.Sleep(time.Minute)
-		synctest.Wait()
-		steadyGeneration, allowed := g.snapshot()
-		require.True(t, allowed)
-		require.Equal(t, generation, steadyGeneration)
-		// The probe, rather than the heartbeat, must detect this gap.
-		simulateMonitorSleep(&g)
-		next, allowed := g.snapshot()
-		assert.False(t, allowed)
-		assert.NotEqual(t, generation, next)
-		time.Sleep(9 * time.Second)
-		_, allowed = g.snapshot()
-		assert.False(t, allowed)
-		time.Sleep(time.Second)
-		end, allowed := g.snapshot()
-		assert.True(t, allowed)
-		assert.Equal(t, next, end)
-	})
-}
-
-func TestMonitorResumeGuardLifecycle(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		pm := newMonitorManagerWithProbe(func(context.Context, monitor.Config) (int64, error) { return 1, nil })
-		defer pm.Stop()
-		assert.Nil(t, pm.resumeGuard.stop)
-		pm.SyncMonitors([]monitor.Config{{ID: "a", Interval: 3600}, {ID: "b", Interval: 3600}})
-		stop := pm.resumeGuard.stop
-		require.NotNil(t, stop)
-		pm.DeleteMonitor("a")
-		assert.Equal(t, stop, pm.resumeGuard.stop)
-		pm.DeleteMonitor("b")
-		assert.Nil(t, pm.resumeGuard.stop)
-		select {
-		case <-stop:
-		default:
-			t.Fatal("heartbeat was not stopped")
-		}
-		time.Sleep(time.Hour)
-		_, err := pm.UpsertMonitor(monitor.Config{ID: "c", Interval: 3600}, false)
-		require.NoError(t, err)
-		_, allowed := pm.resumeGuard.snapshot()
-		assert.True(t, allowed, "idle time must not trigger a resume pause")
-		pm.SyncMonitors(nil)
-		assert.Nil(t, pm.resumeGuard.stop)
-	})
-}
-
-func TestMonitorResumeDiscardsInflightProbe(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		var g monitorResumeGuard
-		g.start()
-		defer g.shutdown()
-		task := newMonitorTask(monitor.Config{ID: "test"})
-		defer task.cancel()
-		task.resumeGuard = &g
-		result := task.runProbe(func(context.Context, monitor.Config) (int64, error) {
-			simulateMonitorSleep(&g)
-			return 0, errors.New("network not ready")
-		})
-		assert.Nil(t, result)
-		assert.Empty(t, task.history.samples)
-		// Explicit requests may still run during the pause and record real failures.
-		result = task.runProbe(func(context.Context, monitor.Config) (int64, error) {
-			return 0, errors.New("unreachable")
-		})
-		require.NotNil(t, result)
-		assert.Equal(t, 100.0, result.PacketLoss)
-	})
-}
-
-func TestMonitorResumeSkipsScheduledProbes(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		var calls atomic.Int32
-		pm := newMonitorManagerWithProbe(func(context.Context, monitor.Config) (int64, error) {
-			calls.Add(1)
-			return 1, nil
-		})
-		defer pm.Stop()
-		pm.SyncMonitors([]monitor.Config{{ID: "test", Interval: 1}})
-		simulateMonitorSleep(&pm.resumeGuard)
-		pm.resumeGuard.snapshot()
-		time.Sleep(9 * time.Second)
-		synctest.Wait()
-		assert.Zero(t, calls.Load())
-		assert.Empty(t, pm.GetResults(1000))
-		time.Sleep(2 * time.Second)
-		synctest.Wait()
-		assert.Positive(t, calls.Load())
-	})
-}

agent/network_monitor_schedule.go

@@ -1,63 +0,0 @@
-package agent
-
-import (
-	"context"
-	"log/slog"
-	"math/rand"
-	"time"
-)
-
-func (pm *MonitorManager) startMonitor(task *monitorTask) {
-	interval := time.Duration(task.config.Interval) * time.Second
-	if interval < time.Second {
-		interval = 30 * time.Second
-	}
-	delay := getStagger(interval.Milliseconds())
-	slog.Debug("starting monitor task", "target", task.config.Target, "delay", delay, "interval", interval)
-	// Certificate checks piggyback on probe ticks, so they run at most once per
-	// probe interval after they become due.
-	go runMonitorSchedule(task.ctx, interval, delay, func() {
-		if _, allowed := task.resumeGuard.snapshot(); allowed {
-			task.runProbe(pm.probe)
-			task.refreshCert(pm.certCheck)
-		}
-	})
-}
-
-// runMonitorSchedule owns only timing. Checks run serially, and slow checks
-// naturally drop missed ticks rather than building an execution backlog.
-func runMonitorSchedule(ctx context.Context, interval, delay time.Duration, run func()) {
-	timer := time.NewTimer(delay)
-	defer timer.Stop()
-	select {
-	case <-ctx.Done():
-		return
-	case <-timer.C:
-	}
-	if ctx.Err() != nil {
-		return
-	}
-	run()
-	ticker := time.NewTicker(interval)
-	defer ticker.Stop()
-	for {
-		select {
-		case <-ctx.Done():
-			return
-		case <-ticker.C:
-			if ctx.Err() != nil {
-				return
-			}
-			run()
-		}
-	}
-}
-
-// getStagger returns an initial delay between half an interval and one interval.
-func getStagger(intervalMilli int64) time.Duration {
-	delay := rand.Intn(int(intervalMilli))
-	if delay < int(intervalMilli)/2 {
-		delay += int(intervalMilli) / 2
-	}
-	return time.Duration(delay) * time.Millisecond
-}

agent/network_monitor_schedule_test.go

@@ -1,167 +0,0 @@
-//go:build testing
-
-package agent
-
-import (
-	"context"
-	"sync/atomic"
-	"testing"
-	"testing/synctest"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/monitor"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestMonitorScheduleTiming(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		ctx, cancel := context.WithCancel(t.Context())
-		defer cancel()
-		var calls atomic.Int32
-		go runMonitorSchedule(ctx, 10*time.Second, 5*time.Second, func() { calls.Add(1) })
-		synctest.Wait()
-		time.Sleep(4 * time.Second)
-		synctest.Wait()
-		assert.Equal(t, 0, int(calls.Load()))
-		time.Sleep(time.Second)
-		synctest.Wait()
-		assert.Equal(t, 1, int(calls.Load()))
-		time.Sleep(10 * time.Second)
-		synctest.Wait()
-		assert.Equal(t, 2, int(calls.Load()))
-		cancel()
-		synctest.Wait()
-		time.Sleep(time.Minute)
-		synctest.Wait()
-		assert.Equal(t, 2, int(calls.Load()))
-	})
-}
-
-func TestMonitorScheduleSlowProbe(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		ctx, cancel := context.WithCancel(t.Context())
-		defer cancel()
-		var calls atomic.Int32
-		release := make(chan struct{})
-		go runMonitorSchedule(ctx, time.Second, 0, func() {
-			calls.Add(1)
-			select {
-			case <-release:
-			case <-ctx.Done():
-			}
-		})
-		synctest.Wait()
-		assert.Equal(t, 1, int(calls.Load()))
-		time.Sleep(time.Minute)
-		synctest.Wait()
-		assert.Equal(t, 1, int(calls.Load()), "a slow probe must not spawn overlapping checks")
-		close(release)
-		synctest.Wait()
-		assert.Equal(t, 1, int(calls.Load()), "missed intervals must not accumulate a backlog")
-		time.Sleep(time.Second)
-		synctest.Wait()
-		assert.Equal(t, 2, int(calls.Load()))
-		cancel()
-		synctest.Wait()
-	})
-}
-
-func TestMonitorScheduledAndImmediateRequestsShareProbe(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		var calls atomic.Int32
-		release := make(chan struct{})
-		cfg := monitor.Config{ID: "test", Interval: 10}
-		pm := newMonitorManagerWithProbe(func(ctx context.Context, config monitor.Config) (int64, error) {
-			assert.Equal(t, cfg, config)
-			calls.Add(1)
-			<-release
-			return 42, nil
-		})
-		defer pm.Stop()
-		task := newMonitorTask(cfg)
-		pm.monitors[cfg.ID] = task
-		go runMonitorSchedule(task.ctx, 10*time.Second, 0, func() { task.runProbe(pm.probe) })
-		synctest.Wait()
-		results := make(chan *monitor.Result, 2)
-		for range 2 {
-			go func() {
-				result, _ := pm.UpsertMonitor(cfg, true)
-				results <- result
-			}()
-		}
-		synctest.Wait()
-		assert.Equal(t, 1, int(calls.Load()))
-		assert.Empty(t, pm.GetResults(1000), "reading history must not wait for network I/O")
-		close(release)
-		synctest.Wait()
-		first, second := <-results, <-results
-		require.NotNil(t, first)
-		require.NotNil(t, second)
-		assert.Equal(t, int64(42), first.AvgResponse)
-		assert.Equal(t, first, second)
-		assert.NotSame(t, first, second, "callers must not share mutable result pointers")
-		assert.Len(t, task.history.samples, 1)
-		// A later explicit request must still perform a fresh probe.
-		_, err := pm.UpsertMonitor(cfg, true)
-		require.NoError(t, err)
-		assert.Equal(t, 2, int(calls.Load()))
-		assert.Len(t, task.history.samples, 2)
-	})
-}
-
-func TestMonitorReplacementCancelsSharedProbe(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		cfg := monitor.Config{ID: "test", Interval: 10}
-		pm := newMonitorManagerWithProbe(func(ctx context.Context, config monitor.Config) (int64, error) {
-			if config.Interval == 10 {
-				<-ctx.Done()
-				return 0, ctx.Err()
-			}
-			return 30, nil
-		})
-		defer pm.Stop()
-		task := newMonitorTask(cfg)
-		task.history.record(monitorSample{responseUs: 10, timestamp: time.Now()})
-		pm.monitors[cfg.ID] = task
-		results := make(chan *monitor.Result, 2)
-		for range 2 {
-			go func() {
-				result, _ := pm.UpsertMonitor(cfg, true)
-				results <- result
-			}()
-		}
-		synctest.Wait()
-		updated := cfg
-		updated.Interval = 20
-		result, err := pm.UpsertMonitor(updated, true)
-		require.NoError(t, err)
-		require.NotNil(t, result)
-		assert.Equal(t, int64(20), result.AvgResponse)
-		assert.Zero(t, result.PacketLoss)
-		synctest.Wait()
-		assert.Nil(t, <-results)
-		assert.Nil(t, <-results)
-		assert.Len(t, task.history.samples, 1)
-		assert.Len(t, pm.monitors[cfg.ID].history.samples, 2)
-	})
-}
-
-func TestMonitorInjectedProbeTimeoutRecordsLoss(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		pm := newMonitorManagerWithProbe(func(ctx context.Context, _ monitor.Config) (int64, error) {
-			ctx, cancel := context.WithTimeout(ctx, 3*time.Second)
-			defer cancel()
-			<-ctx.Done()
-			return 0, ctx.Err()
-		})
-		defer pm.Stop()
-		start := time.Now()
-		result, err := pm.UpsertMonitor(monitor.Config{ID: "test", Interval: 3600}, true)
-		require.NoError(t, err)
-		require.NotNil(t, result)
-		assert.Equal(t, 3*time.Second, time.Since(start))
-		assert.Equal(t, 100.0, result.PacketLoss)
-		assert.NoError(t, pm.monitors["test"].ctx.Err())
-	})
-}

agent/network_monitor_task.go

@@ -1,191 +0,0 @@
-package agent
-
-import (
-	"context"
-	"log/slog"
-	"sync"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/monitor"
-)
-
-const monitorFailureLogInterval = 5 * time.Minute
-
-// monitorTask coordinates a probe and its history for one immutable configuration.
-type monitorTask struct {
-	config         monitor.Config
-	ctx            context.Context
-	cancel         context.CancelFunc
-	history        *monitorHistory
-	resumeGuard    *monitorResumeGuard
-	runMu          sync.Mutex
-	inflight       *monitorRun
-	lastFailureLog int64 // Unix nanoseconds
-
-	certMu        sync.Mutex
-	cert          *monitor.CertInfo
-	certUnsent    bool // cert has not been included in a stats result yet
-	certChecking  bool
-	nextCertCheck time.Time
-}
-
-type monitorRun struct {
-	done   chan struct{}
-	result *monitor.Result // published by closing done; never mutated afterwards
-}
-
-func newMonitorTask(config monitor.Config) *monitorTask {
-	ctx, cancel := context.WithCancel(context.Background())
-	task := &monitorTask{config: config, ctx: ctx, history: newMonitorHistory()}
-	// Serialize cancellation with publication, so canceled probes cannot enter
-	// history copied into a replacement task.
-	task.cancel = func() {
-		task.runMu.Lock()
-		cancel()
-		task.runMu.Unlock()
-	}
-	return task
-}
-
-func newMonitorTaskFromExisting(config monitor.Config, existing *monitorTask) *monitorTask {
-	task := newMonitorTask(config)
-	if existing != nil {
-		task.history = existing.history.clone()
-		// Keep the last known certificate, but check again soon for the new config.
-		// The hub already stores it, so it is not marked unsent.
-		if config.Target == existing.config.Target {
-			task.cert = existing.certInfo()
-		}
-	}
-	return task
-}
-
-// runProbe shares an in-flight check between scheduled and immediate requests.
-// Every completed check contributes exactly one sample, regardless of how many
-// callers were waiting for it. No task or history lock is held during network I/O.
-func (task *monitorTask) runProbe(probe monitorProbe) *monitor.Result {
-	task.runMu.Lock()
-	if task.ctx.Err() != nil {
-		task.runMu.Unlock()
-		return nil
-	}
-	if run := task.inflight; run != nil {
-		task.runMu.Unlock()
-		select {
-		case <-task.ctx.Done():
-			return nil
-		case <-run.done:
-			if task.ctx.Err() != nil {
-				return nil
-			}
-			return copyMonitorResult(run.result)
-		}
-	}
-	run := &monitorRun{done: make(chan struct{})}
-	task.inflight = run
-	task.runMu.Unlock()
-
-	generation, _ := task.resumeGuard.snapshot()
-	responseUs, err := probe(task.ctx, task.config)
-	var logFailure bool
-	task.runMu.Lock()
-	currentGeneration, _ := task.resumeGuard.snapshot()
-	if task.ctx.Err() == nil && generation == currentGeneration {
-		now := time.Now()
-		if err != nil {
-			responseUs = -1
-			logAt := now.UnixNano()
-			if task.lastFailureLog == 0 || logAt < task.lastFailureLog || logAt-task.lastFailureLog >= int64(monitorFailureLogInterval) {
-				logFailure = true
-				task.lastFailureLog = logAt
-			}
-		} else {
-			task.lastFailureLog = 0
-		}
-		result := task.history.record(monitorSample{responseUs: responseUs, timestamp: now})
-		run.result = &result
-	}
-
-	task.inflight = nil
-	close(run.done)
-	task.runMu.Unlock()
-	if logFailure {
-		slog.Warn("monitor failed", "err", err, "target", task.config.Target, "protocol", task.config.Protocol)
-	}
-	if task.ctx.Err() != nil {
-		return nil
-	}
-	return copyMonitorResult(run.result)
-}
-
-// refreshCert checks the certificate of an HTTPS target when due. A failed
-// check keeps the last known certificate and retries sooner, as does a
-// certificate that expires before the next regular check, so renewals show up
-// quickly. Concurrent callers skip rather than wait, and no lock is held during
-// network I/O.
-func (task *monitorTask) refreshCert(check certChecker) {
-	if check == nil || !certCheckEnabled(task.config) {
-		return
-	}
-	task.certMu.Lock()
-	if task.certChecking || time.Now().Before(task.nextCertCheck) {
-		task.certMu.Unlock()
-		return
-	}
-	task.certChecking = true
-	task.certMu.Unlock()
-
-	info, err := check(task.ctx, task.config.Target)
-
-	task.certMu.Lock()
-	defer task.certMu.Unlock()
-	task.certChecking = false
-	if task.ctx.Err() != nil {
-		return
-	}
-	if err != nil {
-		task.nextCertCheck = time.Now().Add(certCheckRetryInterval)
-		slog.Warn("certificate check failed", "err", err, "target", task.config.Target)
-		return
-	}
-	task.cert = &info
-	task.certUnsent = true
-	now := time.Now()
-	interval := certCheckInterval
-	if time.UnixMilli(info.Expires).Before(now.Add(certCheckInterval)) {
-		interval = certCheckRetryInterval
-	}
-	task.nextCertCheck = now.Add(interval)
-}
-
-// certInfo returns a copy of the latest certificate info, or nil if unknown.
-func (task *monitorTask) certInfo() *monitor.CertInfo {
-	task.certMu.Lock()
-	defer task.certMu.Unlock()
-	if task.cert == nil {
-		return nil
-	}
-	cert := *task.cert
-	return &cert
-}
-
-// takeUnsentCert returns the latest certificate info once after each successful
-// check, so unchanged info is not resent with every stats result.
-func (task *monitorTask) takeUnsentCert() *monitor.CertInfo {
-	task.certMu.Lock()
-	defer task.certMu.Unlock()
-	if !task.certUnsent {
-		return nil
-	}
-	task.certUnsent = false
-	cert := *task.cert
-	return &cert
-}
-
-func copyMonitorResult(result *monitor.Result) *monitor.Result {
-	if result == nil {
-		return nil
-	}
-	copy := *result
-	return &copy
-}

agent/network_monitor_task_test.go

@@ -1,79 +0,0 @@
-//go:build testing
-
-package agent
-
-import (
-	"bytes"
-	"context"
-	"errors"
-	"log/slog"
-	"testing"
-	"testing/synctest"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/monitor"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestMonitorFailureLogCooldown(t *testing.T) {
-	var logs bytes.Buffer
-	previous := slog.Default()
-	slog.SetDefault(slog.New(slog.NewTextHandler(&logs, nil)))
-	t.Cleanup(func() { slog.SetDefault(previous) })
-
-	synctest.Test(t, func(t *testing.T) {
-		task := newMonitorTask(monitor.Config{ID: "test", Target: "example.test", Protocol: "tcp"})
-		defer task.cancel()
-		failure := errors.New("connection refused")
-		probe := func(context.Context, monitor.Config) (int64, error) { return 42, failure }
-		var samples int64
-		check := func(wantLog bool) {
-			t.Helper()
-			logs.Reset()
-			result := task.runProbe(probe)
-			require.NotNil(t, result)
-			samples++
-			assert.Equal(t, samples, result.SampleCount, "suppressed warnings must still record samples")
-			if !wantLog {
-				assert.Empty(t, logs.String())
-			} else {
-				assert.Contains(t, logs.String(), `msg="monitor failed"`)
-				assert.Equal(t, 1, bytes.Count(logs.Bytes(), []byte("\n")))
-			}
-		}
-
-		check(true)
-		check(false)
-		time.Sleep(5*time.Minute - time.Nanosecond)
-		check(false)
-		time.Sleep(time.Nanosecond)
-		check(true)
-		check(false)
-		time.Sleep(5 * time.Minute)
-		check(true)
-		check(false)
-
-		// Recovery clears the cooldown.
-		failure = nil
-		check(false)
-		failure = errors.New("connection refused again")
-		check(true)
-
-		// Another monitor has its own cooldown.
-		other := newMonitorTask(task.config)
-		defer other.cancel()
-		logs.Reset()
-		require.NotNil(t, other.runProbe(probe))
-		assert.Contains(t, logs.String(), `msg="monitor failed"`)
-
-		// A canceled probe must not publish a failure or emit a warning.
-		logs.Reset()
-		result := other.runProbe(func(context.Context, monitor.Config) (int64, error) {
-			other.cancel()
-			return -1, context.Canceled
-		})
-		assert.Nil(t, result)
-		assert.Empty(t, logs.String())
-	})
-}

agent/network_monitor_test.go

@@ -1,590 +0,0 @@
-package agent
-
-import (
-	"context"
-	"encoding/binary"
-	"io"
-	"net"
-	"net/http"
-	"net/http/httptest"
-	"testing"
-	"time"
-
-	"github.com/henrygd/beszel"
-	"github.com/henrygd/beszel/internal/entities/monitor"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-	"golang.org/x/net/dns/dnsmessage"
-)
-
-func TestMonitorManagerGetResultsIncludesHourResponseRange(t *testing.T) {
-	now := time.Now().UTC()
-	task := newMonitorTask(monitor.Config{ID: "monitor-1"})
-	task.history.addSampleLocked(monitorSample{responseUs: 10, timestamp: now.Add(-30 * time.Minute)})
-	task.history.addSampleLocked(monitorSample{responseUs: 20, timestamp: now.Add(-9 * time.Minute)})
-	task.history.addSampleLocked(monitorSample{responseUs: 40, timestamp: now.Add(-5 * time.Minute)})
-	task.history.addSampleLocked(monitorSample{responseUs: 30, timestamp: now.Add(-50 * time.Second)})
-	task.history.addSampleLocked(monitorSample{responseUs: -1, timestamp: now.Add(-30 * time.Second)})
-
-	pm := newMonitorManager()
-	pm.monitors = map[string]*monitorTask{"icmp:example.com": task}
-
-	results := pm.GetResults(uint16(time.Minute / time.Millisecond))
-	result, ok := results["monitor-1"]
-	require.True(t, ok)
-	assert.Equal(t, int64(30), result.AvgResponse)
-	assert.Equal(t, int64(25), result.AvgResponse1h)
-	assert.Equal(t, int64(30), result.MinResponse)
-	assert.Equal(t, int64(10), result.MinResponse1h)
-	assert.Equal(t, int64(30), result.MaxResponse)
-	assert.Equal(t, int64(40), result.MaxResponse1h)
-	assert.Equal(t, 50.0, result.PacketLoss)
-	assert.Equal(t, 20.0, result.PacketLoss1h)
-}
-
-func TestMonitorManagerGetResultsIncludesLossOnlyHourData(t *testing.T) {
-	now := time.Now().UTC()
-	task := newMonitorTask(monitor.Config{ID: "monitor-1"})
-	task.history.addSampleLocked(monitorSample{responseUs: -1, timestamp: now.Add(-30 * time.Second)})
-	task.history.addSampleLocked(monitorSample{responseUs: -1, timestamp: now.Add(-10 * time.Second)})
-
-	pm := newMonitorManager()
-	pm.monitors = map[string]*monitorTask{"icmp:example.com": task}
-
-	results := pm.GetResults(uint16(time.Minute / time.Millisecond))
-	result, ok := results["monitor-1"]
-	require.True(t, ok)
-	assert.Equal(t, int64(0), result.AvgResponse)
-	assert.Equal(t, int64(0), result.AvgResponse1h)
-	assert.Equal(t, int64(0), result.MinResponse)
-	assert.Equal(t, int64(0), result.MinResponse1h)
-	assert.Equal(t, int64(0), result.MaxResponse)
-	assert.Equal(t, int64(0), result.MaxResponse1h)
-	assert.Equal(t, 100.0, result.PacketLoss)
-	assert.Equal(t, 100.0, result.PacketLoss1h)
-}
-
-func TestMonitorConfigResultKeyUsesSyncedID(t *testing.T) {
-	cfg := monitor.Config{ID: "monitor-1", Target: "1.1.1.1", Protocol: "icmp", Interval: 10}
-	assert.Equal(t, "monitor-1", cfg.ID)
-}
-
-func TestMonitorManagerSyncMonitorsSkipsConfigsWithoutStableID(t *testing.T) {
-	validCfg := monitor.Config{ID: "monitor-1", Target: "ignored", Protocol: "noop", Interval: 10}
-	invalidCfg := monitor.Config{Target: "ignored", Protocol: "noop", Interval: 10}
-
-	pm := newMonitorManager()
-	pm.SyncMonitors([]monitor.Config{validCfg, invalidCfg})
-	defer pm.Stop()
-
-	_, validExists := pm.monitors[validCfg.ID]
-	_, invalidExists := pm.monitors[invalidCfg.ID]
-	assert.True(t, validExists)
-	assert.False(t, invalidExists)
-}
-
-func TestMonitorManagerSyncMonitorsStopsRemovedTasksButKeepsExisting(t *testing.T) {
-	keepCfg := monitor.Config{ID: "monitor-1", Target: "ignored", Protocol: "noop", Interval: 10}
-	removeCfg := monitor.Config{ID: "monitor-2", Target: "ignored", Protocol: "noop", Interval: 10}
-
-	keptTask := newMonitorTask(keepCfg)
-	removedTask := newMonitorTask(removeCfg)
-	pm := newMonitorManager()
-	pm.monitors = map[string]*monitorTask{
-		keepCfg.ID:   keptTask,
-		removeCfg.ID: removedTask,
-	}
-
-	pm.SyncMonitors([]monitor.Config{keepCfg})
-
-	assert.Same(t, keptTask, pm.monitors[keepCfg.ID])
-	_, exists := pm.monitors[removeCfg.ID]
-	assert.False(t, exists)
-
-	select {
-	case <-removedTask.ctx.Done():
-	default:
-		t.Fatal("expected removed monitor task to be cancelled")
-	}
-
-	select {
-	case <-keptTask.ctx.Done():
-		t.Fatal("expected existing monitor task to remain active")
-	default:
-	}
-}
-
-func TestMonitorManagerSyncMonitorsRestartsChangedConfig(t *testing.T) {
-	originalCfg := monitor.Config{ID: "monitor-1", Target: "ignored-a", Protocol: "noop", Interval: 10}
-	updatedCfg := monitor.Config{ID: "monitor-1", Target: "ignored-b", Protocol: "noop", Interval: 10}
-	originalTask := newMonitorTask(originalCfg)
-	pm := newMonitorManager()
-	pm.monitors = map[string]*monitorTask{
-		originalCfg.ID: originalTask,
-	}
-
-	pm.SyncMonitors([]monitor.Config{updatedCfg})
-	defer pm.Stop()
-
-	restartedTask := pm.monitors[updatedCfg.ID]
-	assert.NotSame(t, originalTask, restartedTask)
-	assert.Equal(t, updatedCfg, restartedTask.config)
-
-	select {
-	case <-originalTask.ctx.Done():
-	default:
-		t.Fatal("expected changed monitor task to be cancelled")
-	}
-}
-
-func TestMonitorManagerApplySyncUpsertRunsImmediatelyAndReturnsResult(t *testing.T) {
-	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		w.WriteHeader(http.StatusNoContent)
-	}))
-	defer server.Close()
-
-	pm := &MonitorManager{
-		monitors: make(map[string]*monitorTask),
-		probe:    networkMonitorProbe(server.Client()),
-	}
-
-	resp, err := pm.HandleSyncRequest(monitor.SyncRequest{
-		Action: monitor.SyncActionUpsert,
-		Config: monitor.Config{ID: "monitor-1", Target: server.URL, Protocol: "http", Interval: 10},
-		RunNow: true,
-	})
-	defer pm.Stop()
-
-	require.NoError(t, err)
-	assert.GreaterOrEqual(t, resp.Result.AvgResponse, int64(0))
-	assert.Equal(t, 0.0, resp.Result.PacketLoss)
-	assert.Equal(t, 0.0, resp.Result.PacketLoss1h)
-
-	task := pm.monitors["monitor-1"]
-	require.NotNil(t, task)
-	task.history.mu.Lock()
-	defer task.history.mu.Unlock()
-	require.Len(t, task.history.samples, 1)
-}
-
-func TestMonitorManagerUpsertMonitorKeepsHistoryWhenOnlyIntervalChanges(t *testing.T) {
-	originalCfg := monitor.Config{ID: "monitor-1", Target: "1.1.1.1", Protocol: "icmp", Interval: 10}
-	updatedCfg := monitor.Config{ID: "monitor-1", Target: "1.1.1.1", Protocol: "icmp", Interval: 30}
-	now := time.Now().UTC()
-
-	existingTask := newMonitorTask(originalCfg)
-	existingTask.history.addSampleLocked(monitorSample{responseUs: 12, timestamp: now.Add(-50 * time.Minute)})
-	existingTask.history.addSampleLocked(monitorSample{responseUs: 24, timestamp: now.Add(-30 * time.Second)})
-
-	pm := newMonitorManager()
-	pm.monitors = map[string]*monitorTask{originalCfg.ID: existingTask}
-
-	result, err := pm.UpsertMonitor(updatedCfg, false)
-	defer pm.Stop()
-
-	require.NoError(t, err)
-	assert.Nil(t, result)
-
-	updatedTask := pm.monitors[updatedCfg.ID]
-	require.NotNil(t, updatedTask)
-	assert.NotSame(t, existingTask, updatedTask)
-	assert.Equal(t, updatedCfg, updatedTask.config)
-
-	updatedTask.history.mu.Lock()
-	defer updatedTask.history.mu.Unlock()
-	require.Len(t, updatedTask.history.samples, 1)
-	assert.Equal(t, int64(24), updatedTask.history.samples[0].responseUs)
-
-	agg := updatedTask.history.aggregateLocked(time.Hour, now)
-	require.True(t, agg.hasData())
-	assert.Equal(t, int64(2), agg.totalCount)
-	assert.Equal(t, int64(2), agg.successCount)
-	assert.Equal(t, int64(18), agg.avgResponse())
-
-	select {
-	case <-existingTask.ctx.Done():
-	default:
-		t.Fatal("expected original monitor task to be cancelled")
-	}
-}
-
-func TestMonitorManagerApplySyncDeleteRemovesTask(t *testing.T) {
-	config := monitor.Config{ID: "monitor-1", Target: "1.1.1.1", Protocol: "icmp", Interval: 10}
-	task := newMonitorTask(config)
-	pm := newMonitorManager()
-	pm.monitors = map[string]*monitorTask{config.ID: task}
-
-	_, err := pm.HandleSyncRequest(monitor.SyncRequest{
-		Action: monitor.SyncActionDelete,
-		Config: monitor.Config{ID: config.ID},
-	})
-
-	require.NoError(t, err)
-	_, exists := pm.monitors[config.ID]
-	assert.False(t, exists)
-
-	select {
-	case <-task.ctx.Done():
-	default:
-		t.Fatal("expected deleted monitor task to be cancelled")
-	}
-}
-
-func TestMonitorManagerGetRandomDelay(t *testing.T) {
-	for i := 1000; i < 360_000; i += 1000 {
-		delay := getStagger(int64(i))
-		assert.GreaterOrEqual(t, delay, time.Duration(i/2)*time.Millisecond)
-		assert.LessOrEqual(t, delay, time.Duration(i)*time.Millisecond)
-	}
-}
-
-func TestMonitorHTTP(t *testing.T) {
-	t.Run("success", func(t *testing.T) {
-		server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-			assert.Equal(t, "Beszel-Agent/"+beszel.Version+" (+https://beszel.dev)", r.Header.Get("User-Agent"))
-			w.WriteHeader(http.StatusNoContent)
-		}))
-		defer server.Close()
-
-		responseUs, err := monitorHTTP(context.Background(), server.Client(), server.URL)
-		require.NoError(t, err)
-		assert.GreaterOrEqual(t, responseUs, int64(0))
-	})
-
-	t.Run("server error", func(t *testing.T) {
-		server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-			http.Error(w, "boom", http.StatusInternalServerError)
-		}))
-		defer server.Close()
-
-		responseUs, err := monitorHTTP(context.Background(), server.Client(), server.URL)
-		assert.Equal(t, int64(-1), responseUs)
-		require.Error(t, err)
-	})
-}
-
-func TestMonitorTCP(t *testing.T) {
-	t.Run("success", func(t *testing.T) {
-		listener, err := net.Listen("tcp", "127.0.0.1:0")
-		require.NoError(t, err)
-		defer listener.Close()
-
-		accepted := make(chan struct{})
-		go func() {
-			defer close(accepted)
-			conn, err := listener.Accept()
-			if err == nil {
-				_ = conn.Close()
-			}
-		}()
-
-		port := uint16(listener.Addr().(*net.TCPAddr).Port)
-		responseUs, err := monitorTCP(context.Background(), "127.0.0.1", port)
-		require.NoError(t, err)
-		assert.GreaterOrEqual(t, responseUs, int64(0))
-		<-accepted
-	})
-
-	t.Run("connection failure", func(t *testing.T) {
-		listener, err := net.Listen("tcp", "127.0.0.1:0")
-		require.NoError(t, err)
-
-		port := uint16(listener.Addr().(*net.TCPAddr).Port)
-		require.NoError(t, listener.Close())
-
-		responseUs, err := monitorTCP(context.Background(), "127.0.0.1", port)
-		assert.Equal(t, int64(-1), responseUs)
-		require.Error(t, err)
-	})
-}
-
-func TestMonitorTCPAddressFallback(t *testing.T) {
-	for _, tc := range []struct {
-		name string
-		ips  []string
-		loss bool
-	}{
-		{"first address fails", []string{"127.0.0.2", "127.0.0.1"}, false},
-		{"first address succeeds", []string{"127.0.0.1", "127.0.0.2"}, false},
-		{"all addresses fail", []string{"127.0.0.2", "127.0.0.3"}, true},
-	} {
-		t.Run(tc.name, func(t *testing.T) {
-			listener, err := net.Listen("tcp4", "127.0.0.1:0")
-			require.NoError(t, err)
-			defer listener.Close()
-			original := net.DefaultResolver
-			net.DefaultResolver = tcpMonitorTestResolver(tc.ips)
-			defer func() { net.DefaultResolver = original }()
-
-			// Verify the resolver preserves the intended order, so success cannot
-			// accidentally bypass the failed first address in the regression case.
-			ips, err := net.DefaultResolver.LookupHost(t.Context(), "tcp-monitor.invalid.")
-			require.NoError(t, err)
-			require.Equal(t, tc.ips, ips)
-			responseUs, err := monitorTCP(t.Context(), "tcp-monitor.invalid.", uint16(listener.Addr().(*net.TCPAddr).Port))
-			if tc.loss {
-				require.Error(t, err)
-				assert.Equal(t, int64(-1), responseUs)
-			} else {
-				require.NoError(t, err)
-				assert.GreaterOrEqual(t, responseUs, int64(0))
-			}
-		})
-	}
-}
-
-// tcpMonitorTestResolver supplies multiple A records without external DNS.
-func tcpMonitorTestResolver(ips []string) *net.Resolver {
-	return &net.Resolver{PreferGo: true, Dial: func(ctx context.Context, network, address string) (net.Conn, error) {
-		client, server := net.Pipe()
-		go func() {
-			defer server.Close()
-			// net.Resolver uses TCP framing when its connection is not a PacketConn.
-			var size uint16
-			if err := binary.Read(server, binary.BigEndian, &size); err != nil {
-				return
-			}
-			packet := make([]byte, size)
-			if _, err := io.ReadFull(server, packet); err != nil {
-				return
-			}
-			var msg dnsmessage.Message
-			if err := msg.Unpack(packet); err != nil {
-				return
-			}
-			msg.Header.Response = true
-			msg.Header.RecursionAvailable = true
-			for _, question := range msg.Questions {
-				if question.Type != dnsmessage.TypeA {
-					continue
-				}
-				for _, ip := range ips {
-					msg.Answers = append(msg.Answers, dnsmessage.Resource{
-						Header: dnsmessage.ResourceHeader{Name: question.Name, Type: dnsmessage.TypeA, Class: dnsmessage.ClassINET},
-						Body:   &dnsmessage.AResource{A: [4]byte(net.ParseIP(ip).To4())},
-					})
-				}
-			}
-			packet, err := msg.Pack()
-			if err != nil {
-				return
-			}
-			response := binary.BigEndian.AppendUint16(nil, uint16(len(packet)))
-			_, _ = server.Write(append(response, packet...))
-		}()
-		return client, nil
-	}}
-}
-
-// udpDNSTestServer starts a UDP server on loopback that answers A queries with the
-// given IPs, and returns its listen address (host:port).
-func udpDNSTestServer(t *testing.T, ips []string) string {
-	t.Helper()
-	conn, err := net.ListenUDP("udp", &net.UDPAddr{IP: net.ParseIP("127.0.0.1")})
-	require.NoError(t, err)
-	t.Cleanup(func() { conn.Close() })
-
-	go func() {
-		buf := make([]byte, 512)
-		for {
-			n, addr, err := conn.ReadFromUDP(buf)
-			if err != nil {
-				return
-			}
-			var msg dnsmessage.Message
-			if err := msg.Unpack(buf[:n]); err != nil {
-				continue
-			}
-			msg.Header.Response = true
-			msg.Header.RecursionAvailable = true
-			for _, question := range msg.Questions {
-				if question.Type != dnsmessage.TypeA {
-					continue
-				}
-				for _, ip := range ips {
-					msg.Answers = append(msg.Answers, dnsmessage.Resource{
-						Header: dnsmessage.ResourceHeader{Name: question.Name, Type: dnsmessage.TypeA, Class: dnsmessage.ClassINET},
-						Body:   &dnsmessage.AResource{A: [4]byte(net.ParseIP(ip).To4())},
-					})
-				}
-			}
-			packet, err := msg.Pack()
-			if err != nil {
-				continue
-			}
-			_, _ = conn.WriteToUDP(packet, addr)
-		}
-	}()
-
-	return conn.LocalAddr().String()
-}
-
-func TestMonitorDNS(t *testing.T) {
-	t.Run("success", func(t *testing.T) {
-		responseUs, err := monitorDNS(context.Background(), "localhost", "")
-		require.NoError(t, err)
-		assert.GreaterOrEqual(t, responseUs, int64(0))
-	})
-
-	t.Run("lookup failure", func(t *testing.T) {
-		responseUs, err := monitorDNS(context.Background(), "", "")
-		assert.Equal(t, int64(-1), responseUs)
-		require.Error(t, err)
-	})
-
-	t.Run("custom server", func(t *testing.T) {
-		serverAddr := udpDNSTestServer(t, []string{"192.0.2.10"})
-		responseUs, err := monitorDNS(context.Background(), "example.test.", serverAddr)
-		require.NoError(t, err)
-		assert.GreaterOrEqual(t, responseUs, int64(0))
-	})
-
-	t.Run("custom server without port defaults to 53", func(t *testing.T) {
-		resolver := dnsResolverForServer("127.0.0.1")
-		conn, err := resolver.Dial(context.Background(), "udp", "")
-		require.NoError(t, err)
-		defer conn.Close()
-		assert.Equal(t, "127.0.0.1:53", conn.RemoteAddr().String())
-	})
-
-	t.Run("custom server unreachable", func(t *testing.T) {
-		responseUs, err := monitorDNS(context.Background(), "example.test.", "127.0.0.1:1")
-		assert.Equal(t, int64(-1), responseUs)
-		require.Error(t, err)
-	})
-}
-
-func TestMonitorManagerCancelsActiveProbe(t *testing.T) {
-	for _, action := range []string{"stop", "delete", "upsert", "sync replace", "sync remove"} {
-		t.Run(action, func(t *testing.T) {
-			started := make(chan struct{})
-			canceled := make(chan struct{})
-			release := make(chan struct{})
-			server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-				close(started)
-				select {
-				case <-r.Context().Done():
-					close(canceled)
-				case <-release:
-				}
-			}))
-			defer server.Close()
-			defer close(release)
-			pm := newMonitorManager()
-			defer pm.Stop()
-			cfg := monitor.Config{ID: "test", Protocol: "http", Target: server.URL, Interval: 3600}
-			task := newMonitorTask(cfg)
-			// Seed history to ensure a canceled RunNow does not return an old result.
-			task.history.addSampleLocked(monitorSample{responseUs: 123, timestamp: time.Now()})
-			pm.monitors[cfg.ID] = task
-			done := make(chan *monitor.Result, 1)
-			go func() {
-				result, _ := pm.UpsertMonitor(cfg, true)
-				done <- result
-			}()
-			select {
-			case <-started:
-			case <-time.After(time.Second):
-				t.Fatal("probe did not start")
-			}
-			updated := cfg
-			updated.Interval--
-			switch action {
-			case "stop":
-				pm.Stop()
-			case "delete":
-				pm.DeleteMonitor(cfg.ID)
-			case "upsert":
-				_, err := pm.UpsertMonitor(updated, false)
-				require.NoError(t, err)
-			case "sync replace":
-				pm.SyncMonitors([]monitor.Config{updated})
-			case "sync remove":
-				pm.SyncMonitors(nil)
-			}
-			select {
-			case <-canceled:
-			case <-time.After(time.Second):
-				t.Fatal("active HTTP request was not canceled")
-			}
-			select {
-			case result := <-done:
-				assert.Nil(t, result)
-			case <-time.After(time.Second):
-				t.Fatal("RunNow did not return after cancellation")
-			}
-			task.history.mu.Lock()
-			assert.Len(t, task.history.samples, 1, "cancellation must not record packet loss")
-			task.history.mu.Unlock()
-		})
-	}
-}
-
-func TestMonitorResolutionCancellation(t *testing.T) {
-	for _, protocol := range []string{"tcp", "dns", "icmp"} {
-		t.Run(protocol, func(t *testing.T) {
-			started := make(chan struct{}, 1)
-			original := net.DefaultResolver
-			net.DefaultResolver = &net.Resolver{PreferGo: true, Dial: func(ctx context.Context, network, address string) (net.Conn, error) {
-				select {
-				case started <- struct{}{}:
-				default:
-				}
-				<-ctx.Done()
-				return nil, ctx.Err()
-			}}
-			defer func() { net.DefaultResolver = original }()
-			ctx, cancel := context.WithCancel(context.Background())
-			defer cancel()
-			done := make(chan error, 1)
-			go func() {
-				var err error
-				switch protocol {
-				case "tcp":
-					_, err = monitorTCP(ctx, "monitor-cancellation.invalid.", 80)
-				case "dns":
-					_, err = monitorDNS(ctx, "monitor-cancellation.invalid.", "")
-				case "icmp":
-					_, err = monitorICMP(ctx, "monitor-cancellation.invalid.")
-				}
-				done <- err
-			}()
-			select {
-			case <-started:
-			case <-time.After(time.Second):
-				t.Fatal("lookup did not start")
-			}
-			cancel()
-			select {
-			case err := <-done:
-				require.Error(t, err)
-			case <-time.After(time.Second):
-				t.Fatal("lookup did not cancel")
-			}
-		})
-	}
-}
-
-func TestMonitorProbeTimeoutRecordsLoss(t *testing.T) {
-	release := make(chan struct{})
-	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		select {
-		case <-r.Context().Done():
-		case <-release:
-		}
-	}))
-	defer server.Close()
-	defer close(release)
-	pm := newMonitorManager()
-	pm.probe = networkMonitorProbe(&http.Client{Timeout: 20 * time.Millisecond})
-	task := newMonitorTask(monitor.Config{ID: "timeout", Protocol: "http", Target: server.URL})
-	defer task.cancel()
-
-	result := task.runProbe(pm.probe)
-	require.NotNil(t, result)
-	assert.Equal(t, 100.0, result.PacketLoss)
-	assert.Equal(t, 100.0, result.PacketLoss1h)
-	require.Len(t, task.history.samples, 1)
-	assert.Equal(t, int64(-1), task.history.samples[0].responseUs)
-	assert.NoError(t, task.ctx.Err(), "a probe timeout must not cancel the task")
-}

agent/package_updates.go

@@ -1,310 +0,0 @@
-package agent
-
-import (
-	"bufio"
-	"context"
-	"errors"
-	"log/slog"
-	"os"
-	"os/exec"
-	"path/filepath"
-	"runtime"
-	"slices"
-	"strings"
-	"sync"
-	"time"
-
-	"github.com/henrygd/beszel/agent/utils"
-)
-
-const (
-	defaultPackageUpdatesInterval = time.Hour
-	packageUpdatesTimeout         = 5 * time.Minute
-	// pacmanSyncInterval limits how often checkupdates downloads fresh sync
-	// databases. Checks in between reuse the last synced copy.
-	pacmanSyncInterval = 12 * time.Hour
-)
-
-// packageUpdatesCheck returns [total] or [total, security] pending package updates.
-type packageUpdatesCheck func(ctx context.Context) ([]uint16, error)
-
-// packageUpdatesManager periodically checks the host package manager for pending
-// updates in the background and caches the result, so checks never delay metrics.
-type packageUpdatesManager struct {
-	sync.Mutex
-	check     packageUpdatesCheck
-	interval  time.Duration
-	counts    []uint16
-	checkedAt time.Time
-	running   bool
-}
-
-// newPackageUpdatesManager returns nil if disabled or no supported package manager
-// is found. Agents running in a container are skipped because the container's
-// package database is not the host's. dataDir holds pacman's private sync databases.
-func newPackageUpdatesManager(dataDir string) *packageUpdatesManager {
-	if runtime.GOOS != "linux" || runningInContainer() {
-		return nil
-	}
-	interval := defaultPackageUpdatesInterval
-	if env, exists := utils.GetEnv("PACKAGE_UPDATES_INTERVAL"); exists {
-		duration, err := time.ParseDuration(env)
-		switch {
-		case err == nil && duration == 0:
-			return nil
-		case err == nil && duration > 0:
-			interval = duration
-		default:
-			slog.Warn("Invalid PACKAGE_UPDATES_INTERVAL", "value", env)
-		}
-	}
-	name, check := detectPackageManager(dataDir)
-	if check == nil {
-		return nil
-	}
-	slog.Debug("Package updates", "manager", name, "interval", interval)
-	return &packageUpdatesManager{check: check, interval: interval}
-}
-
-// get returns the last cached counts and starts a background check if they are stale.
-func (pm *packageUpdatesManager) get(now time.Time) []uint16 {
-	pm.Lock()
-	defer pm.Unlock()
-	if !pm.running && (pm.checkedAt.IsZero() || now.Sub(pm.checkedAt) >= pm.interval) {
-		pm.running = true
-		go pm.refresh()
-	}
-	return pm.counts
-}
-
-func (pm *packageUpdatesManager) refresh() {
-	ctx, cancel := context.WithTimeout(context.Background(), packageUpdatesTimeout)
-	defer cancel()
-	counts, err := pm.check(ctx)
-	if err != nil {
-		slog.Debug("Package updates check failed", "err", err)
-		counts = nil
-	}
-	pm.Lock()
-	pm.counts = counts
-	pm.checkedAt = time.Now()
-	pm.running = false
-	pm.Unlock()
-}
-
-func runningInContainer() bool {
-	for _, path := range []string{"/.dockerenv", "/run/.containerenv"} {
-		if _, err := os.Stat(path); err == nil {
-			return true
-		}
-	}
-	return false
-}
-
-func detectPackageManager(dataDir string) (string, packageUpdatesCheck) {
-	switch {
-	case commandExists("apt-get"):
-		return "apt", checkApt
-	case commandExists("dnf"):
-		return "dnf", checkDnf
-	case commandExists("zypper"):
-		return "zypper", checkZypper
-	case commandExists("checkupdates"):
-		return "pacman", newPacmanCheck(dataDir)
-	case commandExists("apk"):
-		return "apk", checkApk
-	}
-	return "", nil
-}
-
-func commandExists(name string) bool {
-	_, err := exec.LookPath(name)
-	return err == nil
-}
-
-// runPackageCommand runs a read-only package manager command and returns stdout.
-// okCodes lists non-zero exit codes that still mean success.
-func runPackageCommand(ctx context.Context, okCodes []int, name string, args ...string) (string, error) {
-	return runPackageCommandEnv(ctx, nil, okCodes, name, args...)
-}
-
-// runPackageCommandEnv is runPackageCommand with extra environment variables.
-func runPackageCommandEnv(ctx context.Context, env []string, okCodes []int, name string, args ...string) (string, error) {
-	cmd := exec.CommandContext(ctx, name, args...)
-	cmd.Env = append(os.Environ(), "LC_ALL=C")
-	cmd.Env = append(cmd.Env, env...)
-	// checkupdates is a shell script, so a timeout kills only the script and its
-	// children can keep stdout open. WaitDelay stops Output from waiting on them.
-	cmd.WaitDelay = 10 * time.Second
-	out, err := cmd.Output()
-	if exitErr, ok := errors.AsType[*exec.ExitError](err); ok && slices.Contains(okCodes, exitErr.ExitCode()) {
-		return string(out), nil
-	}
-	return string(out), err
-}
-
-// checkApt simulates a full upgrade against the current package lists.
-// It never refreshes the lists; apt-daily or the user does that.
-func checkApt(ctx context.Context) ([]uint16, error) {
-	out, err := runPackageCommand(ctx, nil, "apt-get", "-s", "dist-upgrade")
-	if err != nil {
-		return nil, err
-	}
-	total, security := parseAptSimulate(out)
-	return []uint16{total, security}, nil
-}
-
-// checkDnf uses the system metadata cache only (-C), so it never downloads metadata.
-func checkDnf(ctx context.Context) ([]uint16, error) {
-	out, err := runPackageCommand(ctx, []int{100}, "dnf", "-q", "-C", "check-update")
-	if err != nil {
-		return nil, err
-	}
-	total := parseDnfCheckUpdate(out)
-	out, err = runPackageCommand(ctx, []int{100}, "dnf", "-q", "-C", "check-update", "--security")
-	if err != nil {
-		return []uint16{total}, nil
-	}
-	return []uint16{total, parseDnfCheckUpdate(out)}, nil
-}
-
-func checkZypper(ctx context.Context) ([]uint16, error) {
-	out, err := runPackageCommand(ctx, nil, "zypper", "--no-refresh", "-q", "list-updates")
-	if err != nil {
-		return nil, err
-	}
-	total := parseZypperTable(out)
-	out, err = runPackageCommand(ctx, nil, "zypper", "--no-refresh", "-q", "list-patches", "--category", "security")
-	if err != nil {
-		return []uint16{total}, nil
-	}
-	return []uint16{total, parseZypperTable(out)}, nil
-}
-
-// newPacmanCheck uses checkupdates (pacman-contrib), which syncs a private copy of
-// the databases and never touches pacman's own. The copy lives in dataDir because
-// the systemd unit's ProtectSystem=strict makes the default /tmp location read-only.
-// It syncs every pacmanSyncInterval and uses the existing copy (-n) in between.
-// Local upgrades show up right away since checkupdates links the live local DB.
-// Exit code 2 means no updates.
-func newPacmanCheck(dataDir string) packageUpdatesCheck {
-	var env []string
-	var syncDir string
-	if dataDir != "" {
-		dbPath := filepath.Join(dataDir, "checkup-db")
-		env = []string{"CHECKUPDATES_DB=" + dbPath}
-		syncDir = filepath.Join(dbPath, "sync")
-	}
-	// checks never overlap (packageUpdatesManager.running), so no lock is needed
-	var lastSync time.Time
-	return func(ctx context.Context) ([]uint16, error) {
-		// -n with a missing database reports no updates rather than failing,
-		// so always sync first and whenever the private copy is missing
-		sync := lastSync.IsZero() || time.Since(lastSync) >= pacmanSyncInterval
-		if !sync && syncDir != "" {
-			if _, err := os.Stat(syncDir); err != nil {
-				sync = true
-			}
-		}
-		var args []string
-		if !sync {
-			args = append(args, "-n")
-		}
-		out, err := runPackageCommandEnv(ctx, env, []int{2}, "checkupdates", args...)
-		if err != nil {
-			return nil, err
-		}
-		if sync {
-			lastSync = time.Now()
-		}
-		return []uint16{parsePacmanCheckUpdates(out)}, nil
-	}
-}
-
-func checkApk(ctx context.Context) ([]uint16, error) {
-	out, err := runPackageCommand(ctx, nil, "apk", "--no-network", "-u", "list")
-	if err != nil {
-		return nil, err
-	}
-	return []uint16{parseApkUpgradable(out)}, nil
-}
-
-// parseAptSimulate counts upgrades in `apt-get -s` output. Upgrade lines look like
-// "Inst libc6 [2.35-0ubuntu3.4] (2.35-0ubuntu3.15 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])".
-// New dependencies have no "[old version]" and are not counted.
-func parseAptSimulate(out string) (total, security uint16) {
-	scanner := bufio.NewScanner(strings.NewReader(out))
-	for scanner.Scan() {
-		line := scanner.Text()
-		fields := strings.Fields(line)
-		if len(fields) < 4 || fields[0] != "Inst" || !strings.HasPrefix(fields[2], "[") {
-			continue
-		}
-		total++
-		start := strings.IndexByte(line, '(')
-		end := strings.IndexByte(line, ')')
-		if start >= 0 && end > start && strings.Contains(line[start:end], "-security") {
-			security++
-		}
-	}
-	return total, security
-}
-
-// parseDnfCheckUpdate counts "name.arch version repo" lines, stopping at the
-// obsoletes section so obsoleted packages are not counted twice.
-func parseDnfCheckUpdate(out string) (count uint16) {
-	scanner := bufio.NewScanner(strings.NewReader(out))
-	for scanner.Scan() {
-		line := scanner.Text()
-		if strings.HasPrefix(line, "Obsoleting") {
-			break
-		}
-		fields := strings.Fields(line)
-		if len(fields) == 3 && strings.Contains(fields[0], ".") {
-			count++
-		}
-	}
-	return count
-}
-
-// parseZypperTable counts the data rows of a zypper table (the lines after the
-// "---+---" separator).
-func parseZypperTable(out string) (count uint16) {
-	inTable := false
-	scanner := bufio.NewScanner(strings.NewReader(out))
-	for scanner.Scan() {
-		line := scanner.Text()
-		switch {
-		case !inTable:
-			inTable = strings.HasPrefix(line, "--") && strings.Contains(line, "-+-")
-		case strings.Contains(line, "|"):
-			count++
-		default:
-			return count
-		}
-	}
-	return count
-}
-
-// parsePacmanCheckUpdates counts "name old -> new" lines.
-func parsePacmanCheckUpdates(out string) (count uint16) {
-	scanner := bufio.NewScanner(strings.NewReader(out))
-	for scanner.Scan() {
-		if strings.Contains(scanner.Text(), " -> ") {
-			count++
-		}
-	}
-	return count
-}
-
-// parseApkUpgradable counts lines of `apk -u list`, which look like
-// "musl-1.2.5-r3 aarch64 {musl} (MIT) [upgradable from: musl-1.2.5-r0]".
-func parseApkUpgradable(out string) (count uint16) {
-	scanner := bufio.NewScanner(strings.NewReader(out))
-	for scanner.Scan() {
-		if strings.Contains(scanner.Text(), "[upgradable from:") {
-			count++
-		}
-	}
-	return count
-}

agent/package_updates_test.go

@@ -1,196 +0,0 @@
-//go:build testing
-
-package agent
-
-import (
-	"context"
-	"errors"
-	"os"
-	"path/filepath"
-	"runtime"
-	"strings"
-	"testing"
-	"time"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func readPackageUpdatesTestData(t *testing.T, name string) string {
-	t.Helper()
-	data, err := os.ReadFile(filepath.Join("test-data", "package_updates", name))
-	require.NoError(t, err)
-	return string(data)
-}
-
-// Test data files are real command outputs captured in containers.
-
-func TestParseAptSimulate(t *testing.T) {
-	tests := []struct {
-		file            string
-		total, security uint16
-	}{
-		{"apt_debian12.txt", 44, 5},
-		{"apt_ubuntu2204.txt", 58, 45},
-	}
-	for _, tt := range tests {
-		t.Run(tt.file, func(t *testing.T) {
-			total, security := parseAptSimulate(readPackageUpdatesTestData(t, tt.file))
-			assert.Equal(t, tt.total, total)
-			assert.Equal(t, tt.security, security)
-		})
-	}
-
-	t.Run("new dependencies and trailing brackets", func(t *testing.T) {
-		out := `Inst linux-image-6.8.0-50-generic (6.8.0-50.51 Ubuntu:24.04/noble-updates, Ubuntu:24.04/noble-security [amd64])
-Inst linux-image-generic [6.8.0-49.49] (6.8.0-50.50 Ubuntu:24.04/noble-updates, Ubuntu:24.04/noble-security [amd64])
-Inst gcc-12-base [12.3.0-1ubuntu1~22.04] (12.3.0-1ubuntu1~22.04.3 Ubuntu:22.04/jammy-updates [arm64]) [libstdc++6:arm64 libgcc-s1:arm64 ]
-Conf linux-image-generic (6.8.0-50.50 Ubuntu:24.04/noble-updates, Ubuntu:24.04/noble-security [amd64])
-Remv oldpkg [1.0]`
-		total, security := parseAptSimulate(out)
-		assert.Equal(t, uint16(2), total)
-		assert.Equal(t, uint16(1), security)
-	})
-
-	t.Run("no updates", func(t *testing.T) {
-		total, security := parseAptSimulate("Reading package lists...\n0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.\n")
-		assert.Zero(t, total)
-		assert.Zero(t, security)
-	})
-}
-
-func TestParseDnfCheckUpdate(t *testing.T) {
-	tests := []struct {
-		file  string
-		count uint16
-	}{
-		{"dnf4_rocky9_check_update.txt", 110},
-		{"dnf4_rocky9_check_update_security.txt", 53},
-		{"dnf5_fedora42_check_update.txt", 20},
-		{"dnf5_fedora42_check_update_security.txt", 5},
-	}
-	for _, tt := range tests {
-		t.Run(tt.file, func(t *testing.T) {
-			assert.Equal(t, tt.count, parseDnfCheckUpdate(readPackageUpdatesTestData(t, tt.file)))
-		})
-	}
-
-	t.Run("obsoletes section and notices", func(t *testing.T) {
-		out := `
-kernel.x86_64                     5.14.0-503.el9          baseos
-Security: kernel-core-5.14.0-427.el9.x86_64 is an installed security update
-Obsoleting Packages
-grub2-tools.x86_64                1:2.06-80.el9           baseos
-    grub2-tools.x86_64            1:2.06-77.el9           @baseos
-`
-		assert.Equal(t, uint16(1), parseDnfCheckUpdate(out))
-	})
-}
-
-func TestParseZypperTable(t *testing.T) {
-	tests := []struct {
-		file  string
-		count uint16
-	}{
-		{"zypper_leap155_list_updates.txt", 22},
-		{"zypper_leap155_list_patches_security.txt", 4},
-		{"zypper_leap156_list_updates_none.txt", 0},
-	}
-	for _, tt := range tests {
-		t.Run(tt.file, func(t *testing.T) {
-			assert.Equal(t, tt.count, parseZypperTable(readPackageUpdatesTestData(t, tt.file)))
-		})
-	}
-}
-
-func TestParsePacmanCheckUpdates(t *testing.T) {
-	assert.Equal(t, uint16(4), parsePacmanCheckUpdates(readPackageUpdatesTestData(t, "pacman_checkupdates.txt")))
-	assert.Zero(t, parsePacmanCheckUpdates(""))
-}
-
-func TestParseApkUpgradable(t *testing.T) {
-	assert.Equal(t, uint16(10), parseApkUpgradable(readPackageUpdatesTestData(t, "apk_alpine320_list_upgradable.txt")))
-	assert.Zero(t, parseApkUpgradable(""))
-}
-
-func TestPackageUpdatesManagerCaching(t *testing.T) {
-	calls := make(chan struct{}, 10)
-	result := []uint16{3, 1}
-	var resultErr error
-	pm := &packageUpdatesManager{
-		interval: time.Hour,
-		check: func(context.Context) ([]uint16, error) {
-			calls <- struct{}{}
-			return result, resultErr
-		},
-	}
-	waitIdle := func() {
-		require.Eventually(t, func() bool {
-			pm.Lock()
-			defer pm.Unlock()
-			return !pm.running
-		}, time.Second, time.Millisecond)
-	}
-
-	now := time.Now()
-	// first call starts a background check and returns nothing yet
-	assert.Nil(t, pm.get(now))
-	waitIdle()
-	assert.Len(t, calls, 1)
-
-	// cached result within interval, no new check
-	assert.Equal(t, []uint16{3, 1}, pm.get(now.Add(time.Minute)))
-	assert.Len(t, calls, 1)
-
-	// stale after interval: returns cached value and refreshes in background
-	result, resultErr = nil, errors.New("boom")
-	assert.Equal(t, []uint16{3, 1}, pm.get(now.Add(2*time.Hour)))
-	waitIdle()
-	assert.Len(t, calls, 2)
-
-	// failed check clears the counts
-	assert.Nil(t, pm.get(time.Now()))
-}
-
-func TestPacmanCheckSync(t *testing.T) {
-	if runtime.GOOS == "windows" {
-		t.Skip("requires a shell script on PATH")
-	}
-	binDir := t.TempDir()
-	dataDir := t.TempDir()
-	logFile := filepath.Join(binDir, "calls.log")
-	// fake checkupdates logs its args and db path, and creates the sync dir when syncing
-	script := `#!/bin/sh
-echo "args=[$*] db=$CHECKUPDATES_DB" >> ` + logFile + `
-[ "$1" = "-n" ] || mkdir -p "$CHECKUPDATES_DB/sync"
-echo "linux 6.1-1 -> 6.2-1"
-`
-	require.NoError(t, os.WriteFile(filepath.Join(binDir, "checkupdates"), []byte(script), 0o755))
-	t.Setenv("PATH", binDir+string(os.PathListSeparator)+os.Getenv("PATH"))
-
-	check := newPacmanCheck(dataDir)
-	dbPath := filepath.Join(dataDir, "checkup-db")
-	readCalls := func() []string {
-		data, err := os.ReadFile(logFile)
-		require.NoError(t, err)
-		return strings.Split(strings.TrimSpace(string(data)), "\n")
-	}
-
-	// first check syncs
-	counts, err := check(context.Background())
-	require.NoError(t, err)
-	assert.Equal(t, []uint16{1}, counts)
-	// later checks reuse the synced copy
-	_, err = check(context.Background())
-	require.NoError(t, err)
-	// a missing private copy forces a sync
-	require.NoError(t, os.RemoveAll(dbPath))
-	_, err = check(context.Background())
-	require.NoError(t, err)
-
-	assert.Equal(t, []string{
-		"args=[] db=" + dbPath,
-		"args=[-n] db=" + dbPath,
-		"args=[] db=" + dbPath,
-	}, readCalls())
-}

agent/sensors.go

@@ -5,9 +5,7 @@ import (
 	"errors"
 	"fmt"
 	"log/slog"
-	"os"
 	"path"
-	"path/filepath"
 	"runtime"
 	"strconv"
 	"strings"
@@ -34,8 +32,6 @@ type SensorConfig struct {
 	isBlacklist    bool
 	hasWildcards   bool
 	skipCollection bool
-	skipGPU        bool
-	sensorShadow   string
 	firstRun       bool
 }
 
@@ -45,14 +41,13 @@ func (a *Agent) newSensorConfig() *SensorConfig {
 	sensorsEnvVal, sensorsSet := utils.GetEnv("SENSORS")
 	skipCollection := sensorsSet && sensorsEnvVal == ""
 	sensorsTimeout, _ := utils.GetEnv("SENSORS_TIMEOUT")
-	skipGPU, _ := utils.GetEnv("SKIP_GPU")
 
-	return a.newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal, sensorsTimeout, skipCollection, skipGPU == "true")
+	return a.newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal, sensorsTimeout, skipCollection)
 }
 
 // newSensorConfigWithEnv creates a SensorConfig with the provided environment variables
 // sensorsSet indicates if the SENSORS environment variable was explicitly set (even to empty string)
-func (a *Agent) newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal, sensorsTimeout string, skipCollection, skipGPU bool) *SensorConfig {
+func (a *Agent) newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal, sensorsTimeout string, skipCollection bool) *SensorConfig {
 	timeout := 2 * time.Second
 	if sensorsTimeout != "" {
 		if d, err := time.ParseDuration(sensorsTimeout); err == nil {
@@ -67,7 +62,6 @@ func (a *Agent) newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal,
 		primarySensor:  primarySensor,
 		timeout:        timeout,
 		skipCollection: skipCollection,
-		skipGPU:        skipGPU,
 		firstRun:       true,
 		sensors:        make(map[string]struct{}),
 	}
@@ -79,19 +73,6 @@ func (a *Agent) newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal,
 			common.EnvKey, common.EnvMap{common.HostSysEnvKey: sysSensors},
 		)
 	}
-	if skipGPU && runtime.GOOS == "linux" {
-		// gopsutil reads every temp*_input before results can be filtered, so
-		// point it at a shadow tree built from the effective sysfs root instead.
-		if shadow, err := buildNonGpuSysShadow(effectiveSysRoot(config.context)); err == nil {
-			slog.Info("SKIP_GPU enabled, using non-GPU sensor sysfs shadow", "path", shadow)
-			config.sensorShadow = shadow
-			config.context = context.WithValue(config.context,
-				common.EnvKey, common.EnvMap{common.HostSysEnvKey: shadow},
-			)
-		} else {
-			slog.Warn("SKIP_GPU sensor shadow unavailable, falling back to post-read filtering", "err", err)
-		}
-	}
 
 	// handle blacklist
 	if strings.HasPrefix(sensorsEnvVal, "-") {
@@ -168,9 +149,6 @@ func (a *Agent) updateTemperatures(systemStats *system.Stats) {
 		if !isValidSensor(sensorName, a.sensorConfig) {
 			continue
 		}
-		if a.sensorConfig.skipGPU && isGpuSensorKey(sensorName) {
-			continue
-		}
 		// set dashboard temperature
 		switch a.sensorConfig.primarySensor {
 		case "":
@@ -267,102 +245,3 @@ func scaleTemperature(temp float64) float64 {
 	}
 	return scaled100
 }
-
-// effectiveSysRoot mirrors gopsutil's HostSys lookup, which lives in its
-// internal package: context override, then HOST_SYS env, then /sys.
-func effectiveSysRoot(ctx context.Context) string {
-	if envMap, ok := ctx.Value(common.EnvKey).(common.EnvMap); ok {
-		if v := envMap[common.HostSysEnvKey]; v != "" {
-			return v
-		}
-	}
-	if v := os.Getenv("HOST_SYS"); v != "" {
-		return v
-	}
-	return "/sys"
-}
-
-func (config *SensorConfig) cleanupSensorShadow() {
-	if config.sensorShadow == "" {
-		return
-	}
-	if err := os.RemoveAll(config.sensorShadow); err != nil {
-		slog.Warn("Error removing sensor sysfs shadow", "path", config.sensorShadow, "err", err)
-		return
-	}
-	config.sensorShadow = ""
-}
-
-func (a *Agent) cleanupSensorShadow() {
-	if a.sensorConfig != nil {
-		a.sensorConfig.cleanupSensorShadow()
-	}
-}
-
-func isGpuThermalZone(zoneType string) bool {
-	zoneType = strings.ToLower(strings.TrimSpace(zoneType))
-	return isGpuChipName(zoneType) || strings.Contains(zoneType, "gpu")
-}
-
-// buildNonGpuSysShadow links non-GPU sensor directories into a temp dir. Only
-// static chip names and thermal-zone types are read; no sensor values are touched.
-func buildNonGpuSysShadow(sysRoot string) (string, error) {
-	shadow, err := os.MkdirTemp("", "beszel-sensors-*")
-	if err != nil {
-		return "", err
-	}
-	shadowHwmon := filepath.Join(shadow, "class", "hwmon")
-	if err := os.MkdirAll(shadowHwmon, 0o755); err != nil {
-		os.RemoveAll(shadow)
-		return "", err
-	}
-	entries, err := os.ReadDir(filepath.Join(sysRoot, "class", "hwmon"))
-	if err != nil && !os.IsNotExist(err) {
-		os.RemoveAll(shadow)
-		return "", err
-	}
-	for _, entry := range entries {
-		chipDir := filepath.Join(sysRoot, "class", "hwmon", entry.Name())
-		// Some hwmon devices expose name under device/ (gopsutil's CentOS fallback).
-		name, ok := utils.ReadStringFileOK(filepath.Join(chipDir, "name"))
-		if !ok {
-			name, ok = utils.ReadStringFileOK(filepath.Join(chipDir, "device", "name"))
-		}
-		if !ok || isGpuChipName(name) {
-			continue
-		}
-		if err := os.Symlink(chipDir, filepath.Join(shadowHwmon, entry.Name())); err != nil {
-			os.RemoveAll(shadow)
-			return "", err
-		}
-	}
-
-	thermalEntries, err := os.ReadDir(filepath.Join(sysRoot, "class", "thermal"))
-	if err != nil {
-		if os.IsNotExist(err) {
-			return shadow, nil
-		}
-		os.RemoveAll(shadow)
-		return "", err
-	}
-	shadowThermal := filepath.Join(shadow, "class", "thermal")
-	if err := os.MkdirAll(shadowThermal, 0o755); err != nil {
-		os.RemoveAll(shadow)
-		return "", err
-	}
-	for _, entry := range thermalEntries {
-		if !strings.HasPrefix(entry.Name(), "thermal_zone") {
-			continue
-		}
-		zoneDir := filepath.Join(sysRoot, "class", "thermal", entry.Name())
-		zoneType, ok := utils.ReadStringFileOK(filepath.Join(zoneDir, "type"))
-		if !ok || isGpuThermalZone(zoneType) {
-			continue
-		}
-		if err := os.Symlink(zoneDir, filepath.Join(shadowThermal, entry.Name())); err != nil {
-			os.RemoveAll(shadow)
-			return "", err
-		}
-	}
-	return shadow, nil
-}

agent/sensors_test.go

@@ -5,8 +5,6 @@ package agent
 import (
 	"context"
 	"fmt"
-	"os"
-	"path/filepath"
 	"testing"
 	"time"
 
@@ -330,7 +328,7 @@ func TestNewSensorConfigWithEnv(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			result := agent.newSensorConfigWithEnv(tt.primarySensor, tt.sysSensors, tt.sensors, tt.sensorsTimeout, tt.skipCollection, false)
+			result := agent.newSensorConfigWithEnv(tt.primarySensor, tt.sysSensors, tt.sensors, tt.sensorsTimeout, tt.skipCollection)
 
 			// Check primary sensor
 			assert.Equal(t, tt.expectedConfig.primarySensor, result.primarySensor)
@@ -622,143 +620,3 @@ func TestUpdateTemperaturesSkipsOnTimeout(t *testing.T) {
 	assert.Equal(t, 0.0, agent.systemInfo.DashboardTemp)
 	assert.Equal(t, map[string]float64{}, stats.Temperatures)
 }
-
-func TestIsGpuSensorKey(t *testing.T) {
-	for _, key := range []string{"xe", "XE_temp1", "amdgpu_edge", "NVIDIA"} {
-		assert.True(t, isGpuSensorKey(key), key)
-	}
-	for _, key := range []string{"coretemp_core_0", "acpitz", "xen_temp", "myxe", ""} {
-		assert.False(t, isGpuSensorKey(key), key)
-	}
-}
-
-func TestSkipGpuSensorShadow(t *testing.T) {
-	sysRoot := t.TempDir()
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon0", "name"), "coretemp\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon0", "temp1_input"), "55000\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon1", "name"), "xe\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon1", "temp1_input"), "48000\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "thermal", "thermal_zone0", "type"), "cpu-thermal\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "thermal", "thermal_zone0", "temp"), "55000\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "thermal", "thermal_zone1", "type"), "gpu\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "thermal", "thermal_zone1", "temp"), "48000\n")
-
-	shadow, err := buildNonGpuSysShadow(sysRoot)
-	require.NoError(t, err)
-	t.Cleanup(func() { os.RemoveAll(shadow) })
-
-	assert.FileExists(t, filepath.Join(shadow, "class", "hwmon", "hwmon0", "temp1_input"))
-	assert.NoFileExists(t, filepath.Join(shadow, "class", "hwmon", "hwmon1"))
-	assert.FileExists(t, filepath.Join(shadow, "class", "thermal", "thermal_zone0", "temp"))
-	assert.NoFileExists(t, filepath.Join(shadow, "class", "thermal", "thermal_zone1"))
-}
-
-func TestSkipGpuSensorShadowDeviceName(t *testing.T) {
-	sysRoot := t.TempDir()
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon0", "device", "name"), "coretemp\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon0", "device", "temp1_input"), "55000\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon1", "device", "name"), "xe\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon1", "device", "temp1_input"), "48000\n")
-
-	shadow, err := buildNonGpuSysShadow(sysRoot)
-	require.NoError(t, err)
-	t.Cleanup(func() { os.RemoveAll(shadow) })
-
-	assert.FileExists(t, filepath.Join(shadow, "class", "hwmon", "hwmon0", "device", "temp1_input"))
-	assert.NoFileExists(t, filepath.Join(shadow, "class", "hwmon", "hwmon1"))
-}
-
-func TestSkipGpuSensorShadowKeepsThermalZonesWithoutNonGpuHwmon(t *testing.T) {
-	sysRoot := t.TempDir()
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon0", "name"), "xe\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon0", "temp1_input"), "48000\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "thermal", "thermal_zone0", "type"), "cpu-thermal\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "thermal", "thermal_zone0", "temp"), "55000\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "thermal", "thermal_zone1", "type"), "gpu\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "thermal", "thermal_zone1", "temp"), "48000\n")
-
-	shadow, err := buildNonGpuSysShadow(sysRoot)
-	require.NoError(t, err)
-	t.Cleanup(func() { os.RemoveAll(shadow) })
-
-	hwmonTemps, err := filepath.Glob(filepath.Join(shadow, "class", "hwmon", "hwmon*", "temp*_input"))
-	require.NoError(t, err)
-	assert.Empty(t, hwmonTemps)
-	assert.FileExists(t, filepath.Join(shadow, "class", "thermal", "thermal_zone0", "temp"))
-	assert.NoFileExists(t, filepath.Join(shadow, "class", "thermal", "thermal_zone1"))
-}
-
-func TestNewSensorConfigSkipGpuWiresShadow(t *testing.T) {
-	t.Setenv("SKIP_GPU", "true")
-
-	agent := &Agent{}
-	config := agent.newSensorConfig()
-
-	assert.True(t, config.skipGPU)
-	envMap, ok := config.context.Value(common.EnvKey).(common.EnvMap)
-	require.True(t, ok, "SKIP_GPU should point the sensor context at a sysfs shadow")
-	shadow, ok := envMap[common.HostSysEnvKey]
-	require.True(t, ok)
-	assert.DirExists(t, filepath.Join(shadow, "class", "hwmon"))
-	assert.Equal(t, shadow, config.sensorShadow)
-	config.cleanupSensorShadow()
-	assert.NoDirExists(t, shadow)
-	assert.Empty(t, config.sensorShadow)
-}
-
-func TestSkipGpuShadowUsesSysSensorsRoot(t *testing.T) {
-	sysRoot := t.TempDir()
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon0", "name"), "coretemp\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon0", "temp1_input"), "55000\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon1", "name"), "xe\n")
-	writeFile(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon1", "temp1_input"), "48000\n")
-
-	agent := &Agent{}
-	config := agent.newSensorConfigWithEnv("", sysRoot, "", "", false, true)
-	t.Cleanup(config.cleanupSensorShadow)
-
-	envMap, ok := config.context.Value(common.EnvKey).(common.EnvMap)
-	require.True(t, ok, "SKIP_GPU should point the sensor context at a sysfs shadow")
-	shadow, ok := envMap[common.HostSysEnvKey]
-	require.True(t, ok)
-	require.NotEqual(t, sysRoot, shadow, "shadow must not be the SYS_SENSORS tree itself")
-
-	target, err := os.Readlink(filepath.Join(shadow, "class", "hwmon", "hwmon0"))
-	require.NoError(t, err)
-	assert.Equal(t, filepath.Join(sysRoot, "class", "hwmon", "hwmon0"), target)
-	assert.NoFileExists(t, filepath.Join(shadow, "class", "hwmon", "hwmon1"))
-}
-
-func TestUpdateTemperaturesSkipGpu(t *testing.T) {
-	originalGetSensorTemps := getSensorTemps
-	t.Cleanup(func() {
-		getSensorTemps = originalGetSensorTemps
-	})
-	getSensorTemps = func(ctx context.Context) ([]sensors.TemperatureStat, error) {
-		return []sensors.TemperatureStat{
-			{SensorKey: "coretemp_core_0", Temperature: 55},
-			{SensorKey: "XE", Temperature: 48},
-		}, nil
-	}
-
-	newAgent := func(skipGPU bool) *Agent {
-		agent := &Agent{
-			systemInfo: system.Info{},
-			sensorConfig: &SensorConfig{
-				context: context.Background(),
-				timeout: 2 * time.Second,
-				sensors: map[string]struct{}{},
-				skipGPU: skipGPU,
-			},
-		}
-		return agent
-	}
-
-	stats := &system.Stats{}
-	newAgent(true).updateTemperatures(stats)
-	assert.Equal(t, map[string]float64{"coretemp_core_0": 55}, stats.Temperatures)
-
-	stats = &system.Stats{}
-	newAgent(false).updateTemperatures(stats)
-	assert.Len(t, stats.Temperatures, 2)
-}

agent/smart.go

@@ -931,6 +931,9 @@ func (sm *SmartManager) parseSmartForSata(output []byte, deviceType string) (boo
 		if parsed, ok := smart.ParseSmartRawValueString(attr.Raw.String); ok {
 			rawValue = parsed
 		}
+		if smartData.SmartStatus == "PASSED" && rawValue > 0 && (attr.ID == 5 || attr.ID == 197 || attr.ID == 198) {
+			smartData.SmartStatus = "WARNING"
+		}
 		smartAttr := &smart.SmartAttribute{
 			ID:         attr.ID,
 			Name:       attr.Name,

agent/smart_test.go

@@ -7,6 +7,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"strconv"
 	"testing"
 
 	"github.com/henrygd/beszel/internal/entities/smart"
@@ -89,6 +90,27 @@ func TestParseSmartForSata(t *testing.T) {
 	}
 }
 
+func TestParseSmartForSataWarnsForCriticalAttributes(t *testing.T) {
+	for _, attrID := range []int{5, 197, 198} {
+		t.Run("attribute "+strconv.Itoa(attrID), func(t *testing.T) {
+			jsonPayload := []byte(fmt.Sprintf(`{
+				"smartctl": {"exit_status": 0},
+				"device": {"name": "/dev/sda", "type": "sat"},
+				"model_name": "Example",
+				"serial_number": "WARNING%d",
+				"smart_status": {"passed": true},
+				"temperature": {"current": 30},
+				"ata_smart_attributes": {"table": [{"id": %d, "raw": {"value": 1, "string": "1"}}]}
+			}`, attrID, attrID))
+
+			sm := &SmartManager{SmartDataMap: make(map[string]*smart.SmartData)}
+			hasData, _ := sm.parseSmartForSata(jsonPayload, "")
+			require.True(t, hasData)
+			assert.Equal(t, "WARNING", sm.SmartDataMap[fmt.Sprintf("WARNING%d", attrID)].SmartStatus)
+		})
+	}
+}
+
 func TestParseSmartForSataPreservesFailedAndUnknownStatus(t *testing.T) {
 	for _, test := range []struct {
 		name        string

agent/storage_pool.go

@@ -1,512 +0,0 @@
-package agent
-
-import (
-	"errors"
-	"log/slog"
-	"os/exec"
-	"strings"
-	"sync"
-	"time"
-
-	"github.com/henrygd/beszel/agent/btrfs"
-	"github.com/henrygd/beszel/agent/zfs"
-	"github.com/henrygd/beszel/internal/entities/system"
-	zfsentity "github.com/henrygd/beszel/internal/entities/zfs"
-)
-
-// zfsDatasetUsage holds usage values for a ZFS dataset mountpoint.
-type zfsDatasetUsage struct {
-	used  uint64
-	avail uint64
-}
-
-// datasetUsageRefreshInterval controls how often `zfs list` is re-run for the
-// mountpoint usage map. Dataset inventory changes rarely.
-const datasetUsageRefreshInterval = 5 * time.Minute
-
-// poolStatsRefreshInterval controls how often `zpool list` is re-run for pool
-// capacity. Health and I/O are read from procfs on Linux, so the utility only
-// needs to refresh slow-moving space accounting.
-const poolStatsRefreshInterval = time.Minute
-
-// btrfsFilesystems is the btrfs source; overridable in tests.
-var btrfsFilesystems = btrfs.Filesystems
-
-type poolKernelSample struct {
-	nread  uint64
-	nwrite uint64
-	at     time.Time
-}
-
-// StoragePoolManager combines independent backend inventories. Metrics and
-// dataset usage require the agent lock; GetDetail is safe for concurrent calls.
-type StoragePoolManager struct {
-	backends       []*poolBackend
-	detailInterval time.Duration
-}
-
-// poolBackend owns one backend's collectors and caches. Collector functions
-// are immutable after construction and may run concurrently for metrics/details.
-type poolBackend struct {
-	name           string
-	poolStatsFn    func() ([]zfs.PoolStat, error)       // capacity/health source
-	datasetsFn     func() ([]zfs.Dataset, error)        // dataset inventory source
-	kernelStatsFn  func() ([]zfs.PoolKernelStat, error) // procfs pool state/I/O source
-	poolStatusesFn func() ([]zfs.PoolStatus, error)     // scrub/vdev detail source
-
-	// Utility-backed caches below are refreshed in the background after the
-	// first collection, so cacheMu guards them against those goroutines.
-	cacheMu        sync.Mutex
-	poolData       []zfs.PoolStat // cached pool inventory (TTL below)
-	lastPoolStats  time.Time
-	poolRefreshing bool
-
-	datasetUsage     map[string]zfsDatasetUsage // mountpoint -> usage
-	lastUsageRefresh time.Time
-	usageRefreshing  bool
-
-	kernelSamples map[string]poolKernelSample
-
-	// Detail data (pools, vdevs, scrub, datasets) is cached and refreshed on
-	// an interval. Accessed from handler goroutines, so it is mutex-protected.
-	detailMu          sync.Mutex
-	detail            *zfsentity.ZfsData
-	lastDetailRefresh time.Time
-	detailFailed      bool
-}
-
-func newStoragePoolManager() *StoragePoolManager {
-	return &StoragePoolManager{
-		backends:       []*poolBackend{newZfsBackend(), newBtrfsBackend()},
-		detailInterval: time.Hour,
-	}
-}
-
-func newZfsBackend() *poolBackend {
-	return &poolBackend{
-		name:           "zfs",
-		poolStatsFn:    optionalPoolSource(zfs.PoolStats),
-		datasetsFn:     optionalPoolSource(zfs.Datasets),
-		kernelStatsFn:  optionalPoolSource(zfs.PoolKernelStats),
-		poolStatusesFn: optionalPoolSource(zfs.PoolStatuses),
-	}
-}
-
-func newBtrfsBackend() *poolBackend {
-	return &poolBackend{
-		name:           "btrfs",
-		poolStatsFn:    btrfsSource(btrfsPoolStats),
-		kernelStatsFn:  btrfsSource(btrfsKernelStats),
-		poolStatusesFn: btrfsSource(btrfsPoolStatuses),
-	}
-}
-
-// datasets is optional: only backends that expose datasets provide a collector.
-func (b *poolBackend) datasets() ([]zfs.Dataset, error) {
-	if b.datasetsFn == nil {
-		return nil, nil
-	}
-	return b.datasetsFn()
-}
-
-// A missing utility/interface is a successfully observed absent backend.
-func optionalPoolSource[T any](source func() ([]T, error)) func() ([]T, error) {
-	return func() ([]T, error) {
-		items, err := source()
-		if errors.Is(err, zfs.ErrNoZfs) || errors.Is(err, exec.ErrNotFound) || errors.Is(err, errors.ErrUnsupported) {
-			return nil, nil
-		}
-		return items, err
-	}
-}
-
-func btrfsSource[T any](convert func(btrfs.Filesystem) T) func() ([]T, error) {
-	return func() ([]T, error) {
-		filesystems, err := optionalPoolSource(btrfsFilesystems)()
-		if err != nil {
-			return nil, err
-		}
-		items := make([]T, 0, len(filesystems))
-		for _, fs := range filesystems {
-			items = append(items, convert(fs))
-		}
-		return items, nil
-	}
-}
-
-// Update refreshes systemStats.ZfsPools with the latest pool data. I/O
-// throughput and health come from inexpensive kernel kstats on Linux. Pool
-// capacity and dataset usage come from separately cached utility calls. The
-// pool map is empty when both backends are absent.
-func (m *StoragePoolManager) Update(systemStats *system.Stats) {
-	// Rebuild the combined map so successful pool removals clear old samples.
-	systemStats.ZfsPools = nil
-	for _, backend := range m.backends {
-		backend.updateBackendStats(systemStats)
-	}
-}
-
-func (b *poolBackend) updateBackendStats(systemStats *system.Stats) {
-	pools := b.poolStats()
-	if len(pools) == 0 {
-		b.kernelSamples = nil
-		return
-	}
-
-	kernelStats, ioRates := b.kernelStats()
-
-	if systemStats.ZfsPools == nil {
-		systemStats.ZfsPools = make(map[string]*system.ZfsPool, len(pools))
-	}
-	for i := range pools {
-		pool := &pools[i]
-		// Full precision, matching the dataset values below; the frontend
-		// formats any magnitude.
-		stats := &system.ZfsPool{
-			DisplayName: pool.DisplayName,
-			Raw:         pool.Raw,
-			Total:       float64(pool.Size) / (1024 * 1024 * 1024),
-			Used:        float64(pool.Alloc) / (1024 * 1024 * 1024),
-			Health:      pool.Health,
-		}
-		if kernel, exists := kernelStats[pool.Name]; exists && kernel.Health != "" {
-			stats.Health = kernel.Health
-		}
-		if io, exists := ioRates[pool.Name]; exists {
-			stats.ReadBytes = io.NRead
-			stats.WriteBytes = io.NWrite
-		}
-		slog.Debug("Storage pool sample", "backend", b.name, "pool", pool.Name, "health", stats.Health, "used_gb", stats.Used, "read_bps", stats.ReadBytes, "write_bps", stats.WriteBytes)
-		systemStats.ZfsPools[pool.Name] = stats
-	}
-
-}
-
-// poolStats returns the cached pool inventory, calling its collector at most
-// every poolStatsRefreshInterval. Only the first collection blocks; later
-// refreshes run in the background because utilities like `zpool list` can hang
-// for seconds on busy hosts, which would otherwise delay the hub's stats
-// response. On failure the previous inventory is retained and the refresh is
-// retried on the next cadence.
-func (b *poolBackend) poolStats() []zfs.PoolStat {
-	b.cacheMu.Lock()
-	defer b.cacheMu.Unlock()
-	if b.poolRefreshing || (!b.lastPoolStats.IsZero() && time.Since(b.lastPoolStats) < poolStatsRefreshInterval) {
-		return b.poolData
-	}
-	if b.lastPoolStats.IsZero() {
-		b.storePoolStats(b.poolStatsFn())
-		return b.poolData
-	}
-	b.poolRefreshing = true
-	go func() {
-		pools, err := b.poolStatsFn()
-		b.cacheMu.Lock()
-		defer b.cacheMu.Unlock()
-		b.poolRefreshing = false
-		b.storePoolStats(pools, err)
-	}()
-	return b.poolData
-}
-
-// storePoolStats records a pool inventory result. Callers must hold cacheMu.
-func (b *poolBackend) storePoolStats(pools []zfs.PoolStat, err error) {
-	if err != nil {
-		slog.Debug("Storage pool stats unavailable", "backend", b.name, "err", err)
-	} else {
-		b.poolData = pools
-	}
-	b.lastPoolStats = time.Now()
-}
-
-// kernelStats reads cumulative pool counters and converts them to per-second
-// rates. Counter decreases indicate a pool export/import and reset the
-// baseline instead of producing an underflow spike.
-func (b *poolBackend) kernelStats() (map[string]zfs.PoolKernelStat, map[string]zfs.PoolIoStats) {
-	if b.kernelStatsFn == nil {
-		return nil, nil
-	}
-	stats, err := b.kernelStatsFn()
-	if err != nil {
-		slog.Debug("Storage pool kernel stats unavailable", "backend", b.name, "err", err)
-		return nil, nil
-	}
-	now := time.Now()
-	byName := make(map[string]zfs.PoolKernelStat, len(stats))
-	rates := make(map[string]zfs.PoolIoStats, len(stats))
-	nextSamples := make(map[string]poolKernelSample, len(stats))
-	for _, stat := range stats {
-		byName[stat.Name] = stat
-		if previous, ok := b.kernelSamples[stat.Name]; ok && now.After(previous.at) &&
-			stat.NRead >= previous.nread && stat.NWrite >= previous.nwrite {
-			seconds := now.Sub(previous.at).Seconds()
-			rates[stat.Name] = zfs.PoolIoStats{
-				NRead:  uint64(float64(stat.NRead-previous.nread) / seconds),
-				NWrite: uint64(float64(stat.NWrite-previous.nwrite) / seconds),
-			}
-		}
-		nextSamples[stat.Name] = poolKernelSample{nread: stat.NRead, nwrite: stat.NWrite, at: now}
-	}
-	b.kernelSamples = nextSamples
-	return byName, rates
-}
-
-// refreshDatasetUsage re-runs `zfs list` when the refresh window has elapsed
-// and returns the mountpoint-keyed usage map. Like poolStats, only the first
-// collection blocks and later refreshes run in the background.
-func (b *poolBackend) refreshDatasetUsage() map[string]zfsDatasetUsage {
-	b.cacheMu.Lock()
-	defer b.cacheMu.Unlock()
-	if b.usageRefreshing || (!b.lastUsageRefresh.IsZero() && time.Since(b.lastUsageRefresh) < datasetUsageRefreshInterval) {
-		return b.datasetUsage
-	}
-	if b.lastUsageRefresh.IsZero() {
-		b.storeDatasetUsage(b.datasets())
-		return b.datasetUsage
-	}
-	b.usageRefreshing = true
-	go func() {
-		datasets, err := b.datasets()
-		b.cacheMu.Lock()
-		defer b.cacheMu.Unlock()
-		b.usageRefreshing = false
-		b.storeDatasetUsage(datasets, err)
-	}()
-	return b.datasetUsage
-}
-
-// storeDatasetUsage rebuilds the usage map from a dataset listing. The map is
-// replaced rather than mutated so returned references stay safe to read.
-// Callers must hold cacheMu.
-func (b *poolBackend) storeDatasetUsage(datasets []zfs.Dataset, err error) {
-	if err != nil {
-		slog.Debug("Storage pool dataset usage unavailable", "backend", b.name, "err", err)
-	} else {
-		usage := make(map[string]zfsDatasetUsage, len(datasets))
-		for _, ds := range datasets {
-			if ds.Mountpoint != "" && ds.Mountpoint != "-" {
-				usage[ds.Mountpoint] = zfsDatasetUsage{used: ds.Used, avail: ds.Avail}
-			}
-		}
-		b.datasetUsage = usage
-	}
-	b.lastUsageRefresh = time.Now()
-}
-
-// DatasetUsage returns ZFS dataset usage keyed by mountpoint, refreshed at
-// most every datasetUsageRefreshInterval. On failure the previous map is
-// retained and a debug log is emitted.
-func (m *StoragePoolManager) DatasetUsage() map[string]zfsDatasetUsage {
-	for _, backend := range m.backends {
-		if backend.name == "zfs" {
-			return backend.refreshDatasetUsage()
-		}
-	}
-	return nil
-}
-
-// GetDetail combines backend snapshots, identifying successful inventories so
-// the hub can accept partial updates without deleting failed backend records.
-func (m *StoragePoolManager) GetDetail(force bool) *zfsentity.ZfsData {
-	data := &zfsentity.ZfsData{Complete: true}
-	for _, backend := range m.backends {
-		snapshot := backend.getBackendDetail(force, m.detailInterval)
-		data.Pools = append(data.Pools, snapshot.Pools...)
-		if snapshot.Complete {
-			data.CompleteBackends = append(data.CompleteBackends, backend.name)
-		} else {
-			data.Complete = false
-		}
-	}
-	return data
-}
-
-func (b *poolBackend) getBackendDetail(force bool, interval time.Duration) *zfsentity.ZfsData {
-	b.detailMu.Lock()
-	defer b.detailMu.Unlock()
-
-	if force || b.detailFailed || b.detail == nil || time.Since(b.lastDetailRefresh) >= interval {
-		if data, err := b.collectDetail(b.detail); err != nil {
-			b.detailFailed = true
-			slog.Debug("Storage pool detail collection failed", "backend", b.name, "err", err)
-			if b.detail == nil {
-				return &zfsentity.ZfsData{}
-			}
-			return &zfsentity.ZfsData{Pools: b.detail.Pools}
-		} else {
-			b.detailFailed = false
-			b.detail = data
-			b.lastDetailRefresh = time.Now()
-		}
-	}
-	if b.detail == nil {
-		return &zfsentity.ZfsData{}
-	}
-	return b.detail
-}
-
-// collectDetail builds a ZfsData payload from the current system state.
-func (b *poolBackend) collectDetail(previous *zfsentity.ZfsData) (*zfsentity.ZfsData, error) {
-	pools, err := b.poolStatsFn()
-	if err != nil {
-		return nil, err
-	}
-	if len(pools) == 0 {
-		return &zfsentity.ZfsData{Pools: []*zfsentity.PoolDetail{}, Complete: true}, nil
-	}
-
-	statuses, statusErr := b.poolStatusesFn()
-	if statusErr != nil {
-		slog.Debug("Storage pool status unavailable", "backend", b.name, "err", statusErr)
-	}
-	datasets, datasetsErr := b.datasets()
-	if datasetsErr != nil {
-		slog.Debug("Storage pool datasets unavailable", "backend", b.name, "err", datasetsErr)
-	}
-
-	statusByPool := make(map[string]zfs.PoolStatus, len(statuses))
-	for _, st := range statuses {
-		statusByPool[st.Name] = st
-	}
-
-	previousByPool := make(map[string]*zfsentity.PoolDetail)
-	if previous != nil {
-		for _, pool := range previous.Pools {
-			if pool != nil {
-				previousByPool[pool.Name] = pool
-			}
-		}
-	}
-
-	data := &zfsentity.ZfsData{Pools: make([]*zfsentity.PoolDetail, 0, len(pools)), Complete: true}
-	for i := range pools {
-		p := &pools[i]
-		detail := &zfsentity.PoolDetail{
-			DisplayName: p.DisplayName,
-			Raw:         p.Raw,
-			Name:        p.Name,
-			Health:      p.Health,
-			Size:        p.Size,
-			Alloc:       p.Alloc,
-			Free:        p.Free,
-		}
-		if st, ok := statusByPool[p.Name]; statusErr == nil && ok {
-			if st.Scrub.State != "" && st.Scrub.State != "NONE" {
-				detail.Scrub = &zfsentity.Scrub{
-					State:    st.Scrub.State,
-					Progress: st.Scrub.Progress,
-					Errors:   st.Scrub.Errors,
-				}
-			}
-			for _, v := range st.Vdevs {
-				detail.Vdevs = append(detail.Vdevs, &zfsentity.Vdev{
-					Name:         v.Name,
-					State:        v.State,
-					ReadErrs:     v.ReadErrs,
-					WriteErrs:    v.WriteErrs,
-					ChecksumErrs: v.ChecksumErrs,
-				})
-			}
-		} else {
-			if cached := previousByPool[p.Name]; cached != nil {
-				detail.Scrub = cached.Scrub
-				detail.Vdevs = cached.Vdevs
-			}
-		}
-		if datasetsErr == nil {
-			foundDataset := false
-			for _, ds := range datasets {
-				if poolOfDataset(ds.Name) == p.Name {
-					foundDataset = true
-					detail.Datasets = append(detail.Datasets, &zfsentity.Dataset{
-						Name:       ds.Name,
-						Used:       ds.Used,
-						Avail:      ds.Avail,
-						Mountpoint: ds.Mountpoint,
-					})
-				}
-			}
-			if !foundDataset {
-				if cached := previousByPool[p.Name]; cached != nil {
-					detail.Datasets = cached.Datasets
-				}
-			}
-		} else if cached := previousByPool[p.Name]; cached != nil {
-			detail.Datasets = cached.Datasets
-		}
-		data.Pools = append(data.Pools, detail)
-	}
-	return data, nil
-}
-
-// poolOfDataset returns the pool name for a dataset name (everything before
-// the first '/'). Datasets without a separator belong to a pool of the same
-// name.
-func poolOfDataset(name string) string {
-	if idx := strings.IndexByte(name, '/'); idx >= 0 {
-		return name[:idx]
-	}
-	return name
-}
-
-// ZfsMountpoints returns the set of mountpoints backed by ZFS datasets.
-func (m *StoragePoolManager) ZfsMountpoints() map[string]bool {
-	usage := m.DatasetUsage()
-	mountpoints := make(map[string]bool, len(usage))
-	for mountpoint := range usage {
-		mountpoints[mountpoint] = true
-	}
-	return mountpoints
-}
-
-func btrfsPoolStats(fs btrfs.Filesystem) zfs.PoolStat {
-	return zfs.PoolStat{MountID: fs.MountID, IODevice: fs.IODevice, Raw: fs.Raw, DisplayName: fs.Name, Name: "b:" + fs.UUID, Size: fs.Size, Alloc: fs.Alloc, Free: fs.Size - min(fs.Alloc, fs.Size), Health: fs.Health}
-}
-
-func btrfsKernelStats(fs btrfs.Filesystem) zfs.PoolKernelStat {
-	return zfs.PoolKernelStat{Name: "b:" + fs.UUID, Health: fs.Health, NRead: fs.NRead, NWrite: fs.NWrite}
-}
-
-func btrfsPoolStatuses(fs btrfs.Filesystem) zfs.PoolStatus {
-	status := zfs.PoolStatus{Name: "b:" + fs.UUID, State: fs.Health, Scrub: zfs.ScrubStatus{State: "NONE"}}
-	for _, dev := range fs.Devices {
-		status.Vdevs = append(status.Vdevs, zfs.VdevStatus{
-			Name: dev.Name, State: dev.State,
-			ReadErrs: dev.ReadErrs, WriteErrs: dev.WriteErrs, ChecksumErrs: dev.CorruptionErrs,
-		})
-	}
-	return status
-}
-
-// markDuplicateCharts leaves pool telemetry and detail intact, but tells the
-// hub which charts already have a filesystem equivalent. Only exact kernel
-// filesystem and I/O-device matches qualify; labels are never used.
-func (m *StoragePoolManager) markDuplicateCharts(stats *system.Stats, filesystems map[string]*system.FsStats, mountID func(string) string) {
-	identities := make(map[string]string, len(filesystems))
-	for device, fs := range filesystems {
-		if fs.DiskTotal > 0 {
-			identities[device] = mountID(fs.Mountpoint)
-		}
-	}
-	for _, backend := range m.backends {
-		backend.cacheMu.Lock()
-		pools := backend.poolData
-		backend.cacheMu.Unlock()
-		for _, pool := range pools {
-			sample := stats.ZfsPools[pool.Name]
-			if sample == nil || pool.MountID == "" {
-				continue
-			}
-			for device, identity := range identities {
-				if identity != pool.MountID {
-					continue
-				}
-				// Raw physical usage is not equivalent to a filesystem usage chart.
-				sample.HideUsage = !pool.Raw
-				if pool.IODevice != "" && pool.IODevice == device {
-					sample.HideIO = true
-				}
-			}
-		}
-	}
-}

agent/storage_pool_test.go

@@ -1,548 +0,0 @@
-//go:build testing
-
-package agent
-
-import (
-	"errors"
-	"fmt"
-	"os/exec"
-	"strings"
-	"sync"
-	"testing"
-	"time"
-
-	"github.com/henrygd/beszel/agent/btrfs"
-	"github.com/henrygd/beszel/agent/zfs"
-	"github.com/henrygd/beszel/internal/entities/system"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestOptionalPoolSource(t *testing.T) {
-	failure := errors.New("timeout")
-	for _, err := range []error{nil, zfs.ErrNoZfs, fmt.Errorf("zpool: %w", exec.ErrNotFound), errors.ErrUnsupported, failure} {
-		_, got := optionalPoolSource(func() ([]zfs.PoolStat, error) { return nil, err })()
-		if err == failure {
-			assert.ErrorIs(t, got, failure)
-		} else {
-			assert.NoError(t, got)
-		}
-	}
-}
-
-type poolTestBackend struct {
-	name  string
-	err   error
-	alloc uint64
-	read  uint64
-	empty bool
-}
-
-func (state *poolTestBackend) backend() *poolBackend {
-	name := "zfs"
-	if strings.HasPrefix(state.name, "b:") {
-		name = "btrfs"
-	}
-	return &poolBackend{
-		name: name,
-		poolStatsFn: func() ([]zfs.PoolStat, error) {
-			if state.empty {
-				return nil, state.err
-			}
-			return []zfs.PoolStat{{Name: state.name, Size: 100, Alloc: state.alloc}}, state.err
-		},
-		kernelStatsFn: func() ([]zfs.PoolKernelStat, error) {
-			return []zfs.PoolKernelStat{{Name: state.name, NRead: state.read}}, state.err
-		},
-		poolStatusesFn: func() ([]zfs.PoolStatus, error) { return nil, nil },
-		datasetsFn:     func() ([]zfs.Dataset, error) { return nil, nil },
-	}
-}
-
-func TestIndependentPoolBackendCaches(t *testing.T) {
-	for _, failed := range []int{0, 1} {
-		t.Run([]string{"zfs", "btrfs"}[failed], func(t *testing.T) {
-			states := []*poolTestBackend{{name: "tank", alloc: 10}, {name: "b:uuid", alloc: 10}}
-			managers := []*poolBackend{states[0].backend(), states[1].backend()}
-			zm := &StoragePoolManager{backends: managers, detailInterval: time.Hour}
-			var stats system.Stats
-			zm.Update(&stats)
-			require.Len(t, stats.ZfsPools, 2)
-			require.True(t, zm.GetDetail(true).Complete)
-			baseline := poolKernelSample{at: time.Now().Add(-time.Second)}
-			for i, m := range managers {
-				m.lastPoolStats = time.Time{}
-				m.kernelSamples[states[i].name] = baseline
-				states[i].alloc = 20
-				states[i].read = 100
-			}
-			states[failed].err = errors.New("collection failed")
-			zm.Update(&stats)
-			healthy := 1 - failed
-			assert.Equal(t, uint64(10), managers[failed].poolData[0].Alloc)
-			assert.Equal(t, uint64(20), managers[healthy].poolData[0].Alloc)
-			assert.Equal(t, baseline, managers[failed].kernelSamples[states[failed].name])
-			assert.Zero(t, stats.ZfsPools[states[failed].name].ReadBytes)
-			assert.Positive(t, stats.ZfsPools[states[healthy].name].ReadBytes)
-			partial := zm.GetDetail(true)
-			assert.False(t, partial.Complete)
-			assert.False(t, partial.CanRefreshPool(states[failed].name))
-			assert.True(t, partial.CanRefreshPool(states[healthy].name))
-			assert.Equal(t, uint64(10), partial.Pools[failed].Alloc)
-			assert.Equal(t, uint64(20), partial.Pools[healthy].Alloc)
-			assert.False(t, zm.GetDetail(false).Complete, "a failed forced refresh must not become complete from cache")
-
-			// Successful empty inventory removes only the healthy backend's pool.
-			states[healthy].empty = true
-			managers[healthy].lastPoolStats = time.Time{}
-			zm.Update(&stats)
-			require.Len(t, stats.ZfsPools, 1)
-			assert.Contains(t, stats.ZfsPools, states[failed].name)
-			partial = zm.GetDetail(true)
-			require.Len(t, partial.Pools, 1)
-			assert.True(t, partial.CanRefreshPool(states[healthy].name))
-
-			// Recovery uses the retained I/O baseline, then normal removal works.
-			states[failed].err = nil
-			managers[failed].lastPoolStats = time.Time{}
-			zm.Update(&stats)
-			assert.Positive(t, stats.ZfsPools[states[failed].name].ReadBytes)
-			assert.True(t, zm.GetDetail(true).Complete)
-			states[failed].empty = true
-			managers[failed].lastPoolStats = time.Time{}
-			zm.Update(&stats)
-			assert.Empty(t, stats.ZfsPools)
-			assert.Empty(t, zm.GetDetail(true).Pools)
-		})
-	}
-}
-
-func TestIndependentBackendsWithoutCache(t *testing.T) {
-	z := &poolTestBackend{name: "tank", err: errors.New("ZFS failure")}
-	b := &poolTestBackend{name: "b:uuid", alloc: 20}
-	zm := &StoragePoolManager{backends: []*poolBackend{z.backend(), b.backend()}, detailInterval: time.Hour}
-	var stats system.Stats
-	zm.Update(&stats)
-	require.Len(t, stats.ZfsPools, 1)
-	assert.Contains(t, stats.ZfsPools, "b:uuid")
-	detail := zm.GetDetail(true)
-	require.Len(t, detail.Pools, 1)
-	assert.False(t, detail.Complete)
-	assert.Equal(t, []string{"btrfs"}, detail.CompleteBackends)
-}
-
-func TestConcurrentBackendDetailsAndMetrics(t *testing.T) {
-	zm := &StoragePoolManager{backends: []*poolBackend{(&poolTestBackend{name: "tank"}).backend(), (&poolTestBackend{name: "b:uuid"}).backend()}, detailInterval: time.Hour}
-	var wg sync.WaitGroup
-	for i := 0; i < 3; i++ {
-		wg.Add(1)
-		go func(metrics bool) {
-			defer wg.Done()
-			for j := 0; j < 10; j++ {
-				if metrics {
-					zm.Update(&system.Stats{})
-				} else {
-					zm.GetDetail(true)
-				}
-			}
-		}(i == 0)
-	}
-	wg.Wait()
-}
-
-func TestStoragePoolBackendOrder(t *testing.T) {
-	z := (&poolTestBackend{name: "tank"}).backend()
-	b := (&poolTestBackend{name: "b:uuid"}).backend()
-	b.datasetsFn = nil // Btrfs does not expose datasets.
-	z.datasetsFn = func() ([]zfs.Dataset, error) {
-		return []zfs.Dataset{{Name: "tank/data", Mountpoint: "/tank", Used: 10}}, nil
-	}
-	m := &StoragePoolManager{backends: []*poolBackend{b, z}, detailInterval: time.Hour}
-	var stats system.Stats
-	m.Update(&stats)
-	require.Len(t, stats.ZfsPools, 2)
-	detail := m.GetDetail(true)
-	require.True(t, detail.Complete)
-	assert.Equal(t, []string{"btrfs", "zfs"}, detail.CompleteBackends)
-	assert.Empty(t, detail.Pools[0].Datasets)
-	assert.Len(t, detail.Pools[1].Datasets, 1)
-	assert.Equal(t, uint64(10), m.DatasetUsage()["/tank"].used)
-
-	b.poolData[0].MountID = "uuid"
-	b.poolData[0].IODevice = "sda"
-	calls := 0
-	m.markDuplicateCharts(&stats, map[string]*system.FsStats{
-		"sda": {Mountpoint: "/", DiskTotal: 100},
-	}, func(string) string { calls++; return "uuid" })
-	assert.Equal(t, 1, calls, "resolve each filesystem once across all backends")
-	assert.True(t, stats.ZfsPools["b:uuid"].HideUsage)
-	assert.True(t, stats.ZfsPools["b:uuid"].HideIO)
-}
-
-func TestUpdatePopulatesZfsPools(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) {
-		return []zfs.PoolStat{{Name: "tank", Size: 23999000000000, Alloc: 12000000000000, Free: 11999000000000, Health: "DEGRADED"}}, nil
-	}
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) {
-		return []zfs.Dataset{
-			{Name: "tank/apps", Used: 5000000000000, Avail: 11999000000000, Mountpoint: "/tank/apps"},
-			{Name: "tank/backup", Used: 6000000000000, Avail: 11999000000000, Mountpoint: "/tank/backup"},
-			// Small zvol (Proxmox VM EFI disk): must not round to zero.
-			{Name: "rpool/vm-100-disk-2", Used: 4194304, Avail: 0, Mountpoint: "-"},
-		}, nil
-	}
-	var kernelCalls int
-	zm.backends[0].kernelStatsFn = func() ([]zfs.PoolKernelStat, error) {
-		kernelCalls++
-		return []zfs.PoolKernelStat{{
-			Name: "tank", Health: "ONLINE",
-			NRead: uint64(kernelCalls-1) * 1250, NWrite: uint64(kernelCalls-1) * 5120,
-		}}, nil
-	}
-
-	var stats system.Stats
-	// The first kernel sample establishes the cumulative-counter baseline.
-	zm.Update(&stats)
-	zm.backends[0].kernelSamples["tank"] = poolKernelSample{at: time.Now().Add(-time.Second)}
-	zm.Update(&stats)
-	require.NotNil(t, stats.ZfsPools)
-	require.Contains(t, stats.ZfsPools, "tank")
-	assert.InDelta(t, 22350.8105, stats.ZfsPools["tank"].Total, 0.0001) // Size in GiB
-	assert.InDelta(t, 11175.8709, stats.ZfsPools["tank"].Used, 0.0001)  // Alloc in GiB
-	assert.Equal(t, "ONLINE", stats.ZfsPools["tank"].Health)
-	assert.InDelta(t, 1250, stats.ZfsPools["tank"].ReadBytes, 5)
-	assert.InDelta(t, 5120, stats.ZfsPools["tank"].WriteBytes, 5)
-
-}
-
-// TestUpdateKernelStatsMissing verifies pools without a kernel sample report zero
-// I/O instead of erroring.
-func TestUpdateKernelStatsMissing(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) {
-		return []zfs.PoolStat{{Name: "tank", Size: 1, Alloc: 1, Health: "ONLINE"}}, nil
-	}
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
-	zm.backends[0].kernelStatsFn = func() ([]zfs.PoolKernelStat, error) {
-		return nil, zfs.ErrNoZfs
-	}
-
-	var stats system.Stats
-	zm.Update(&stats)
-	require.NotNil(t, stats.ZfsPools)
-	assert.Equal(t, uint64(0), stats.ZfsPools["tank"].ReadBytes)
-	assert.Equal(t, uint64(0), stats.ZfsPools["tank"].WriteBytes)
-}
-
-func TestUpdateKernelCounterReset(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) {
-		return []zfs.PoolStat{{Name: "tank", Health: "ONLINE"}}, nil
-	}
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
-	zm.backends[0].kernelSamples = map[string]poolKernelSample{
-		"tank": {nread: 100, nwrite: 200, at: time.Now().Add(-time.Second)},
-	}
-	zm.backends[0].kernelStatsFn = func() ([]zfs.PoolKernelStat, error) {
-		return []zfs.PoolKernelStat{{Name: "tank", Health: "ONLINE", NRead: 10, NWrite: 20}}, nil
-	}
-
-	var stats system.Stats
-	zm.Update(&stats)
-	assert.Equal(t, uint64(0), stats.ZfsPools["tank"].ReadBytes)
-	assert.Equal(t, uint64(0), stats.ZfsPools["tank"].WriteBytes)
-}
-
-func TestUpdateNoZfs(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	calls := 0
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) {
-		calls++
-		return nil, zfs.ErrNoZfs
-	}
-
-	var stats system.Stats
-	zm.Update(&stats)
-	zm.Update(&stats)
-	assert.Nil(t, stats.ZfsPools)
-	assert.Equal(t, 1, calls, "failed pool discovery should be cached until the next refresh interval")
-}
-
-func TestUpdateEmptyPools(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	calls := 0
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) {
-		calls++
-		return nil, nil
-	}
-
-	var stats system.Stats
-	zm.Update(&stats)
-	zm.Update(&stats)
-	assert.Nil(t, stats.ZfsPools)
-	assert.Equal(t, 1, calls, "an empty pool inventory should be cached until the next refresh interval")
-}
-
-func TestDatasetUsage(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	calls := 0
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) {
-		calls++
-		return []zfs.Dataset{
-			{Name: "tank", Used: 12000000000000, Avail: 11999000000000, Mountpoint: "/tank"},
-			{Name: "tank/apps", Used: 1000000000000, Avail: 11999000000000, Mountpoint: "/tank/apps"},
-			{Name: "rpool", Used: 900000000000, Avail: 300000000000, Mountpoint: "-"}, // zvol/unmounted: excluded
-		}, nil
-	}
-
-	usage := zm.DatasetUsage()
-	require.Len(t, usage, 2)
-	assert.Equal(t, zfsDatasetUsage{used: 12000000000000, avail: 11999000000000}, usage["/tank"])
-	assert.Equal(t, zfsDatasetUsage{used: 1000000000000, avail: 11999000000000}, usage["/tank/apps"])
-	assert.Equal(t, 1, calls)
-
-	// Second call within the refresh window must not re-run the collector.
-	zm.DatasetUsage()
-	assert.Equal(t, 1, calls)
-}
-
-func TestDatasetUsageRefreshOnErrorKeepsPrevious(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) {
-		return []zfs.Dataset{{Name: "tank", Used: 1, Avail: 1, Mountpoint: "/tank"}}, nil
-	}
-	assert.Len(t, zm.DatasetUsage(), 1)
-
-	// Force refresh window expiry, then a failing collector.
-	zm.backends[0].lastUsageRefresh = time.Now().Add(-10 * time.Minute)
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) {
-		return nil, zfs.ErrNoZfs
-	}
-	usage := zm.DatasetUsage()
-	assert.Len(t, usage, 1, "previous usage should be retained on error")
-}
-
-func TestDatasetUsageClearsAbsentBackend(t *testing.T) {
-	b := newZfsBackend()
-	b.datasetUsage = map[string]zfsDatasetUsage{"/tank": {used: 1, avail: 1}}
-	b.datasetsFn = optionalPoolSource(func() ([]zfs.Dataset, error) {
-		return nil, zfs.ErrNoZfs
-	})
-
-	datasets, err := b.datasets()
-	require.NoError(t, err, "an absent backend must not produce an error to log")
-	assert.Empty(t, datasets)
-	b.refreshDatasetUsage()
-	assert.Empty(t, b.datasetUsage)
-	assert.False(t, b.lastUsageRefresh.IsZero())
-}
-
-func TestGetDetailForceRefresh(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	poolCalls := 0
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) {
-		poolCalls++
-		return []zfs.PoolStat{{Name: "tank", Alloc: uint64(poolCalls)}}, nil
-	}
-	zm.backends[0].poolStatusesFn = func() ([]zfs.PoolStatus, error) { return nil, nil }
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
-
-	first := zm.GetDetail(false)
-	assert.True(t, first.Complete)
-	require.Len(t, first.Pools, 1)
-	assert.Equal(t, uint64(1), first.Pools[0].Alloc)
-
-	cached := zm.GetDetail(false)
-	require.Len(t, cached.Pools, 1)
-	assert.Equal(t, uint64(1), cached.Pools[0].Alloc)
-	assert.Equal(t, 1, poolCalls)
-
-	refreshed := zm.GetDetail(true)
-	assert.True(t, refreshed.Complete)
-	require.Len(t, refreshed.Pools, 1)
-	assert.Equal(t, uint64(2), refreshed.Pools[0].Alloc)
-	assert.Equal(t, 2, poolCalls)
-}
-
-func TestGetDetailSuccessfulEmptyInventoryClearsCache(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) {
-		return []zfs.PoolStat{{Name: "tank"}}, nil
-	}
-	zm.backends[0].poolStatusesFn = func() ([]zfs.PoolStatus, error) { return nil, nil }
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
-
-	require.Len(t, zm.GetDetail(false).Pools, 1)
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) { return nil, nil }
-	empty := zm.GetDetail(true)
-	assert.True(t, empty.Complete)
-	assert.Empty(t, empty.Pools)
-}
-
-func TestGetDetailFailureReturnsIncompleteCachedInventory(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) {
-		return []zfs.PoolStat{{Name: "tank"}}, nil
-	}
-	zm.backends[0].poolStatusesFn = func() ([]zfs.PoolStatus, error) {
-		return []zfs.PoolStatus{{Name: "tank", Vdevs: []zfs.VdevStatus{{Name: "mirror-0"}}}}, nil
-	}
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) {
-		return []zfs.Dataset{{Name: "tank/data"}}, nil
-	}
-	first := zm.GetDetail(false)
-	require.True(t, first.Complete)
-	require.Len(t, first.Pools[0].Vdevs, 1)
-	require.Len(t, first.Pools[0].Datasets, 1)
-
-	zm.backends[0].poolStatusesFn = func() ([]zfs.PoolStatus, error) { return nil, zfs.ErrNoZfs }
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) { return nil, zfs.ErrNoZfs }
-	partial := zm.GetDetail(true)
-	require.True(t, partial.Complete)
-	require.Len(t, partial.Pools[0].Vdevs, 1)
-	require.Len(t, partial.Pools[0].Datasets, 1)
-
-	zm.backends[0].poolStatsFn = func() ([]zfs.PoolStat, error) { return nil, zfs.ErrNoZfs }
-	lastSuccessfulRefresh := zm.backends[0].lastDetailRefresh
-	failed := zm.GetDetail(true)
-	assert.False(t, failed.Complete)
-	require.Len(t, failed.Pools, 1)
-	assert.Equal(t, "tank", failed.Pools[0].Name)
-	assert.Equal(t, lastSuccessfulRefresh, zm.backends[0].lastDetailRefresh)
-}
-
-func TestZfsMountpoints(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs"}}}
-	zm.backends[0].datasetsFn = func() ([]zfs.Dataset, error) {
-		return []zfs.Dataset{
-			{Name: "tank", Mountpoint: "/tank"},
-			{Name: "rpool/ROOT/pve-1", Mountpoint: "/"},
-		}, nil
-	}
-	mountpoints := zm.ZfsMountpoints()
-	assert.Len(t, mountpoints, 2)
-	assert.True(t, mountpoints["/tank"])
-	assert.True(t, mountpoints["/"])
-}
-
-func TestBtrfsRawCapacityPropagates(t *testing.T) {
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs", poolStatsFn: func() ([]zfs.PoolStat, error) {
-		return []zfs.PoolStat{btrfsPoolStats(btrfs.Filesystem{UUID: "raw", Name: "raw", Size: 200, Alloc: 100, Raw: true})}, nil
-	},
-		poolStatusesFn: func() ([]zfs.PoolStatus, error) { return nil, nil },
-		datasetsFn:     func() ([]zfs.Dataset, error) { return nil, nil }}}}
-	var stats system.Stats
-	zm.Update(&stats)
-	require.True(t, stats.ZfsPools["b:raw"].Raw)
-	detail := zm.GetDetail(true)
-	require.True(t, detail.Complete)
-	require.Len(t, detail.Pools, 1)
-	assert.True(t, detail.Pools[0].Raw)
-}
-
-func TestMarkDuplicatePoolCharts(t *testing.T) {
-	for _, tc := range []struct {
-		name, poolID, device string
-		raw                  bool
-		diskTotal            float64
-		wantUsage, wantIO    bool
-	}{
-		{"single device root", "fs1", "dm-0", false, 100, true, true},
-		{"multi device", "fs1", "", false, 100, true, false},
-		{"different IO device", "fs1", "nvme0n1", false, 100, true, false},
-		{"different filesystem", "fs2", "dm-0", false, 100, false, false},
-		{"unknown identity", "", "dm-0", false, 100, false, false},
-		{"raw usage", "fs1", "dm-0", true, 100, false, true},
-		{"failed disk collection", "fs1", "dm-0", false, 0, false, false},
-	} {
-		t.Run(tc.name, func(t *testing.T) {
-			zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs", poolData: []zfs.PoolStat{{Name: "arbitrary label", MountID: tc.poolID, IODevice: tc.device, Raw: tc.raw}}}}}
-			stats := &system.Stats{ZfsPools: map[string]*system.ZfsPool{"arbitrary label": {}}}
-			fs := map[string]*system.FsStats{"dm-0": {Root: true, Mountpoint: "/", DiskTotal: tc.diskTotal}}
-			zm.markDuplicateCharts(stats, fs, func(string) string { return "fs1" })
-			assert.Equal(t, tc.wantUsage, stats.ZfsPools["arbitrary label"].HideUsage)
-			assert.Equal(t, tc.wantIO, stats.ZfsPools["arbitrary label"].HideIO)
-			// Bind mounts and custom extra-filesystem names have the same identity.
-			fs["dm-0"].Root = false
-			fs["dm-0"].Mountpoint = "/extra-filesystems/storage"
-			fs["dm-0"].Name = "custom name"
-			stats.ZfsPools["arbitrary label"] = &system.ZfsPool{}
-			zm.markDuplicateCharts(stats, fs, func(string) string { return "fs1" })
-			assert.Equal(t, tc.wantUsage, stats.ZfsPools["arbitrary label"].HideUsage)
-			assert.Equal(t, tc.wantIO, stats.ZfsPools["arbitrary label"].HideIO)
-		})
-	}
-}
-
-func TestBtrfsPoolIdentities(t *testing.T) {
-	old := btrfsFilesystems
-	t.Cleanup(func() { btrfsFilesystems = old })
-	label := "tank"
-	btrfsFilesystems = func() ([]btrfs.Filesystem, error) {
-		return []btrfs.Filesystem{
-			{UUID: "11111111-1111-4111-8111-111111111111", Name: label, Size: 100, Health: "ONLINE", NRead: 100, Devices: []btrfs.Device{{Name: "first"}}},
-			{UUID: "22222222-2222-4222-8222-222222222222", Name: "tank", Size: 200, Health: "DEGRADED", NRead: 200, Devices: []btrfs.Device{{Name: "second"}}},
-		}, nil
-	}
-	zm := &StoragePoolManager{detailInterval: time.Hour, backends: []*poolBackend{{name: "zfs", poolStatsFn: func() ([]zfs.PoolStat, error) { return []zfs.PoolStat{{Name: "tank", Size: 300}}, nil },
-		kernelStatsFn: func() ([]zfs.PoolKernelStat, error) { return []zfs.PoolKernelStat{{Name: "tank", NRead: 300}}, nil },
-		poolStatusesFn: func() ([]zfs.PoolStatus, error) {
-			return []zfs.PoolStatus{{Name: "tank", Vdevs: []zfs.VdevStatus{{Name: "zfs-device"}}}}, nil
-		},
-
-		datasetsFn: func() ([]zfs.Dataset, error) { return []zfs.Dataset{{Name: "tank/data"}}, nil }}, newBtrfsBackend()}}
-	first := "b:11111111-1111-4111-8111-111111111111"
-	second := "b:22222222-2222-4222-8222-222222222222"
-	var stats system.Stats
-	zm.Update(&stats)
-	require.Len(t, stats.ZfsPools, 3)
-	assert.Contains(t, stats.ZfsPools, "tank")
-	assert.Equal(t, "ONLINE", stats.ZfsPools[first].Health)
-	assert.Equal(t, "DEGRADED", stats.ZfsPools[second].Health)
-	assert.Equal(t, uint64(100), zm.backends[1].kernelSamples[first].nread)
-	assert.Equal(t, uint64(200), zm.backends[1].kernelSamples[second].nread)
-	detail := zm.GetDetail(true)
-	require.Len(t, detail.Pools, 3)
-	assert.Equal(t, "zfs-device", detail.Pools[0].Vdevs[0].Name)
-	assert.Len(t, detail.Pools[0].Datasets, 1)
-	assert.Equal(t, "first", detail.Pools[1].Vdevs[0].Name)
-	assert.Empty(t, detail.Pools[1].Datasets)
-	assert.Equal(t, "second", detail.Pools[2].Vdevs[0].Name)
-	label = "renamed"
-	zm.backends[0].lastPoolStats = time.Time{}
-	zm.backends[1].lastPoolStats = time.Time{}
-	zm.Update(&stats)
-	require.Len(t, stats.ZfsPools, 3)
-	assert.Equal(t, "renamed", stats.ZfsPools[first].DisplayName)
-	assert.Equal(t, first, zm.GetDetail(true).Pools[1].Name)
-	assert.Equal(t, "renamed", zm.GetDetail(true).Pools[1].DisplayName)
-}
-
-func TestStaleUtilityCachesRefreshInBackground(t *testing.T) {
-	release := make(chan struct{})
-	b := &poolBackend{name: "zfs"}
-	b.poolStatsFn = func() ([]zfs.PoolStat, error) {
-		<-release
-		return []zfs.PoolStat{{Name: "new"}}, nil
-	}
-	b.datasetsFn = func() ([]zfs.Dataset, error) {
-		<-release
-		return []zfs.Dataset{{Name: "new", Mountpoint: "/new"}}, nil
-	}
-	b.poolData = []zfs.PoolStat{{Name: "old"}}
-	b.lastPoolStats = time.Now().Add(-2 * poolStatsRefreshInterval)
-	b.datasetUsage = map[string]zfsDatasetUsage{"/old": {}}
-	b.lastUsageRefresh = time.Now().Add(-2 * datasetUsageRefreshInterval)
-
-	// A hung utility must not block collection; cached data is served meanwhile.
-	for range 2 {
-		assert.Equal(t, "old", b.poolStats()[0].Name)
-		assert.Contains(t, b.refreshDatasetUsage(), "/old")
-	}
-
-	close(release)
-	require.Eventually(t, func() bool {
-		return b.poolStats()[0].Name == "new" && b.refreshDatasetUsage()["/new"] == zfsDatasetUsage{}
-	}, time.Second, time.Millisecond)
-}

agent/system.go

@@ -12,7 +12,6 @@ import (
 
 	"github.com/henrygd/beszel"
 	"github.com/henrygd/beszel/agent/battery"
-	"github.com/henrygd/beszel/agent/btrfs"
 	"github.com/henrygd/beszel/agent/utils"
 	"github.com/henrygd/beszel/agent/zfs"
 	"github.com/henrygd/beszel/internal/entities/container"
@@ -220,9 +219,8 @@ func (a *Agent) getSystemStats(cacheTimeMs uint16) system.Stats {
 	// disk i/o (cache-aware per interval)
 	a.updateDiskIo(cacheTimeMs, &systemStats)
 
-	// storage pool stats
-	a.storagePoolManager.Update(&systemStats)
-	a.storagePoolManager.markDuplicateCharts(&systemStats, a.fsStats, btrfs.MountID)
+	// zfs pool stats
+	a.zfsManager.Update(&systemStats)
 
 	// network stats (per cache interval)
 	a.updateNetworkStats(cacheTimeMs, &systemStats)

agent/test-data/package_updates/apk_alpine320_list_upgradable.txt

@@ -1,10 +0,0 @@
-apk-tools-2.14.4-r1 aarch64 {apk-tools} (GPL-2.0-only) [upgradable from: apk-tools-2.14.4-r0]
-busybox-1.36.1-r31 aarch64 {busybox} (GPL-2.0-only) [upgradable from: busybox-1.36.1-r28]
-busybox-binsh-1.36.1-r31 aarch64 {busybox} (GPL-2.0-only) [upgradable from: busybox-binsh-1.36.1-r28]
-ca-certificates-bundle-20260413-r0 aarch64 {ca-certificates} (MPL-2.0 AND MIT) [upgradable from: ca-certificates-bundle-20240226-r0]
-libcrypto3-3.3.7-r0 aarch64 {openssl} (Apache-2.0) [upgradable from: libcrypto3-3.3.0-r2]
-libssl3-3.3.7-r0 aarch64 {openssl} (Apache-2.0) [upgradable from: libssl3-3.3.0-r2]
-musl-1.2.5-r3 aarch64 {musl} (MIT) [upgradable from: musl-1.2.5-r0]
-musl-utils-1.2.5-r3 aarch64 {musl} (MIT AND BSD-2-Clause AND GPL-2.0-or-later) [upgradable from: musl-utils-1.2.5-r0]
-ssl_client-1.36.1-r31 aarch64 {busybox} (GPL-2.0-only) [upgradable from: ssl_client-1.36.1-r28]
-zlib-1.3.2-r0 aarch64 {zlib} (Zlib) [upgradable from: zlib-1.3.1-r1]

agent/test-data/package_updates/apt_debian12.txt

@@ -1,101 +0,0 @@
-Reading package lists...
-Building dependency tree...
-Reading state information...
-Calculating upgrade...
-The following packages will be upgraded:
-  base-files bash bsdutils debian-archive-keyring dpkg e2fsprogs gcc-12-base
-  gpgv init-system-helpers libblkid1 libc-bin libc6 libcap2 libcom-err2
-  libext2fs2 libgcc-s1 libgcrypt20 libgnutls30 liblzma5 libmount1
-  libpam-modules libpam-modules-bin libpam-runtime libpam0g libpcre2-8-0
-  libseccomp2 libsmartcols1 libss2 libstdc++6 libsystemd0 libtasn1-6 libudev1
-  libuuid1 login logsave mount passwd perl-base sed tar tzdata usr-is-merged
-  util-linux util-linux-extra
-44 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.
-Inst base-files [12.4+deb12u4] (12.4+deb12u15 Debian:12.15/oldstable [arm64])
-Conf base-files (12.4+deb12u15 Debian:12.15/oldstable [arm64])
-Inst bash [5.2.15-2+b2] (5.2.15-2+b13 Debian:12.15/oldstable [arm64])
-Conf bash (5.2.15-2+b13 Debian:12.15/oldstable [arm64])
-Inst bsdutils [1:2.38.1-5+b1] (1:2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Conf bsdutils (1:2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Inst tar [1.34+dfsg-1.2] (1.34+dfsg-1.2+deb12u1 Debian:12.15/oldstable [arm64])
-Conf tar (1.34+dfsg-1.2+deb12u1 Debian:12.15/oldstable [arm64])
-Inst dpkg [1.21.22] (1.21.23 Debian:12.15/oldstable [arm64])
-Conf dpkg (1.21.23 Debian:12.15/oldstable [arm64])
-Inst login [1:4.13+dfsg1-1+b1] (1:4.13+dfsg1-1+deb12u2 Debian:12.15/oldstable [arm64])
-Conf login (1:4.13+dfsg1-1+deb12u2 Debian:12.15/oldstable [arm64])
-Inst perl-base [5.36.0-7+deb12u1] (5.36.0-7+deb12u3 Debian:12.15/oldstable [arm64])
-Conf perl-base (5.36.0-7+deb12u3 Debian:12.15/oldstable [arm64])
-Inst sed [4.9-1] (4.9-1+deb12u1 Debian:12.15/oldstable [arm64])
-Conf sed (4.9-1+deb12u1 Debian:12.15/oldstable [arm64])
-Inst gcc-12-base [12.2.0-14] (12.2.0-14+deb12u1 Debian:12.15/oldstable [arm64]) [libstdc++6:arm64 libgcc-s1:arm64 ]
-Conf gcc-12-base (12.2.0-14+deb12u1 Debian:12.15/oldstable [arm64]) [libstdc++6:arm64 libgcc-s1:arm64 ]
-Inst libgcc-s1 [12.2.0-14] (12.2.0-14+deb12u1 Debian:12.15/oldstable [arm64]) [libstdc++6:arm64 ]
-Conf libgcc-s1 (12.2.0-14+deb12u1 Debian:12.15/oldstable [arm64]) [libstdc++6:arm64 ]
-Inst libstdc++6 [12.2.0-14] (12.2.0-14+deb12u1 Debian:12.15/oldstable [arm64])
-Conf libstdc++6 (12.2.0-14+deb12u1 Debian:12.15/oldstable [arm64])
-Inst libc6 [2.36-9+deb12u3] (2.36-9+deb12u14 Debian:12.15/oldstable [arm64])
-Conf libc6 (2.36-9+deb12u14 Debian:12.15/oldstable [arm64])
-Inst libsmartcols1 [2.38.1-5+b1] (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Conf libsmartcols1 (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Inst util-linux-extra [2.38.1-5+b1] (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Conf util-linux-extra (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Inst util-linux [2.38.1-5+b1] (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Conf util-linux (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Inst usr-is-merged [35] (37~deb12u1 Debian:12.15/oldstable [all])
-Conf usr-is-merged (37~deb12u1 Debian:12.15/oldstable [all])
-Inst init-system-helpers [1.65.2] (1.65.2+deb12u1 Debian:12.15/oldstable [all])
-Conf init-system-helpers (1.65.2+deb12u1 Debian:12.15/oldstable [all])
-Inst libc-bin [2.36-9+deb12u3] (2.36-9+deb12u14 Debian:12.15/oldstable [arm64])
-Conf libc-bin (2.36-9+deb12u14 Debian:12.15/oldstable [arm64])
-Inst libpam0g [1.5.2-6+deb12u1] (1.5.2-6+deb12u2 Debian:12.15/oldstable [arm64])
-Conf libpam0g (1.5.2-6+deb12u2 Debian:12.15/oldstable [arm64])
-Inst libpam-modules-bin [1.5.2-6+deb12u1] (1.5.2-6+deb12u2 Debian:12.15/oldstable [arm64]) [libpam-modules:arm64 on libpam-modules-bin:arm64] [libpam-modules:arm64 ]
-Conf libpam-modules-bin (1.5.2-6+deb12u2 Debian:12.15/oldstable [arm64]) [libpam-modules:arm64 ]
-Inst libpam-modules [1.5.2-6+deb12u1] (1.5.2-6+deb12u2 Debian:12.15/oldstable [arm64])
-Conf libpam-modules (1.5.2-6+deb12u2 Debian:12.15/oldstable [arm64])
-Inst logsave [1.47.0-2] (1.47.0-2+b2 Debian:12.15/oldstable [arm64])
-Inst libext2fs2 [1.47.0-2] (1.47.0-2+b2 Debian:12.15/oldstable [arm64]) [e2fsprogs:arm64 on libext2fs2:arm64] [e2fsprogs:arm64 ]
-Conf libext2fs2 (1.47.0-2+b2 Debian:12.15/oldstable [arm64]) [e2fsprogs:arm64 ]
-Inst e2fsprogs [1.47.0-2] (1.47.0-2+b2 Debian:12.15/oldstable [arm64])
-Inst mount [2.38.1-5+b1] (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Inst libpam-runtime [1.5.2-6+deb12u1] (1.5.2-6+deb12u2 Debian:12.15/oldstable [all])
-Conf libpam-runtime (1.5.2-6+deb12u2 Debian:12.15/oldstable [all])
-Inst passwd [1:4.13+dfsg1-1+b1] (1:4.13+dfsg1-1+deb12u2 Debian:12.15/oldstable [arm64])
-Conf passwd (1:4.13+dfsg1-1+deb12u2 Debian:12.15/oldstable [arm64])
-Inst debian-archive-keyring [2023.3+deb12u1] (2023.3+deb12u2 Debian:12.15/oldstable [all])
-Conf debian-archive-keyring (2023.3+deb12u2 Debian:12.15/oldstable [all])
-Inst libgcrypt20 [1.10.1-3] (1.10.1-3+deb12u1 Debian:12.15/oldstable, Debian-Security:12/oldstable-security [arm64])
-Conf libgcrypt20 (1.10.1-3+deb12u1 Debian:12.15/oldstable, Debian-Security:12/oldstable-security [arm64])
-Inst gpgv [2.2.40-1.1] (2.2.40-1.1+deb12u2 Debian:12.15/oldstable [arm64])
-Conf gpgv (2.2.40-1.1+deb12u2 Debian:12.15/oldstable [arm64])
-Inst libblkid1 [2.38.1-5+b1] (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Conf libblkid1 (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Inst libcap2 [1:2.66-4] (1:2.66-4+deb12u3+b1 Debian:12.15/oldstable [arm64])
-Conf libcap2 (1:2.66-4+deb12u3+b1 Debian:12.15/oldstable [arm64])
-Inst libtasn1-6 [4.19.0-2] (4.19.0-2+deb12u1 Debian:12.15/oldstable, Debian-Security:12/oldstable-security [arm64])
-Conf libtasn1-6 (4.19.0-2+deb12u1 Debian:12.15/oldstable, Debian-Security:12/oldstable-security [arm64])
-Inst libgnutls30 [3.7.9-2+deb12u1] (3.7.9-2+deb12u7 Debian:12.15/oldstable, Debian-Security:12/oldstable-security [arm64])
-Conf libgnutls30 (3.7.9-2+deb12u7 Debian:12.15/oldstable, Debian-Security:12/oldstable-security [arm64])
-Inst liblzma5 [5.4.1-0.2] (5.4.1-1+deb12u2 Debian-Security:12/oldstable-security [arm64])
-Conf liblzma5 (5.4.1-1+deb12u2 Debian-Security:12/oldstable-security [arm64])
-Inst libmount1 [2.38.1-5+b1] (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Conf libmount1 (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Inst libpcre2-8-0 [10.42-1] (10.42-1+deb12u1 Debian-Security:12/oldstable-security [arm64])
-Conf libpcre2-8-0 (10.42-1+deb12u1 Debian-Security:12/oldstable-security [arm64])
-Inst libseccomp2 [2.5.4-1+b3] (2.5.4-1+deb12u1 Debian:12.15/oldstable [arm64])
-Conf libseccomp2 (2.5.4-1+deb12u1 Debian:12.15/oldstable [arm64])
-Inst libsystemd0 [252.19-1~deb12u1] (252.39-1~deb12u2 Debian:12.15/oldstable [arm64])
-Conf libsystemd0 (252.39-1~deb12u2 Debian:12.15/oldstable [arm64])
-Inst libudev1 [252.19-1~deb12u1] (252.39-1~deb12u2 Debian:12.15/oldstable [arm64])
-Conf libudev1 (252.39-1~deb12u2 Debian:12.15/oldstable [arm64])
-Inst libuuid1 [2.38.1-5+b1] (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Conf libuuid1 (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Inst tzdata [2023c-5+deb12u1] (2026b-0+deb12u1 Debian:12.15/oldstable [all])
-Inst libcom-err2 [1.47.0-2] (1.47.0-2+b2 Debian:12.15/oldstable [arm64])
-Inst libss2 [1.47.0-2] (1.47.0-2+b2 Debian:12.15/oldstable [arm64])
-Conf logsave (1.47.0-2+b2 Debian:12.15/oldstable [arm64])
-Conf e2fsprogs (1.47.0-2+b2 Debian:12.15/oldstable [arm64])
-Conf mount (2.38.1-5+deb12u3 Debian:12.15/oldstable [arm64])
-Conf tzdata (2026b-0+deb12u1 Debian:12.15/oldstable [all])
-Conf libcom-err2 (1.47.0-2+b2 Debian:12.15/oldstable [arm64])
-Conf libss2 (1.47.0-2+b2 Debian:12.15/oldstable [arm64])

agent/test-data/package_updates/apt_ubuntu2204.txt

@@ -1,130 +0,0 @@
-Reading package lists...
-Building dependency tree...
-Reading state information...
-Calculating upgrade...
-The following packages will be upgraded:
-  apt base-files bash bsdutils coreutils diffutils dpkg e2fsprogs gcc-12-base
-  gpgv gzip libapt-pkg6.0 libattr1 libblkid1 libbz2-1.0 libc-bin libc6 libcap2
-  libcom-err2 libext2fs2 libgcc-s1 libgcrypt20 libgnutls30 libgssapi-krb5-2
-  libk5crypto3 libkrb5-3 libkrb5support0 liblzma5 libmount1 libncurses6
-  libncursesw6 libp11-kit0 libpam-modules libpam-modules-bin libpam-runtime
-  libpam0g libprocps8 libseccomp2 libsmartcols1 libss2 libssl3 libstdc++6
-  libsystemd0 libtasn1-6 libtinfo6 libudev1 libuuid1 login logsave mount
-  ncurses-base ncurses-bin passwd perl-base procps sed tar util-linux
-58 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.
-Inst gcc-12-base [12.3.0-1ubuntu1~22.04] (12.3.0-1ubuntu1~22.04.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64]) [libstdc++6:arm64 libgcc-s1:arm64 ]
-Conf gcc-12-base (12.3.0-1ubuntu1~22.04.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64]) [libstdc++6:arm64 libgcc-s1:arm64 ]
-Inst libgcc-s1 [12.3.0-1ubuntu1~22.04] (12.3.0-1ubuntu1~22.04.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64]) [libstdc++6:arm64 ]
-Conf libgcc-s1 (12.3.0-1ubuntu1~22.04.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64]) [libstdc++6:arm64 ]
-Inst libstdc++6 [12.3.0-1ubuntu1~22.04] (12.3.0-1ubuntu1~22.04.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libstdc++6 (12.3.0-1ubuntu1~22.04.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libc6 [2.35-0ubuntu3.4] (2.35-0ubuntu3.15 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libc6 (2.35-0ubuntu3.15 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst base-files [12ubuntu4.4] (12ubuntu4.7 Ubuntu:22.04/jammy-updates [arm64])
-Conf base-files (12ubuntu4.7 Ubuntu:22.04/jammy-updates [arm64])
-Inst bash [5.1-6ubuntu1] (5.1-6ubuntu1.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf bash (5.1-6ubuntu1.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst bsdutils [1:2.37.2-4ubuntu3] (1:2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf bsdutils (1:2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst coreutils [8.32-4.1ubuntu1] (8.32-4.1ubuntu1.4 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf coreutils (8.32-4.1ubuntu1.4 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst diffutils [1:3.8-0ubuntu2] (1:3.8-0ubuntu2.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf diffutils (1:3.8-0ubuntu2.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libbz2-1.0 [1.0.8-5build1] (1.0.8-5ubuntu0.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libbz2-1.0 (1.0.8-5ubuntu0.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libgcrypt20 [1.9.4-3ubuntu3] (1.9.4-3ubuntu3.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libgcrypt20 (1.9.4-3ubuntu3.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst liblzma5 [5.2.5-2ubuntu1] (5.2.5-2ubuntu1.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf liblzma5 (5.2.5-2ubuntu1.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libsystemd0 [249.11-0ubuntu3.10] (249.11-0ubuntu3.22 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libsystemd0 (249.11-0ubuntu3.22 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libudev1 [249.11-0ubuntu3.10] (249.11-0ubuntu3.22 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libudev1 (249.11-0ubuntu3.22 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libapt-pkg6.0 [2.4.10] (2.4.14 Ubuntu:22.04/jammy-updates [arm64])
-Conf libapt-pkg6.0 (2.4.14 Ubuntu:22.04/jammy-updates [arm64])
-Inst tar [1.34+dfsg-1ubuntu0.1.22.04.1] (1.34+dfsg-1ubuntu0.1.22.04.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf tar (1.34+dfsg-1ubuntu0.1.22.04.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst dpkg [1.21.1ubuntu2.2] (1.21.1ubuntu2.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf dpkg (1.21.1ubuntu2.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst gzip [1.10-4ubuntu4.1] (1.10-4ubuntu4.2 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf gzip (1.10-4ubuntu4.2 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst login [1:4.8.1-2ubuntu2.1] (1:4.8.1-2ubuntu2.2 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf login (1:4.8.1-2ubuntu2.2 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst ncurses-bin [6.3-2ubuntu0.1] (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf ncurses-bin (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst perl-base [5.34.0-3ubuntu1.2] (5.34.0-3ubuntu1.9 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf perl-base (5.34.0-3ubuntu1.9 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst sed [4.8-1ubuntu2] (4.8-1ubuntu2.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf sed (4.8-1ubuntu2.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst util-linux [2.37.2-4ubuntu3] (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf util-linux (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libc-bin [2.35-0ubuntu3.4] (2.35-0ubuntu3.15 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libc-bin (2.35-0ubuntu3.15 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst ncurses-base [6.3-2ubuntu0.1] (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [all])
-Conf ncurses-base (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [all])
-Inst gpgv [2.2.27-3ubuntu2.1] (2.2.27-3ubuntu2.5 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf gpgv (2.2.27-3ubuntu2.5 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libp11-kit0 [0.24.0-6build1] (0.24.0-6ubuntu0.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libp11-kit0 (0.24.0-6ubuntu0.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libtasn1-6 [4.18.0-4build1] (4.18.0-4ubuntu0.2 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libtasn1-6 (4.18.0-4ubuntu0.2 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libgnutls30 [3.7.3-4ubuntu1.2] (3.7.3-4ubuntu1.9 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libgnutls30 (3.7.3-4ubuntu1.9 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libseccomp2 [2.5.3-2ubuntu2] (2.5.3-2ubuntu3~22.04.1 Ubuntu:22.04/jammy-updates [arm64])
-Conf libseccomp2 (2.5.3-2ubuntu3~22.04.1 Ubuntu:22.04/jammy-updates [arm64])
-Inst apt [2.4.10] (2.4.14 Ubuntu:22.04/jammy-updates [arm64])
-Conf apt (2.4.14 Ubuntu:22.04/jammy-updates [arm64])
-Inst libpam0g [1.4.0-11ubuntu2.3] (1.4.0-11ubuntu2.8 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libpam0g (1.4.0-11ubuntu2.8 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libpam-modules-bin [1.4.0-11ubuntu2.3] (1.4.0-11ubuntu2.8 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64]) [libpam-modules:arm64 on libpam-modules-bin:arm64] [libpam-modules:arm64 ]
-Conf libpam-modules-bin (1.4.0-11ubuntu2.8 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64]) [libpam-modules:arm64 ]
-Inst libpam-modules [1.4.0-11ubuntu2.3] (1.4.0-11ubuntu2.8 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libpam-modules (1.4.0-11ubuntu2.8 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst logsave [1.46.5-2ubuntu1.1] (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64])
-Inst libext2fs2 [1.46.5-2ubuntu1.1] (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64]) [e2fsprogs:arm64 on libext2fs2:arm64] [e2fsprogs:arm64 ]
-Conf libext2fs2 (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64]) [e2fsprogs:arm64 ]
-Inst e2fsprogs [1.46.5-2ubuntu1.1] (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64])
-Inst mount [2.37.2-4ubuntu3] (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libattr1 [1:2.5.1-1build1] (1:2.5.1-1ubuntu0.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libattr1 (1:2.5.1-1ubuntu0.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libblkid1 [2.37.2-4ubuntu3] (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libblkid1 (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libcap2 [1:2.44-1ubuntu0.22.04.1] (1:2.44-1ubuntu0.22.04.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libcap2 (1:2.44-1ubuntu0.22.04.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libcom-err2 [1.46.5-2ubuntu1.1] (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64])
-Conf libcom-err2 (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64])
-Inst libk5crypto3 [1.19.2-2ubuntu0.2] (1.19.2-2ubuntu0.10 Ubuntu:22.04/jammy-updates [arm64])
-Conf libk5crypto3 (1.19.2-2ubuntu0.10 Ubuntu:22.04/jammy-updates [arm64])
-Inst libkrb5support0 [1.19.2-2ubuntu0.2] (1.19.2-2ubuntu0.10 Ubuntu:22.04/jammy-updates [arm64]) [libkrb5-3:arm64 ]
-Conf libkrb5support0 (1.19.2-2ubuntu0.10 Ubuntu:22.04/jammy-updates [arm64]) [libkrb5-3:arm64 ]
-Inst libkrb5-3 [1.19.2-2ubuntu0.2] (1.19.2-2ubuntu0.10 Ubuntu:22.04/jammy-updates [arm64]) [libgssapi-krb5-2:arm64 ]
-Conf libkrb5-3 (1.19.2-2ubuntu0.10 Ubuntu:22.04/jammy-updates [arm64]) [libgssapi-krb5-2:arm64 ]
-Inst libgssapi-krb5-2 [1.19.2-2ubuntu0.2] (1.19.2-2ubuntu0.10 Ubuntu:22.04/jammy-updates [arm64])
-Conf libgssapi-krb5-2 (1.19.2-2ubuntu0.10 Ubuntu:22.04/jammy-updates [arm64])
-Inst libssl3 [3.0.2-0ubuntu1.10] (3.0.2-0ubuntu1.29 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libssl3 (3.0.2-0ubuntu1.29 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libmount1 [2.37.2-4ubuntu3] (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libmount1 (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libpam-runtime [1.4.0-11ubuntu2.3] (1.4.0-11ubuntu2.8 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [all])
-Conf libpam-runtime (1.4.0-11ubuntu2.8 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [all])
-Inst libsmartcols1 [2.37.2-4ubuntu3] (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libsmartcols1 (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libncurses6 [6.3-2ubuntu0.1] (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64]) []
-Inst libncursesw6 [6.3-2ubuntu0.1] (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64]) []
-Inst libtinfo6 [6.3-2ubuntu0.1] (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libtinfo6 (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libuuid1 [2.37.2-4ubuntu3] (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libuuid1 (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst passwd [1:4.8.1-2ubuntu2.1] (1:4.8.1-2ubuntu2.2 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf passwd (1:4.8.1-2ubuntu2.2 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libprocps8 [2:3.3.17-6ubuntu2] (2:3.3.17-6ubuntu2.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Inst libss2 [1.46.5-2ubuntu1.1] (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64])
-Inst procps [2:3.3.17-6ubuntu2] (2:3.3.17-6ubuntu2.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf logsave (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64])
-Conf e2fsprogs (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64])
-Conf mount (2.37.2-4ubuntu3.6 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libncurses6 (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libncursesw6 (6.3-2ubuntu0.3 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libprocps8 (2:3.3.17-6ubuntu2.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])
-Conf libss2 (1.46.5-2ubuntu1.2 Ubuntu:22.04/jammy-updates [arm64])
-Conf procps (2:3.3.17-6ubuntu2.1 Ubuntu:22.04/jammy-updates, Ubuntu:22.04/jammy-security [arm64])

agent/test-data/package_updates/dnf4_rocky9_check_update.txt

@@ -1,111 +0,0 @@
-
-alternatives.aarch64                     1.24-2.el9                       baseos
-audit-libs.aarch64                       3.1.5-8.el9                      baseos
-basesystem.noarch                        11-13.el9.0.1                    baseos
-bash.aarch64                             5.1.8-9.el9                      baseos
-binutils.aarch64                         2.35.2-72.el9                    baseos
-binutils-gold.aarch64                    2.35.2-72.el9                    baseos
-bzip2-libs.aarch64                       1.0.8-11.el9                     baseos
-ca-certificates.noarch                   2025.2.80_v9.0.305-91.el9        baseos
-coreutils-single.aarch64                 8.32-41.el9_8.1                  baseos
-cracklib.aarch64                         2.9.6-28.el9                     baseos
-cracklib-dicts.aarch64                   2.9.6-28.el9                     baseos
-crypto-policies.noarch                   20260224-1.gitea0f072.el9        baseos
-crypto-policies-scripts.noarch           20260224-1.gitea0f072.el9        baseos
-curl-minimal.aarch64                     7.76.1-40.el9_8.5                baseos
-cyrus-sasl-lib.aarch64                   2.1.27-22.el9_7                  baseos
-dnf.noarch                               4.14.0-34.el9_8.rocky.0.1        baseos
-dnf-data.noarch                          4.14.0-34.el9_8.rocky.0.1        baseos
-elfutils-debuginfod-client.aarch64       0.194-1.el9.rocky.0.1            baseos
-elfutils-default-yama-scope.noarch       0.194-1.el9.rocky.0.1            baseos
-elfutils-libelf.aarch64                  0.194-1.el9.rocky.0.1            baseos
-elfutils-libs.aarch64                    0.194-1.el9.rocky.0.1            baseos
-expat.aarch64                            2.5.0-6.el9_8.3                  baseos
-file-libs.aarch64                        5.39-17.el9                      baseos
-filesystem.aarch64                       3.16-5.el9                       baseos
-findutils.aarch64                        1:4.8.0-7.el9                    baseos
-gdbm-libs.aarch64                        1:1.23-1.el9                     baseos
-glib2.aarch64                            2.68.4-19.el9_8.10               baseos
-glibc.aarch64                            2.34-275.el9_8                   baseos
-glibc-common.aarch64                     2.34-275.el9_8                   baseos
-glibc-minimal-langpack.aarch64           2.34-275.el9_8                   baseos
-gnupg2.aarch64                           2.3.3-5.el9_7                    baseos
-gnutls.aarch64                           3.8.10-8.el9_8                   baseos
-gzip.aarch64                             1.12-2.el9_8                     baseos
-ima-evm-utils.aarch64                    1.6.2-2.el9.rocky.0.2            baseos
-krb5-libs.aarch64                        1.21.1-10.el9_8                  baseos
-less.aarch64                             590-6.el9                        baseos
-libacl.aarch64                           2.4.0-1.el9_8                    baseos
-libarchive.aarch64                       3.5.3-11.el9_8                   baseos
-libatomic.aarch64                        11.5.0-14.el9                    baseos
-libattr.aarch64                          2.6.0-1.el9_8                    baseos
-libblkid.aarch64                         2.37.4-25.el9                    baseos
-libcap.aarch64                           2.48-10.el9_7.1                  baseos
-libcom_err.aarch64                       1.46.5-8.el9                     baseos
-libcurl-minimal.aarch64                  7.76.1-40.el9_8.5                baseos
-libdb.aarch64                            5.3.28-57.el9_6                  baseos
-libdnf.aarch64                           0.69.0-18.el9.rocky.0.1          baseos
-libeconf.aarch64                         0.4.1-7.el9_8                    baseos
-libevent.aarch64                         2.1.13-1.el9_8                   baseos
-libfdisk.aarch64                         2.37.4-25.el9                    baseos
-libgcc.aarch64                           11.5.0-14.el9                    baseos
-libgcrypt.aarch64                        1.10.0-13.el9_8                  baseos
-libgomp.aarch64                          11.5.0-14.el9                    baseos
-libksba.aarch64                          1.5.1-7.el9                      baseos
-libmount.aarch64                         2.37.4-25.el9                    baseos
-libnghttp2.aarch64                       1.43.0-6.el9_8.2                 baseos
-librepo.aarch64                          1.19.0-1.el9                     baseos
-libselinux.aarch64                       3.6-3.el9                        baseos
-libsemanage.aarch64                      3.6-5.el9_6                      baseos
-libsepol.aarch64                         3.6-3.el9                        baseos
-libsmartcols.aarch64                     2.37.4-25.el9                    baseos
-libsolv.aarch64                          0.7.24-6.el9_8                   baseos
-libstdc++.aarch64                        11.5.0-14.el9                    baseos
-libtasn1.aarch64                         4.16.0-10.el9_8                  baseos
-libusbx.aarch64                          1.0.30-1.el9_8                   baseos
-libuser.aarch64                          0.63-17.el9                      baseos
-libuuid.aarch64                          2.37.4-25.el9                    baseos
-libxml2.aarch64                          2.9.13-14.el9_8.4                baseos
-libzstd.aarch64                          1.5.5-1.el9                      baseos
-mpfr.aarch64                             4.1.0-10.el9                     baseos
-ncurses-base.noarch                      6.2-12.20210508.el9              baseos
-ncurses-libs.aarch64                     6.2-12.20210508.el9              baseos
-nettle.aarch64                           3.10.1-1.el9                     baseos
-openldap.aarch64                         2.6.8-4.el9.0.1                  baseos
-openssl.aarch64                          1:3.5.8-1.el9_8                  baseos
-openssl-libs.aarch64                     1:3.5.8-1.el9_8                  baseos
-p11-kit.aarch64                          0.26.4-1.el9_8                   baseos
-p11-kit-trust.aarch64                    0.26.4-1.el9_8                   baseos
-pam.aarch64                              1.5.1-28.el9_8.1                 baseos
-pcre.aarch64                             8.44-4.el9                       baseos
-pcre2.aarch64                            10.40-6.el9                      baseos
-pcre2-syntax.noarch                      10.40-6.el9                      baseos
-python3.aarch64                          3.9.25-7.el9_8.3                 baseos
-python3-dnf.noarch                       4.14.0-34.el9_8.rocky.0.1        baseos
-python3-hawkey.aarch64                   0.69.0-18.el9.rocky.0.1          baseos
-python3-libdnf.aarch64                   0.69.0-18.el9.rocky.0.1          baseos
-python3-libs.aarch64                     3.9.25-7.el9_8.3                 baseos
-python3-pip-wheel.noarch                 21.3.1-2.el9_8.rocky.0.1         baseos
-python3-rpm.aarch64                      4.16.1.3-40.el9                  baseos
-python3-setuptools-wheel.noarch          53.0.0-15.el9                    baseos
-rocky-gpg-keys.noarch                    9.8-1.2.el9                      baseos
-rocky-release.noarch                     9.8-1.2.el9                      baseos
-rocky-repos.noarch                       9.8-1.2.el9                      baseos
-rootfiles.noarch                         8.1-35.el9                       baseos
-rpm.aarch64                              4.16.1.3-40.el9                  baseos
-rpm-build-libs.aarch64                   4.16.1.3-40.el9                  baseos
-rpm-libs.aarch64                         4.16.1.3-40.el9                  baseos
-rpm-sign-libs.aarch64                    4.16.1.3-40.el9                  baseos
-sed.aarch64                              4.8-10.el9_8                     baseos
-setup.noarch                             2.13.7-10.el9                    baseos
-shadow-utils.aarch64                     2:4.9-16.el9                     baseos
-sqlite-libs.aarch64                      3.34.1-11.el9_8                  baseos
-systemd-libs.aarch64                     252-67.el9_8.6.rocky.0.1         baseos
-tar.aarch64                              2:1.34-13.el9_8                  baseos
-tpm2-tss.aarch64                         3.2.3-1.el9                      baseos
-tzdata.noarch                            2026c-1.el9_8                    baseos
-usermode.aarch64                         1.114-7.el9                      baseos
-util-linux.aarch64                       2.37.4-25.el9                    baseos
-util-linux-core.aarch64                  2.37.4-25.el9                    baseos
-vim-minimal.aarch64                      2:8.2.2637-26.el9_8.21           baseos
-yum.noarch                               4.14.0-34.el9_8.rocky.0.1        baseos

agent/test-data/package_updates/dnf4_rocky9_check_update_security.txt

@@ -1,54 +0,0 @@
-
-binutils.aarch64                        2.35.2-72.el9                     baseos
-binutils-gold.aarch64                   2.35.2-72.el9                     baseos
-bzip2-libs.aarch64                      1.0.8-11.el9                      baseos
-coreutils-single.aarch64                8.32-41.el9_8.1                   baseos
-curl-minimal.aarch64                    7.76.1-40.el9_8.5                 baseos
-expat.aarch64                           2.5.0-6.el9_8.3                   baseos
-file-libs.aarch64                       5.39-17.el9                       baseos
-glib2.aarch64                           2.68.4-19.el9_8.10                baseos
-glibc.aarch64                           2.34-275.el9_8                    baseos
-glibc-common.aarch64                    2.34-275.el9_8                    baseos
-glibc-minimal-langpack.aarch64          2.34-275.el9_8                    baseos
-gnupg2.aarch64                          2.3.3-5.el9_7                     baseos
-gnutls.aarch64                          3.8.10-8.el9_8                    baseos
-gzip.aarch64                            1.12-2.el9_8                      baseos
-krb5-libs.aarch64                       1.21.1-10.el9_8                   baseos
-less.aarch64                            590-6.el9                         baseos
-libacl.aarch64                          2.4.0-1.el9_8                     baseos
-libarchive.aarch64                      3.5.3-11.el9_8                    baseos
-libatomic.aarch64                       11.5.0-14.el9                     baseos
-libattr.aarch64                         2.6.0-1.el9_8                     baseos
-libblkid.aarch64                        2.37.4-25.el9                     baseos
-libcap.aarch64                          2.48-10.el9_7.1                   baseos
-libcurl-minimal.aarch64                 7.76.1-40.el9_8.5                 baseos
-libevent.aarch64                        2.1.13-1.el9_8                    baseos
-libfdisk.aarch64                        2.37.4-25.el9                     baseos
-libgcc.aarch64                          11.5.0-14.el9                     baseos
-libgcrypt.aarch64                       1.10.0-13.el9_8                   baseos
-libgomp.aarch64                         11.5.0-14.el9                     baseos
-libmount.aarch64                        2.37.4-25.el9                     baseos
-libnghttp2.aarch64                      1.43.0-6.el9_8.2                  baseos
-libsmartcols.aarch64                    2.37.4-25.el9                     baseos
-libsolv.aarch64                         0.7.24-6.el9_8                    baseos
-libstdc++.aarch64                       11.5.0-14.el9                     baseos
-libtasn1.aarch64                        4.16.0-10.el9_8                   baseos
-libuuid.aarch64                         2.37.4-25.el9                     baseos
-libxml2.aarch64                         2.9.13-14.el9_8.4                 baseos
-ncurses-base.noarch                     6.2-12.20210508.el9               baseos
-ncurses-libs.aarch64                    6.2-12.20210508.el9               baseos
-openssl.aarch64                         1:3.5.8-1.el9_8                   baseos
-openssl-libs.aarch64                    1:3.5.8-1.el9_8                   baseos
-p11-kit.aarch64                         0.26.4-1.el9_8                    baseos
-p11-kit-trust.aarch64                   0.26.4-1.el9_8                    baseos
-pam.aarch64                             1.5.1-28.el9_8.1                  baseos
-python3.aarch64                         3.9.25-7.el9_8.3                  baseos
-python3-libs.aarch64                    3.9.25-7.el9_8.3                  baseos
-python3-setuptools-wheel.noarch         53.0.0-15.el9                     baseos
-shadow-utils.aarch64                    2:4.9-16.el9                      baseos
-sqlite-libs.aarch64                     3.34.1-11.el9_8                   baseos
-systemd-libs.aarch64                    252-67.el9_8.6.rocky.0.1          baseos
-tar.aarch64                             2:1.34-13.el9_8                   baseos
-util-linux.aarch64                      2.37.4-25.el9                     baseos
-util-linux-core.aarch64                 2.37.4-25.el9                     baseos
-vim-minimal.aarch64                     2:8.2.2637-26.el9_8.21            baseos

agent/test-data/package_updates/dnf5_fedora42_check_update.txt

@@ -1,20 +0,0 @@
-dnf5.aarch64                             5.2.18.0-3.fc42  updates
-dnf5-plugins.aarch64                     5.2.18.0-3.fc42  updates
-elfutils-default-yama-scope.noarch       0.195-1.fc42     updates
-elfutils-libelf.aarch64                  0.195-1.fc42     updates
-elfutils-libs.aarch64                    0.195-1.fc42     updates
-fedora-release-common.noarch             42-31            updates
-fedora-release-container.noarch          42-31            updates
-fedora-release-identity-container.noarch 42-31            updates
-glibc.aarch64                            2.41-18.fc42     updates
-glibc-common.aarch64                     2.41-18.fc42     updates
-glibc-minimal-langpack.aarch64           2.41-18.fc42     updates
-krb5-libs.aarch64                        1.21.3-7.fc42    updates
-libdnf5.aarch64                          5.2.18.0-3.fc42  updates
-libdnf5-cli.aarch64                      5.2.18.0-3.fc42  updates
-libsolv.aarch64                          0.7.37-2.fc42    updates
-openssl-libs.aarch64                     1:3.2.6-4.fc42   updates
-rpm-sequoia.aarch64                      1.10.2-2.fc42    updates
-tzdata.noarch                            2026b-1.fc42     updates
-vim-data.noarch                          2:9.2.390-1.fc42 updates
-vim-minimal.aarch64                      2:9.2.390-1.fc42 updates

agent/test-data/package_updates/dnf5_fedora42_check_update_security.txt

@@ -1,5 +0,0 @@
-krb5-libs.aarch64    1.21.3-7.fc42    updates
-openssl-libs.aarch64 1:3.2.6-4.fc42   updates
-rpm-sequoia.aarch64  1.10.2-2.fc42    updates
-vim-data.noarch      2:9.2.390-1.fc42 updates
-vim-minimal.aarch64  2:9.2.390-1.fc42 updates

agent/test-data/package_updates/pacman_checkupdates.txt

@@ -1,4 +0,0 @@
-libpcap 1.10.7-1 -> 1.11.0-1
-libsecret 0.21.7-1 -> 0.21.8.2-1
-libtirpc 1.3.7-1 -> 1.3.8-1
-tzdata 2026c-1 -> 2026d-1

agent/test-data/package_updates/zypper_leap155_list_patches_security.txt

@@ -1,15 +0,0 @@
-Warning: Repository 'Update repository of openSUSE Backports' metadata expired since 2025-03-02 19:18:12 UTC.
-Warning: Repository 'Main Update Repository' metadata expired since 2025-08-30 08:17:31 UTC.
-Warning: Repository 'Update Repository (Non-Oss)' metadata expired since 2025-04-10 11:03:28 UTC.
-
-
-
-Repository                                                   | Name                        | Category | Severity  | Interactive | Status | Summary
--------------------------------------------------------------+-----------------------------+----------+-----------+-------------+--------+--------------------------------
-Update repository with updates from SUSE Linux Enterprise 15 | openSUSE-SLE-15.5-2024-3765 | security | moderate  | ---         | needed | Security update for openssl-1_1
-Update repository with updates from SUSE Linux Enterprise 15 | openSUSE-SLE-15.5-2024-3926 | security | moderate  | ---         | needed | Security update for curl
-Update repository with updates from SUSE Linux Enterprise 15 | openSUSE-SLE-15.5-2024-4078 | security | important | ---         | needed | Security update for glib2
-Update repository with updates from SUSE Linux Enterprise 15 | openSUSE-SLE-15.5-2024-4359 | security | moderate  | ---         | needed | Security update for curl
-
-4 patches needed (4 security patches)
-

agent/test-data/package_updates/zypper_leap155_list_updates.txt

@@ -1,29 +0,0 @@
-Warning: Repository 'Update repository of openSUSE Backports' metadata expired since 2025-03-02 19:18:12 UTC.
-Warning: Repository 'Main Update Repository' metadata expired since 2025-08-30 08:17:31 UTC.
-Warning: Repository 'Update Repository (Non-Oss)' metadata expired since 2025-04-10 11:03:28 UTC.
-
-
-S  | Repository                                                   | Name               | Current Version                          | Available Version                        | Arch
----+--------------------------------------------------------------+--------------------+------------------------------------------+------------------------------------------+--------
-v  | Update repository with updates from SUSE Linux Enterprise 15 | aaa_base           | 84.87+git20180409.04c9dae-150300.10.20.1 | 84.87+git20180409.04c9dae-150300.10.23.1 | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | bash               | 4.4-150400.25.22                         | 4.4-150400.27.3.2                        | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | bash-sh            | 4.4-150400.25.22                         | 4.4-150400.27.3.2                        | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | crypto-policies    | 20210917.c9d86d1-150400.3.6.1            | 20210917.c9d86d1-150400.3.8.1            | noarch
-v  | Update repository with updates from SUSE Linux Enterprise 15 | curl               | 8.0.1-150400.5.50.1                      | 8.0.1-150400.5.59.1                      | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | glibc              | 2.31-150300.86.3                         | 2.31-150300.89.2                         | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libcom_err2        | 1.46.4-150400.3.6.2                      | 1.46.4-150400.3.9.2                      | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libcurl4           | 8.0.1-150400.5.50.1                      | 8.0.1-150400.5.59.1                      | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libgcc_s1          | 13.3.0+git8781-150000.1.12.1             | 14.2.0+git10526-150000.1.6.1             | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libglib-2_0-0      | 2.70.5-150400.3.14.1                     | 2.70.5-150400.3.17.1                     | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libopenssl1_1      | 1.1.1l-150500.17.34.1                    | 1.1.1l-150500.17.37.1                    | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libopenssl1_1-hmac | 1.1.1l-150500.17.34.1                    | 1.1.1l-150500.17.37.1                    | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libreadline7       | 7.0-150400.25.22                         | 7.0-150400.27.3.2                        | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libsolv-tools      | 0.7.30-150400.3.27.2                     | 0.7.31-150500.6.5.1                      | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libsolv-tools-base | 0.7.30-150400.3.27.2                     | 0.7.31-150500.6.5.1                      | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libstdc++6         | 13.3.0+git8781-150000.1.12.1             | 14.2.0+git10526-150000.1.6.1             | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libudev1           | 249.17-150400.8.43.1                     | 249.17-150400.8.46.1                     | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | libzypp            | 17.35.8-150500.6.13.1                    | 17.35.16-150500.6.31.1                   | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | login_defs         | 4.8.1-150400.10.21.1                     | 4.8.1-150400.10.24.1                     | noarch
-v  | Update repository with updates from SUSE Linux Enterprise 15 | openssl-1_1        | 1.1.1l-150500.17.34.1                    | 1.1.1l-150500.17.37.1                    | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | shadow             | 4.8.1-150400.10.21.1                     | 4.8.1-150400.10.24.1                     | aarch64
-v  | Update repository with updates from SUSE Linux Enterprise 15 | zypper             | 1.14.76-150500.6.6.15                    | 1.14.78-150500.6.14.1                    | aarch64

agent/test-data/package_updates/zypper_leap156_list_updates_none.txt

@@ -1,3 +0,0 @@
-Warning: Repository 'Update repository of openSUSE Backports' metadata expired since 2026-07-10 11:19:15 UTC.
-
-

agent/zfs/zfs.go

@@ -33,15 +33,11 @@ var ErrNoZfs = errors.New("zfs utilities unavailable")
 
 // PoolStat is a snapshot of a ZFS pool's capacity and health.
 type PoolStat struct {
-	DisplayName string // optional friendly name; Name remains the stable key
-	MountID     string // Btrfs filesystem identity, empty for other backends
-	IODevice    string // sole Btrfs member device, if known
-	Raw         bool   // physical accounting rather than usable filesystem space
-	Name        string
-	Size        uint64 // total capacity in bytes
-	Alloc       uint64 // allocated bytes
-	Free        uint64 // free bytes
-	Health      string // ONLINE, DEGRADED, FAULTED, ...
+	Name   string
+	Size   uint64 // total capacity in bytes
+	Alloc  uint64 // allocated bytes
+	Free   uint64 // free bytes
+	Health string // ONLINE, DEGRADED, FAULTED, ...
 }
 
 // PoolKernelStat is the inexpensive pool telemetry exposed by the ZFS kernel.
@@ -70,9 +66,6 @@ type Dataset struct {
 // PoolStats returns capacity and health for all pools on the system using
 // `zpool list`. Frequent health and I/O sampling uses PoolKernelStats instead.
 func PoolStats() ([]PoolStat, error) {
-	if err := checkZfsDevice(); err != nil {
-		return nil, err
-	}
 	out, err := commandOutput("zpool", "list", "-Hp", "-o", "name,size,alloc,free,health")
 	if err != nil {
 		var exitErr *exec.ExitError
@@ -87,9 +80,6 @@ func PoolStats() ([]PoolStat, error) {
 // Datasets returns all datasets on the system with usage and mountpoint
 // information using `zfs list` (recursive by default).
 func Datasets() ([]Dataset, error) {
-	if err := checkZfsDevice(); err != nil {
-		return nil, err
-	}
 	out, err := commandOutput("zfs", "list", "-Hp", "-o", "name,used,avail,mountpoint")
 	if err != nil {
 		return nil, fmt.Errorf("zfs list: %w", err)

agent/zfs/zfs_linux.go

@@ -13,10 +13,7 @@ import (
 	"strings"
 )
 
-var (
-	procZfsPath = "/proc/spl/kstat/zfs"
-	devZfsPath  = "/dev/zfs"
-)
+var procZfsPath = "/proc/spl/kstat/zfs"
 
 func ARCSize() (uint64, error) {
 	file, err := os.Open(filepath.Join(procZfsPath, "arcstats"))
@@ -43,19 +40,6 @@ func ARCSize() (uint64, error) {
 	return 0, fmt.Errorf("size field not found in arcstats")
 }
 
-// checkZfsDevice lets containers without /dev/zfs fail fast instead of
-// waiting for ZFS utility commands to time out.
-func checkZfsDevice() error {
-	_, err := os.Stat(devZfsPath)
-	if err != nil {
-		if errors.Is(err, os.ErrNotExist) {
-			return ErrNoZfs
-		}
-		return err
-	}
-	return nil
-}
-
 // PoolKernelStats reads pool state and cumulative I/O counters directly from
 // procfs. These kstats are the same interfaces used by node_exporter's Linux
 // ZFS collector and avoid keeping a `zpool iostat` subprocess alive.

agent/zfs/zfs_linux_test.go

@@ -88,66 +88,3 @@ func TestReadObjsetIORequiresAllCounters(t *testing.T) {
 	_, _, err := readObjsetIO(path)
 	require.Error(t, err)
 }
-
-func TestCollectorsSkipCommandsWhenDevZfsMissing(t *testing.T) {
-	root := t.TempDir()
-	oldDevZfsPath := devZfsPath
-	devZfsPath = filepath.Join(root, "missing")
-	t.Cleanup(func() { devZfsPath = oldDevZfsPath })
-
-	oldCommandOutput := commandOutput
-	commandOutput = func(name string, args ...string) ([]byte, error) {
-		t.Fatalf("unexpected %s call with %v", name, args)
-		return nil, nil
-	}
-	t.Cleanup(func() { commandOutput = oldCommandOutput })
-
-	_, err := PoolStats()
-	assert.ErrorIs(t, err, ErrNoZfs)
-	_, err = Datasets()
-	assert.ErrorIs(t, err, ErrNoZfs)
-}
-
-func TestDatasetsDelegatesWhenDevZfsPresent(t *testing.T) {
-	oldDevZfsPath := devZfsPath
-	devZfsPath = filepath.Join(t.TempDir(), "zfs")
-	require.NoError(t, os.WriteFile(devZfsPath, nil, 0o644))
-	t.Cleanup(func() { devZfsPath = oldDevZfsPath })
-
-	oldCommandOutput := commandOutput
-	commandOutput = func(name string, args ...string) ([]byte, error) {
-		assert.Equal(t, "zfs", name)
-		assert.Equal(t, []string{"list", "-Hp", "-o", "name,used,avail,mountpoint"}, args)
-		return []byte("tank\t50\t50\t/tank\n"), nil
-	}
-	t.Cleanup(func() { commandOutput = oldCommandOutput })
-
-	datasets, err := Datasets()
-	require.NoError(t, err)
-	assert.Equal(t, []Dataset{{Name: "tank", Used: 50, Avail: 50, Mountpoint: "/tank"}}, datasets)
-}
-
-func TestPoolStatsDelegatesToZpoolWhenDevZfsPresent(t *testing.T) {
-	root := t.TempDir()
-	devFile := filepath.Join(root, "zfs")
-	require.NoError(t, os.WriteFile(devFile, []byte(""), 0o644))
-
-	oldDevZfsPath := devZfsPath
-	devZfsPath = devFile
-	t.Cleanup(func() { devZfsPath = oldDevZfsPath })
-
-	oldCommandOutput := commandOutput
-	called := false
-	commandOutput = func(name string, args ...string) ([]byte, error) {
-		called = true
-		assert.Equal(t, "zpool", name)
-		assert.Equal(t, []string{"list", "-Hp", "-o", "name,size,alloc,free,health"}, args)
-		return []byte("tank\t100\t50\t50\tONLINE\n"), nil
-	}
-	t.Cleanup(func() { commandOutput = oldCommandOutput })
-
-	pools, err := PoolStats()
-	require.NoError(t, err)
-	assert.True(t, called)
-	assert.Equal(t, []PoolStat{{Name: "tank", Size: 100, Alloc: 50, Free: 50, Health: "ONLINE"}}, pools)
-}

agent/zfs/zfs_nonlinux.go

@@ -1,9 +0,0 @@
-//go:build !linux
-
-package zfs
-
-// The /dev/zfs probe is Linux-specific. Other platforms detect availability
-// through the ZFS utilities themselves.
-func checkZfsDevice() error {
-	return nil
-}

agent/zfs/zfs_nonlinux_test.go

@@ -1,33 +0,0 @@
-//go:build testing && !linux
-
-package zfs
-
-import (
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestCollectorsUseUtilitiesOnNonLinux(t *testing.T) {
-	oldCommandOutput := commandOutput
-	commandOutput = func(name string, args ...string) ([]byte, error) {
-		switch name {
-		case "zpool":
-			return []byte("tank\t100\t50\t50\tONLINE\n"), nil
-		case "zfs":
-			return []byte("tank\t50\t50\t/tank\n"), nil
-		default:
-			t.Fatalf("unexpected command %s", name)
-			return nil, nil
-		}
-	}
-	t.Cleanup(func() { commandOutput = oldCommandOutput })
-
-	pools, err := PoolStats()
-	require.NoError(t, err)
-	assert.Equal(t, []PoolStat{{Name: "tank", Size: 100, Alloc: 50, Free: 50, Health: "ONLINE"}}, pools)
-	datasets, err := Datasets()
-	require.NoError(t, err)
-	assert.Equal(t, []Dataset{{Name: "tank", Used: 50, Avail: 50, Mountpoint: "/tank"}}, datasets)
-}

agent/zfs_pool.go

@@ -0,0 +1,320 @@
+package agent
+
+import (
+	"log/slog"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/henrygd/beszel/agent/zfs"
+	"github.com/henrygd/beszel/internal/entities/system"
+	zfsentity "github.com/henrygd/beszel/internal/entities/zfs"
+)
+
+// zfsDatasetUsage holds usage values for a ZFS dataset mountpoint.
+type zfsDatasetUsage struct {
+	used  uint64
+	avail uint64
+}
+
+// datasetUsageRefreshInterval controls how often `zfs list` is re-run for the
+// mountpoint usage map. Dataset inventory changes rarely.
+const datasetUsageRefreshInterval = 5 * time.Minute
+
+// poolStatsRefreshInterval controls how often `zpool list` is re-run for pool
+// capacity. Health and I/O are read from procfs on Linux, so the utility only
+// needs to refresh slow-moving space accounting.
+const poolStatsRefreshInterval = time.Minute
+
+type poolKernelSample struct {
+	nread  uint64
+	nwrite uint64
+	at     time.Time
+}
+
+// ZfsManager collects ZFS pool and dataset statistics. Collection functions
+// are fields so unit tests can substitute them (same pattern as
+// diskDiscovery.usageFn). It is safe for concurrent use by a single goroutine
+// only; callers must hold the agent lock like updateDiskUsage does.
+type ZfsManager struct {
+	poolStatsFn    func() ([]zfs.PoolStat, error)       // capacity/health source
+	datasetsFn     func() ([]zfs.Dataset, error)        // dataset inventory source
+	kernelStatsFn  func() ([]zfs.PoolKernelStat, error) // procfs pool state/I/O source
+	poolStatusesFn func() ([]zfs.PoolStatus, error)     // scrub/vdev detail source
+
+	poolData      []zfs.PoolStat // cached pool inventory (TTL below)
+	lastPoolStats time.Time
+	kernelSamples map[string]poolKernelSample
+
+	datasetUsage     map[string]zfsDatasetUsage // mountpoint -> usage
+	lastUsageRefresh time.Time
+
+	// Detail data (pools, vdevs, scrub, datasets) is cached and refreshed on
+	// an interval. Accessed from handler goroutines, so it is mutex-protected.
+	detailMu          sync.Mutex
+	detail            *zfsentity.ZfsData
+	lastDetailRefresh time.Time
+	detailInterval    time.Duration
+}
+
+// newZfsManager creates a ZfsManager wired to the system's ZFS utilities.
+func newZfsManager() *ZfsManager {
+	return &ZfsManager{
+		poolStatsFn:    zfs.PoolStats,
+		datasetsFn:     zfs.Datasets,
+		kernelStatsFn:  zfs.PoolKernelStats,
+		poolStatusesFn: zfs.PoolStatuses,
+		detailInterval: time.Hour,
+	}
+}
+
+// Update refreshes systemStats.ZfsPools with the latest pool data. I/O
+// throughput and health come from inexpensive kernel kstats on Linux. Pool
+// capacity and dataset usage come from separately cached utility calls. It is
+// a no-op when ZFS is absent.
+func (zm *ZfsManager) Update(systemStats *system.Stats) {
+	pools := zm.poolStats()
+	if len(pools) == 0 {
+		return
+	}
+
+	kernelStats, ioRates := zm.kernelStats()
+
+	if systemStats.ZfsPools == nil {
+		systemStats.ZfsPools = make(map[string]*system.ZfsPool, len(pools))
+	}
+	for i := range pools {
+		pool := &pools[i]
+		// Full precision, matching the dataset values below; the frontend
+		// formats any magnitude.
+		stats := &system.ZfsPool{
+			Total:  float64(pool.Size) / (1024 * 1024 * 1024),
+			Used:   float64(pool.Alloc) / (1024 * 1024 * 1024),
+			Health: pool.Health,
+		}
+		if kernel, exists := kernelStats[pool.Name]; exists && kernel.Health != "" {
+			stats.Health = kernel.Health
+		}
+		if io, exists := ioRates[pool.Name]; exists {
+			stats.ReadBytes = io.NRead
+			stats.WriteBytes = io.NWrite
+		}
+		slog.Debug("ZFS pool sample", "pool", pool.Name, "health", stats.Health, "used_gb", stats.Used, "read_bps", stats.ReadBytes, "write_bps", stats.WriteBytes)
+		systemStats.ZfsPools[pool.Name] = stats
+	}
+
+}
+
+// poolStats returns the cached pool inventory, re-running `zpool list` at most
+// every poolStatsRefreshInterval. On failure the previous inventory is
+// retained and the refresh is retried on the next cadence.
+func (zm *ZfsManager) poolStats() []zfs.PoolStat {
+	if zm.lastPoolStats.IsZero() || time.Since(zm.lastPoolStats) >= poolStatsRefreshInterval {
+		pools, err := zm.poolStatsFn()
+		if err != nil {
+			slog.Debug("ZFS pool stats unavailable", "err", err)
+		} else {
+			zm.poolData = pools
+		}
+		zm.lastPoolStats = time.Now()
+	}
+	return zm.poolData
+}
+
+// kernelStats reads cumulative pool counters and converts them to per-second
+// rates. Counter decreases indicate a pool export/import and reset the
+// baseline instead of producing an underflow spike.
+func (zm *ZfsManager) kernelStats() (map[string]zfs.PoolKernelStat, map[string]zfs.PoolIoStats) {
+	if zm.kernelStatsFn == nil {
+		return nil, nil
+	}
+	stats, err := zm.kernelStatsFn()
+	if err != nil {
+		slog.Debug("ZFS kernel stats unavailable", "err", err)
+		return nil, nil
+	}
+	now := time.Now()
+	byName := make(map[string]zfs.PoolKernelStat, len(stats))
+	rates := make(map[string]zfs.PoolIoStats, len(stats))
+	nextSamples := make(map[string]poolKernelSample, len(stats))
+	for _, stat := range stats {
+		byName[stat.Name] = stat
+		if previous, ok := zm.kernelSamples[stat.Name]; ok && now.After(previous.at) &&
+			stat.NRead >= previous.nread && stat.NWrite >= previous.nwrite {
+			seconds := now.Sub(previous.at).Seconds()
+			rates[stat.Name] = zfs.PoolIoStats{
+				NRead:  uint64(float64(stat.NRead-previous.nread) / seconds),
+				NWrite: uint64(float64(stat.NWrite-previous.nwrite) / seconds),
+			}
+		}
+		nextSamples[stat.Name] = poolKernelSample{nread: stat.NRead, nwrite: stat.NWrite, at: now}
+	}
+	zm.kernelSamples = nextSamples
+	return byName, rates
+}
+
+// refreshDatasetUsage re-runs `zfs list` when the refresh window has elapsed
+// and rebuilds the mountpoint-keyed usage map.
+func (zm *ZfsManager) refreshDatasetUsage() {
+	if !zm.lastUsageRefresh.IsZero() && time.Since(zm.lastUsageRefresh) < datasetUsageRefreshInterval {
+		return
+	}
+	datasets, err := zm.datasetsFn()
+	if err != nil {
+		slog.Debug("ZFS dataset usage unavailable", "err", err)
+	} else {
+		usage := make(map[string]zfsDatasetUsage, len(datasets))
+		for _, ds := range datasets {
+			if ds.Mountpoint != "" && ds.Mountpoint != "-" {
+				usage[ds.Mountpoint] = zfsDatasetUsage{used: ds.Used, avail: ds.Avail}
+			}
+		}
+		zm.datasetUsage = usage
+	}
+	zm.lastUsageRefresh = time.Now()
+}
+
+// DatasetUsage returns ZFS dataset usage keyed by mountpoint, refreshed at
+// most every datasetUsageRefreshInterval. On failure the previous map is
+// retained and a debug log is emitted.
+func (zm *ZfsManager) DatasetUsage() map[string]zfsDatasetUsage {
+	zm.refreshDatasetUsage()
+	return zm.datasetUsage
+}
+
+// GetDetail returns ZFS detail data (pool health, scrub, vdevs, datasets).
+// Scheduled requests use the cached snapshot until stale; manual requests can
+// force collection. On failure the previous snapshot is retained.
+func (zm *ZfsManager) GetDetail(force bool) *zfsentity.ZfsData {
+	zm.detailMu.Lock()
+	defer zm.detailMu.Unlock()
+
+	if force || zm.detail == nil || time.Since(zm.lastDetailRefresh) >= zm.detailInterval {
+		if data, err := zm.collectDetail(zm.detail); err != nil {
+			slog.Debug("ZFS detail collection failed", "err", err)
+			if zm.detail == nil {
+				return &zfsentity.ZfsData{}
+			}
+			return &zfsentity.ZfsData{Pools: zm.detail.Pools}
+		} else {
+			zm.detail = data
+			zm.lastDetailRefresh = time.Now()
+		}
+	}
+	if zm.detail == nil {
+		return &zfsentity.ZfsData{}
+	}
+	return zm.detail
+}
+
+// collectDetail builds a ZfsData payload from the current system state.
+func (zm *ZfsManager) collectDetail(previous *zfsentity.ZfsData) (*zfsentity.ZfsData, error) {
+	pools, err := zm.poolStatsFn()
+	if err != nil {
+		return nil, err
+	}
+	if len(pools) == 0 {
+		return &zfsentity.ZfsData{Pools: []*zfsentity.PoolDetail{}, Complete: true}, nil
+	}
+
+	statuses, statusErr := zm.poolStatusesFn()
+	if statusErr != nil {
+		slog.Debug("ZFS pool status unavailable", "err", statusErr)
+	}
+	datasets, datasetsErr := zm.datasetsFn()
+	if datasetsErr != nil {
+		slog.Debug("ZFS datasets unavailable", "err", datasetsErr)
+	}
+
+	statusByPool := make(map[string]zfs.PoolStatus, len(statuses))
+	for _, st := range statuses {
+		statusByPool[st.Name] = st
+	}
+
+	previousByPool := make(map[string]*zfsentity.PoolDetail)
+	if previous != nil {
+		for _, pool := range previous.Pools {
+			if pool != nil {
+				previousByPool[pool.Name] = pool
+			}
+		}
+	}
+
+	data := &zfsentity.ZfsData{Pools: make([]*zfsentity.PoolDetail, 0, len(pools)), Complete: true}
+	for i := range pools {
+		p := &pools[i]
+		detail := &zfsentity.PoolDetail{
+			Name:   p.Name,
+			Health: p.Health,
+			Size:   p.Size,
+			Alloc:  p.Alloc,
+			Free:   p.Free,
+		}
+		if st, ok := statusByPool[p.Name]; statusErr == nil && ok {
+			if st.Scrub.State != "" && st.Scrub.State != "NONE" {
+				detail.Scrub = &zfsentity.Scrub{
+					State:    st.Scrub.State,
+					Progress: st.Scrub.Progress,
+					Errors:   st.Scrub.Errors,
+				}
+			}
+			for _, v := range st.Vdevs {
+				detail.Vdevs = append(detail.Vdevs, &zfsentity.Vdev{
+					Name:         v.Name,
+					State:        v.State,
+					ReadErrs:     v.ReadErrs,
+					WriteErrs:    v.WriteErrs,
+					ChecksumErrs: v.ChecksumErrs,
+				})
+			}
+		} else {
+			if cached := previousByPool[p.Name]; cached != nil {
+				detail.Scrub = cached.Scrub
+				detail.Vdevs = cached.Vdevs
+			}
+		}
+		if datasetsErr == nil {
+			foundDataset := false
+			for _, ds := range datasets {
+				if poolOfDataset(ds.Name) == p.Name {
+					foundDataset = true
+					detail.Datasets = append(detail.Datasets, &zfsentity.Dataset{
+						Name:       ds.Name,
+						Used:       ds.Used,
+						Avail:      ds.Avail,
+						Mountpoint: ds.Mountpoint,
+					})
+				}
+			}
+			if !foundDataset {
+				if cached := previousByPool[p.Name]; cached != nil {
+					detail.Datasets = cached.Datasets
+				}
+			}
+		} else if cached := previousByPool[p.Name]; cached != nil {
+			detail.Datasets = cached.Datasets
+		}
+		data.Pools = append(data.Pools, detail)
+	}
+	return data, nil
+}
+
+// poolOfDataset returns the pool name for a dataset name (everything before
+// the first '/'). Datasets without a separator belong to a pool of the same
+// name.
+func poolOfDataset(name string) string {
+	if idx := strings.IndexByte(name, '/'); idx >= 0 {
+		return name[:idx]
+	}
+	return name
+}
+
+// ZfsMountpoints returns the set of mountpoints backed by ZFS datasets.
+func (zm *ZfsManager) ZfsMountpoints() map[string]bool {
+	usage := zm.DatasetUsage()
+	mountpoints := make(map[string]bool, len(usage))
+	for mountpoint := range usage {
+		mountpoints[mountpoint] = true
+	}
+	return mountpoints
+}

agent/zfs_pool_test.go

@@ -0,0 +1,245 @@
+//go:build testing
+
+package agent
+
+import (
+	"testing"
+	"time"
+
+	"github.com/henrygd/beszel/agent/zfs"
+	"github.com/henrygd/beszel/internal/entities/system"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestUpdatePopulatesZfsPools(t *testing.T) {
+	zm := &ZfsManager{}
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) {
+		return []zfs.PoolStat{{Name: "tank", Size: 23999000000000, Alloc: 12000000000000, Free: 11999000000000, Health: "DEGRADED"}}, nil
+	}
+	zm.datasetsFn = func() ([]zfs.Dataset, error) {
+		return []zfs.Dataset{
+			{Name: "tank/apps", Used: 5000000000000, Avail: 11999000000000, Mountpoint: "/tank/apps"},
+			{Name: "tank/backup", Used: 6000000000000, Avail: 11999000000000, Mountpoint: "/tank/backup"},
+			// Small zvol (Proxmox VM EFI disk): must not round to zero.
+			{Name: "rpool/vm-100-disk-2", Used: 4194304, Avail: 0, Mountpoint: "-"},
+		}, nil
+	}
+	var kernelCalls int
+	zm.kernelStatsFn = func() ([]zfs.PoolKernelStat, error) {
+		kernelCalls++
+		return []zfs.PoolKernelStat{{
+			Name: "tank", Health: "ONLINE",
+			NRead: uint64(kernelCalls-1) * 1250, NWrite: uint64(kernelCalls-1) * 5120,
+		}}, nil
+	}
+
+	var stats system.Stats
+	// The first kernel sample establishes the cumulative-counter baseline.
+	zm.Update(&stats)
+	zm.kernelSamples["tank"] = poolKernelSample{at: time.Now().Add(-time.Second)}
+	zm.Update(&stats)
+	require.NotNil(t, stats.ZfsPools)
+	require.Contains(t, stats.ZfsPools, "tank")
+	assert.InDelta(t, 22350.8105, stats.ZfsPools["tank"].Total, 0.0001) // Size in GiB
+	assert.InDelta(t, 11175.8709, stats.ZfsPools["tank"].Used, 0.0001)  // Alloc in GiB
+	assert.Equal(t, "ONLINE", stats.ZfsPools["tank"].Health)
+	assert.InDelta(t, 1250, stats.ZfsPools["tank"].ReadBytes, 5)
+	assert.InDelta(t, 5120, stats.ZfsPools["tank"].WriteBytes, 5)
+
+}
+
+// TestUpdateKernelStatsMissing verifies pools without a kernel sample report zero
+// I/O instead of erroring.
+func TestUpdateKernelStatsMissing(t *testing.T) {
+	zm := &ZfsManager{}
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) {
+		return []zfs.PoolStat{{Name: "tank", Size: 1, Alloc: 1, Health: "ONLINE"}}, nil
+	}
+	zm.datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
+	zm.kernelStatsFn = func() ([]zfs.PoolKernelStat, error) {
+		return nil, zfs.ErrNoZfs
+	}
+
+	var stats system.Stats
+	zm.Update(&stats)
+	require.NotNil(t, stats.ZfsPools)
+	assert.Equal(t, uint64(0), stats.ZfsPools["tank"].ReadBytes)
+	assert.Equal(t, uint64(0), stats.ZfsPools["tank"].WriteBytes)
+}
+
+func TestUpdateKernelCounterReset(t *testing.T) {
+	zm := &ZfsManager{}
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) {
+		return []zfs.PoolStat{{Name: "tank", Health: "ONLINE"}}, nil
+	}
+	zm.datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
+	zm.kernelSamples = map[string]poolKernelSample{
+		"tank": {nread: 100, nwrite: 200, at: time.Now().Add(-time.Second)},
+	}
+	zm.kernelStatsFn = func() ([]zfs.PoolKernelStat, error) {
+		return []zfs.PoolKernelStat{{Name: "tank", Health: "ONLINE", NRead: 10, NWrite: 20}}, nil
+	}
+
+	var stats system.Stats
+	zm.Update(&stats)
+	assert.Equal(t, uint64(0), stats.ZfsPools["tank"].ReadBytes)
+	assert.Equal(t, uint64(0), stats.ZfsPools["tank"].WriteBytes)
+}
+
+func TestUpdateNoZfs(t *testing.T) {
+	zm := &ZfsManager{}
+	calls := 0
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) {
+		calls++
+		return nil, zfs.ErrNoZfs
+	}
+
+	var stats system.Stats
+	zm.Update(&stats)
+	zm.Update(&stats)
+	assert.Nil(t, stats.ZfsPools)
+	assert.Equal(t, 1, calls, "failed pool discovery should be cached until the next refresh interval")
+}
+
+func TestUpdateEmptyPools(t *testing.T) {
+	zm := &ZfsManager{}
+	calls := 0
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) {
+		calls++
+		return nil, nil
+	}
+
+	var stats system.Stats
+	zm.Update(&stats)
+	zm.Update(&stats)
+	assert.Nil(t, stats.ZfsPools)
+	assert.Equal(t, 1, calls, "an empty pool inventory should be cached until the next refresh interval")
+}
+
+func TestDatasetUsage(t *testing.T) {
+	zm := &ZfsManager{}
+	calls := 0
+	zm.datasetsFn = func() ([]zfs.Dataset, error) {
+		calls++
+		return []zfs.Dataset{
+			{Name: "tank", Used: 12000000000000, Avail: 11999000000000, Mountpoint: "/tank"},
+			{Name: "tank/apps", Used: 1000000000000, Avail: 11999000000000, Mountpoint: "/tank/apps"},
+			{Name: "rpool", Used: 900000000000, Avail: 300000000000, Mountpoint: "-"}, // zvol/unmounted: excluded
+		}, nil
+	}
+
+	usage := zm.DatasetUsage()
+	require.Len(t, usage, 2)
+	assert.Equal(t, zfsDatasetUsage{used: 12000000000000, avail: 11999000000000}, usage["/tank"])
+	assert.Equal(t, zfsDatasetUsage{used: 1000000000000, avail: 11999000000000}, usage["/tank/apps"])
+	assert.Equal(t, 1, calls)
+
+	// Second call within the refresh window must not re-run the collector.
+	zm.DatasetUsage()
+	assert.Equal(t, 1, calls)
+}
+
+func TestDatasetUsageRefreshOnErrorKeepsPrevious(t *testing.T) {
+	zm := &ZfsManager{}
+	zm.datasetsFn = func() ([]zfs.Dataset, error) {
+		return []zfs.Dataset{{Name: "tank", Used: 1, Avail: 1, Mountpoint: "/tank"}}, nil
+	}
+	assert.Len(t, zm.DatasetUsage(), 1)
+
+	// Force refresh window expiry, then a failing collector.
+	zm.lastUsageRefresh = time.Now().Add(-10 * time.Minute)
+	zm.datasetsFn = func() ([]zfs.Dataset, error) {
+		return nil, zfs.ErrNoZfs
+	}
+	usage := zm.DatasetUsage()
+	assert.Len(t, usage, 1, "previous usage should be retained on error")
+}
+
+func TestGetDetailForceRefresh(t *testing.T) {
+	zm := &ZfsManager{detailInterval: time.Hour}
+	poolCalls := 0
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) {
+		poolCalls++
+		return []zfs.PoolStat{{Name: "tank", Alloc: uint64(poolCalls)}}, nil
+	}
+	zm.poolStatusesFn = func() ([]zfs.PoolStatus, error) { return nil, nil }
+	zm.datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
+
+	first := zm.GetDetail(false)
+	assert.True(t, first.Complete)
+	require.Len(t, first.Pools, 1)
+	assert.Equal(t, uint64(1), first.Pools[0].Alloc)
+
+	cached := zm.GetDetail(false)
+	require.Len(t, cached.Pools, 1)
+	assert.Equal(t, uint64(1), cached.Pools[0].Alloc)
+	assert.Equal(t, 1, poolCalls)
+
+	refreshed := zm.GetDetail(true)
+	assert.True(t, refreshed.Complete)
+	require.Len(t, refreshed.Pools, 1)
+	assert.Equal(t, uint64(2), refreshed.Pools[0].Alloc)
+	assert.Equal(t, 2, poolCalls)
+}
+
+func TestGetDetailSuccessfulEmptyInventoryClearsCache(t *testing.T) {
+	zm := &ZfsManager{detailInterval: time.Hour}
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) {
+		return []zfs.PoolStat{{Name: "tank"}}, nil
+	}
+	zm.poolStatusesFn = func() ([]zfs.PoolStatus, error) { return nil, nil }
+	zm.datasetsFn = func() ([]zfs.Dataset, error) { return nil, nil }
+
+	require.Len(t, zm.GetDetail(false).Pools, 1)
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) { return nil, nil }
+	empty := zm.GetDetail(true)
+	assert.True(t, empty.Complete)
+	assert.Empty(t, empty.Pools)
+}
+
+func TestGetDetailFailureReturnsIncompleteCachedInventory(t *testing.T) {
+	zm := &ZfsManager{detailInterval: time.Hour}
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) {
+		return []zfs.PoolStat{{Name: "tank"}}, nil
+	}
+	zm.poolStatusesFn = func() ([]zfs.PoolStatus, error) {
+		return []zfs.PoolStatus{{Name: "tank", Vdevs: []zfs.VdevStatus{{Name: "mirror-0"}}}}, nil
+	}
+	zm.datasetsFn = func() ([]zfs.Dataset, error) {
+		return []zfs.Dataset{{Name: "tank/data"}}, nil
+	}
+	first := zm.GetDetail(false)
+	require.True(t, first.Complete)
+	require.Len(t, first.Pools[0].Vdevs, 1)
+	require.Len(t, first.Pools[0].Datasets, 1)
+
+	zm.poolStatusesFn = func() ([]zfs.PoolStatus, error) { return nil, zfs.ErrNoZfs }
+	zm.datasetsFn = func() ([]zfs.Dataset, error) { return nil, zfs.ErrNoZfs }
+	partial := zm.GetDetail(true)
+	require.True(t, partial.Complete)
+	require.Len(t, partial.Pools[0].Vdevs, 1)
+	require.Len(t, partial.Pools[0].Datasets, 1)
+
+	zm.poolStatsFn = func() ([]zfs.PoolStat, error) { return nil, zfs.ErrNoZfs }
+	lastSuccessfulRefresh := zm.lastDetailRefresh
+	failed := zm.GetDetail(true)
+	assert.False(t, failed.Complete)
+	require.Len(t, failed.Pools, 1)
+	assert.Equal(t, "tank", failed.Pools[0].Name)
+	assert.Equal(t, lastSuccessfulRefresh, zm.lastDetailRefresh)
+}
+
+func TestZfsMountpoints(t *testing.T) {
+	zm := &ZfsManager{}
+	zm.datasetsFn = func() ([]zfs.Dataset, error) {
+		return []zfs.Dataset{
+			{Name: "tank", Mountpoint: "/tank"},
+			{Name: "rpool/ROOT/pve-1", Mountpoint: "/"},
+		}, nil
+	}
+	mountpoints := zm.ZfsMountpoints()
+	assert.Len(t, mountpoints, 2)
+	assert.True(t, mountpoints["/tank"])
+	assert.True(t, mountpoints["/"])
+}

beszel.go

@@ -6,7 +6,7 @@ import "github.com/blang/semver"
 
 const (
 	// Version is the current version of the application.
-	Version = "0.20.0"
+	Version = "0.19.0"
 	// AppName is the name of the application.
 	AppName = "beszel"
 )
@@ -19,6 +19,3 @@ var MinVersionAgentResponse = semver.MustParse("0.13.0")
 
 // MinVersionZfsData is the minimum agent version that supports ZFS detail requests.
 var MinVersionZfsData = semver.MustParse("0.18.9")
-
-// MinVersionNetworkMonitors is the minimum agent version that supports network monitor sync.
-var MinVersionNetworkMonitors = semver.MustParse("0.20.0")

go.mod

@@ -5,25 +5,23 @@ go 1.27.1
 require (
 	github.com/blang/semver v3.5.1+incompatible
 	github.com/coreos/go-systemd/v22 v22.7.0
-	github.com/distribution/reference v0.6.0
 	github.com/ebitengine/purego v0.11.0
-	github.com/fxamacker/cbor/v2 v2.9.4
+	github.com/fxamacker/cbor/v2 v2.9.3
 	github.com/gliderlabs/ssh v0.3.8
-	github.com/lxzan/gws v1.10.2
-	github.com/nicholas-fedor/shoutrrr v0.21.0
-	github.com/opencontainers/go-digest v1.0.0
+	github.com/google/uuid v1.6.0
+	github.com/lxzan/gws v1.10.1
+	github.com/nicholas-fedor/shoutrrr v0.20.0
 	github.com/pocketbase/dbx v1.12.0
-	github.com/pocketbase/pocketbase v0.40.4
+	github.com/pocketbase/pocketbase v0.40.2
 	github.com/shirou/gopsutil/v4 v4.26.8
 	github.com/spf13/cast v1.10.0
 	github.com/spf13/cobra v1.10.2
 	github.com/spf13/pflag v1.0.10
 	github.com/stretchr/testify v1.12.1
-	golang.org/x/crypto v0.57.0
-	golang.org/x/exp v0.0.0-20260908205506-85c1c2202aba
-	golang.org/x/net v0.59.0
-	golang.org/x/oauth2 v0.37.0
-	golang.org/x/sys v0.48.0
+	golang.org/x/crypto v0.56.0
+	golang.org/x/exp v0.0.0-20260824195058-e88cd73687aa
+	golang.org/x/net v0.58.0
+	golang.org/x/sys v0.47.0
 	gopkg.in/yaml.v3 v3.0.1
 	howett.net/plist v1.0.1
 )
@@ -33,7 +31,7 @@ require (
 	github.com/asaskevich/govalidator v0.0.0-20230301143203-a9d515a09cc2 // indirect
 	github.com/disintegration/imaging v1.6.2 // indirect
 	github.com/domodwyer/mailyak/v3 v3.6.2 // indirect
-	github.com/dustin/go-humanize v1.1.0 // indirect
+	github.com/dustin/go-humanize v1.0.1 // indirect
 	github.com/eclipse/paho.golang v0.23.0 // indirect
 	github.com/fatih/color v1.19.0 // indirect
 	github.com/fsnotify/fsnotify v1.10.1 // indirect
@@ -43,7 +41,6 @@ require (
 	github.com/go-sql-driver/mysql v1.9.1 // indirect
 	github.com/godbus/dbus/v5 v5.2.2 // indirect
 	github.com/golang-jwt/jwt/v5 v5.3.1 // indirect
-	github.com/google/uuid v1.6.0 // indirect
 	github.com/gorilla/websocket v1.5.3 // indirect
 	github.com/inconshreveable/mousetrap v1.1.0 // indirect
 	github.com/klauspost/compress v1.20.0 // indirect
@@ -52,23 +49,18 @@ require (
 	github.com/mattn/go-isatty v0.0.24 // indirect
 	github.com/ncruces/go-strftime v1.0.0 // indirect
 	github.com/pocketbase/ozzo-validation/v4 v4.3.0 // indirect
-	github.com/power-devops/perfstat v0.0.0-20260916203055-22a1a467d9f0 // indirect
+	github.com/power-devops/perfstat v0.0.0-20260805114148-88456608a4f6 // indirect
 	github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec // indirect
 	github.com/tklauser/go-sysconf v0.4.0 // indirect
 	github.com/tklauser/numcpus v0.12.0 // indirect
 	github.com/x448/float16 v0.8.4 // indirect
 	github.com/yusufpapurcu/wmi v1.2.4 // indirect
 	go.yaml.in/yaml/v3 v3.0.5 // indirect
-	golang.org/x/image v0.46.0 // indirect
-	golang.org/x/mod v0.41.0 // indirect
-	golang.org/x/sync v0.23.0 // indirect
-	golang.org/x/term v0.46.0 // indirect
-	golang.org/x/text v0.42.0 // indirect
-	golang.org/x/tools v0.50.0 // indirect
-	mellium.im/reader v0.1.0 // indirect
-	mellium.im/sasl v0.3.2 // indirect
-	mellium.im/xmlstream v0.15.4 // indirect
-	mellium.im/xmpp v0.23.0 // indirect
+	golang.org/x/image v0.45.0 // indirect
+	golang.org/x/oauth2 v0.36.0 // indirect
+	golang.org/x/sync v0.22.0 // indirect
+	golang.org/x/term v0.45.0 // indirect
+	golang.org/x/text v0.41.0 // indirect
 	modernc.org/libc v1.74.4 // indirect
 	modernc.org/mathutil v1.7.1 // indirect
 	modernc.org/memory v1.12.1 // indirect

go.sum

@@ -15,12 +15,10 @@ github.com/cpuguy83/go-md2man/v2 v2.0.6/go.mod h1:oOW0eioCTA6cOiMLiUPZOpcVxMig6N
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/disintegration/imaging v1.6.2 h1:w1LecBlG2Lnp8B3jk5zSuNqd7b4DXhcjwek1ei82L+c=
 github.com/disintegration/imaging v1.6.2/go.mod h1:44/5580QXChDfwIclfc/PCwrr44amcmDAg8hxG0Ewe4=
-github.com/distribution/reference v0.6.0 h1:0IXCQ5g4/QMHHkarYzh5l+u8T3t73zM5QvfrDyIgxBk=
-github.com/distribution/reference v0.6.0/go.mod h1:BbU0aIcezP1/5jX/8MP0YiH4SdvB5Y4f/wlDRiLyi3E=
 github.com/domodwyer/mailyak/v3 v3.6.2 h1:x3tGMsyFhTCaxp6ycgR0FE/bu5QiNp+hetUuCOBXMn8=
 github.com/domodwyer/mailyak/v3 v3.6.2/go.mod h1:lOm/u9CyCVWHeaAmHIdF4RiKVxKUT/H5XX10lIKAL6c=
-github.com/dustin/go-humanize v1.1.0 h1:dbKTrvD0klcbBV/h4AWJdMuZogJACoMlvWIWZ5b2xWg=
-github.com/dustin/go-humanize v1.1.0/go.mod h1:hc1CvRkJMsgxqjmjMQF3QNRAZBwY8AXBAzKYoSX9sFI=
+github.com/dustin/go-humanize v1.0.1 h1:GzkhY7T5VNhEkwH0PVJgjz+fX1rhBrR7pRT3mDkpeCY=
+github.com/dustin/go-humanize v1.0.1/go.mod h1:Mu1zIs6XwVuF/gI1OepvI0qD18qycQx+mFykh5fBlto=
 github.com/ebitengine/purego v0.11.0 h1:jhp/D+Nyv7UUW8HAcmcjt2N2rYrYi9m3SL21k0Ua/NI=
 github.com/ebitengine/purego v0.11.0/go.mod h1:DCHPP08djqhNSoTfImcnHYQRZmd0qhakvrozqaEYhGQ=
 github.com/eclipse/paho.golang v0.23.0 h1:KHgl2wz6EJo7cMBmkuhpt7C576vP+kpPv7jjvSyR6Mk=
@@ -31,8 +29,8 @@ github.com/frankban/quicktest v1.14.6 h1:7Xjx+VpznH+oBnejlPUj8oUpdxnVs4f8XU8WnHk
 github.com/frankban/quicktest v1.14.6/go.mod h1:4ptaffx2x8+WTWXmUCuVU6aPUX1/Mz7zb5vbUoiM6w0=
 github.com/fsnotify/fsnotify v1.10.1 h1:b0/UzAf9yR5rhf3RPm9gf3ehBPpf0oZKIjtpKrx59Ho=
 github.com/fsnotify/fsnotify v1.10.1/go.mod h1:TLheqan6HD6GBK6PrDWyDPBaEV8LspOxvPSjC+bVfgo=
-github.com/fxamacker/cbor/v2 v2.9.4 h1:xwjVlxEMR3S605oUlgBjKLTTeGFciYPGYCtF/35LKGo=
-github.com/fxamacker/cbor/v2 v2.9.4/go.mod h1:vM4b+DJCtHn+zz7h3FFp/hDAI9WNWCsZj23V5ytsSxQ=
+github.com/fxamacker/cbor/v2 v2.9.3 h1:oQBnFATpNdY8gJHTndDDv5Xl4QqNaz51G5LLEPhng3Q=
+github.com/fxamacker/cbor/v2 v2.9.3/go.mod h1:vM4b+DJCtHn+zz7h3FFp/hDAI9WNWCsZj23V5ytsSxQ=
 github.com/gabriel-vasile/mimetype v1.4.15 h1:05iP/CYtZ/w455R/KZM6rZ5ieAdh99UPtd+d3YzLmaI=
 github.com/gabriel-vasile/mimetype v1.4.15/go.mod h1:azpTcoLcDZRNgFou5j+APrqQx9HqVPWa6ijYQIIVswQ=
 github.com/ganigeorgiev/fexpr v0.6.0 h1:Fza3O/QMBKEudUvxV862qe6GjxM60GJjjKytdp+VQus=
@@ -77,31 +75,29 @@ github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
 github.com/kr/text v0.2.0/go.mod h1:eLer722TekiGuMkidMxC/pM04lWEeraHUUmBw8l2grE=
 github.com/lufia/plan9stats v0.0.0-20260802145828-341c2f0c90b5 h1:eveIIGn4BGM3qknO74omf6HYr30/exH+eVUTuAgwjZ0=
 github.com/lufia/plan9stats v0.0.0-20260802145828-341c2f0c90b5/go.mod h1:autxFIvghDt3jPTLoqZ9OZ7s9qTGNAWmYCjVFWPX/zg=
-github.com/lxzan/gws v1.10.2 h1:htReTvcY89iMk1ScVtUbk6J96kIZWaafj6r/lasK/NA=
-github.com/lxzan/gws v1.10.2/go.mod h1:gXHSCPmTGryWJ4icuqy8Yho32E4YIMHH0fkDRYJRbdc=
+github.com/lxzan/gws v1.10.1 h1:1xG+tDOV0lgDeVPf0wNT74u3cn0K3LpcavRrTPTrMwQ=
+github.com/lxzan/gws v1.10.1/go.mod h1:gXHSCPmTGryWJ4icuqy8Yho32E4YIMHH0fkDRYJRbdc=
 github.com/mattn/go-colorable v0.1.15 h1:+u9SLTRGnXv73cEsnsmoZBom+dMU88B2M0aDcWy0/jY=
 github.com/mattn/go-colorable v0.1.15/go.mod h1:6LmQG8QLFO4G5z1gPvYEzlUgJ2wF+stgPZH1UqBm1s8=
 github.com/mattn/go-isatty v0.0.24 h1:tGZZoVgT/KiqK1c8ocVLeDS8BSWMRd47J3Lbz7vsReI=
 github.com/mattn/go-isatty v0.0.24/go.mod h1:nMCL3Zebbrt45jsMDgnfIwz6ydEQApk5oEI3HqDio6A=
 github.com/ncruces/go-strftime v1.0.0 h1:HMFp8mLCTPp341M/ZnA4qaf7ZlsbTc+miZjCLOFAw7w=
 github.com/ncruces/go-strftime v1.0.0/go.mod h1:Fwc5htZGVVkseilnfgOVb9mKy6w1naJmn9CehxcKcls=
-github.com/nicholas-fedor/shoutrrr v0.21.0 h1:as/mEwdaZMijCVu0FkTUEXashhvC3Y7C5g9dsXMcmQc=
-github.com/nicholas-fedor/shoutrrr v0.21.0/go.mod h1:dgg4kJv9K0tLXBH/1TXiSibNbM2hcd4SK6xb0sglyU4=
-github.com/onsi/ginkgo/v2 v2.32.2 h1:2o6vyFvR6snrJWgRVztC+OwuqqPEMI1UzYl2s2iU7Cg=
-github.com/onsi/ginkgo/v2 v2.32.2/go.mod h1:+aXOY+vzZ5mu2iI2HpTZUPmM//oQfsNFX6gU9kNcA44=
+github.com/nicholas-fedor/shoutrrr v0.20.0 h1:hMAxIYlfAeZ1FcTDgU0kUOvVXUsOirWo8IWlnzGLkac=
+github.com/nicholas-fedor/shoutrrr v0.20.0/go.mod h1:hgde37yNWCXh8+N6WemyDRMNYLOFTf326GsBx8Z7CFA=
+github.com/onsi/ginkgo/v2 v2.32.1 h1:6tlvcDm/3sE8lGJbZ4+d4mO3RLy24/tQWOFzVSQNIfw=
+github.com/onsi/ginkgo/v2 v2.32.1/go.mod h1:+aXOY+vzZ5mu2iI2HpTZUPmM//oQfsNFX6gU9kNcA44=
 github.com/onsi/gomega v1.43.0 h1:VlG/1FxqNxhSO+lq/OHBNaaqwiBK/mO8JbVkX9Y+FeU=
 github.com/onsi/gomega v1.43.0/go.mod h1:REff/hsDsodHoKlWsP2mAPhu1+5/6hVYNf9rIEBpeSg=
-github.com/opencontainers/go-digest v1.0.0 h1:apOUWs51W5PlhuyGyz9FCeeBIOUDA/6nW8Oi/yOhh5U=
-github.com/opencontainers/go-digest v1.0.0/go.mod h1:0JzlMkj0TRzQZfJkVvzbP0HBR3IKzErnv2BNG4W4MAM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/pocketbase/dbx v1.12.0 h1:/oLErM+A0b4xI0PWTGPqSDVjzix48PqI/bng2l0PzoA=
 github.com/pocketbase/dbx v1.12.0/go.mod h1:xXRCIAKTHMgUCyCKZm55pUOdvFziJjQfXaWKhu2vhMs=
 github.com/pocketbase/ozzo-validation/v4 v4.3.0 h1:uKBDVma7bZqgR2a6AwE+k9hkuDFfiZMpBHQdZ1z3iQs=
 github.com/pocketbase/ozzo-validation/v4 v4.3.0/go.mod h1:6XNjSTw/Jb2F8LOkKO3oyzIWExbrGiYoS4uVxVwz90g=
-github.com/pocketbase/pocketbase v0.40.4 h1:0SvSUreR3NhUMCs9LchE59oEG53efZ3cKiMGyAGBN9U=
-github.com/pocketbase/pocketbase v0.40.4/go.mod h1:2mU+80FLiY1fb13WZRg8Xx/lKg4nTjgKVJMigyRH6k0=
-github.com/power-devops/perfstat v0.0.0-20260916203055-22a1a467d9f0 h1:XA01Vk/wv9YikCi1V51yRzIHPMT5of9+cMpZoZvDn/M=
-github.com/power-devops/perfstat v0.0.0-20260916203055-22a1a467d9f0/go.mod h1:OmDBASR4679mdNQnz2pUhc2G8CO2JrUAVFDRBDP/hJE=
+github.com/pocketbase/pocketbase v0.40.2 h1:7gTqvt3bmilkphyZZ1QNhX19g3BXHqT7ynDyU81RVT4=
+github.com/pocketbase/pocketbase v0.40.2/go.mod h1:jc3YuyToy+ZXM4CeO7uSCN/htgR8yv+tjSE3eJZ8eh8=
+github.com/power-devops/perfstat v0.0.0-20260805114148-88456608a4f6 h1:jL3a8soXdzuTCcRnKhOmtcsVOObdDTFf4O2B403HPRU=
+github.com/power-devops/perfstat v0.0.0-20260805114148-88456608a4f6/go.mod h1:OmDBASR4679mdNQnz2pUhc2G8CO2JrUAVFDRBDP/hJE=
 github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec h1:W09IVJc94icq4NjY3clb7Lk8O1qJ8BdBEF8z0ibU0rE=
 github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec/go.mod h1:qqbHyh8v60DhA7CoWK5oRCqLrMHRGoxYCSS9EjAz6Eo=
 github.com/rogpeppe/go-internal v1.9.0 h1:73kH8U+JUqXU8lRuOHeVHaa/SZPifC7BkcraZVejAe8=
@@ -136,37 +132,37 @@ go.yaml.in/yaml/v3 v3.0.4/go.mod h1:DhzuOOF2ATzADvBadXxruRBLzYTpT36CKvDb3+aBEFg=
 go.yaml.in/yaml/v3 v3.0.5 h1:N6y/pJk8buWs9NY5ERU2HSMfm+IuD/OtfdAnq6kESPw=
 go.yaml.in/yaml/v3 v3.0.5/go.mod h1:HVTZu1O7/Vkt2N+BFy8Zza+lnLsABggaTM2ZpNIGuKg=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
-golang.org/x/crypto v0.57.0 h1:3ZVCjf8Ggz7zneR/EHRVx68Ctf+2pmIMP2UFhh9cC6M=
-golang.org/x/crypto v0.57.0/go.mod h1:Fdz0i5U6CoizGwLda9DttjSk6qlZo25zYNtR+ycvuZA=
-golang.org/x/exp v0.0.0-20260908205506-85c1c2202aba h1:Ck8QetSgk912qxWLMCKxd0in+aiyBQyDSMae6e/xmpU=
-golang.org/x/exp v0.0.0-20260908205506-85c1c2202aba/go.mod h1:50RgIsmK7OwqzTTeqcSXQW8SswW0o8fRcDxmqGluJ8E=
+golang.org/x/crypto v0.56.0 h1:GUh5Ii4J5jtcseSMiRqr1jXCNHoxjeV9Fmekc2oLy6Y=
+golang.org/x/crypto v0.56.0/go.mod h1:OMW5y6CY9l38uPLmxU6l6pwcXp1obtLo3e6gT7gQR2I=
+golang.org/x/exp v0.0.0-20260824195058-e88cd73687aa h1:QSyA8ishJCyT21kER9KwNt0b7BM3iRK4x9QXhjN5Fdk=
+golang.org/x/exp v0.0.0-20260824195058-e88cd73687aa/go.mod h1:zeBbvyFKDaLwa7CH/zI8KXt7gTl14SF7sO08Pl5jBCM=
 golang.org/x/image v0.0.0-20191009234506-e7c1f5e7dbb8/go.mod h1:FeLwcggjj3mMvU+oOTbSwawSJRM1uh48EjtB4UJZlP0=
-golang.org/x/image v0.46.0 h1:b1+oYj0Jbp6K5MDT4i4/eZpYlk3V8SJhhDKh6LBHAyQ=
-golang.org/x/image v0.46.0/go.mod h1:3B3W05VGVQyuXucLINLjXKrqISASfi4Xj+iCVkLMwew=
-golang.org/x/mod v0.41.0 h1:qJmnOUb4YB+FsEuM3HcWucdZASCPGhsX6uljO6pog0c=
-golang.org/x/mod v0.41.0/go.mod h1:Ek9pY8RKWXwsWvd3rQiHYtMqkjSUV+s1Rj7j4H5Ur6o=
+golang.org/x/image v0.45.0 h1:FMb1nTbH5H9vF55SriQHgFw5GnNL9Jg6L25BwXKzhB0=
+golang.org/x/image v0.45.0/go.mod h1:n62x/7RqlwXDvGsSU4u6IUTUf6KghUZ9Bt7cG/T9Fx4=
+golang.org/x/mod v0.40.0 h1:hUv+3cXcdRHz08UmSiOob7sadHig73uo5bkXxQ/tvUs=
+golang.org/x/mod v0.40.0/go.mod h1:0/weTWkPWGBikyTWAX3dkjVztMmBA5hM0DH6BElSupE=
 golang.org/x/net v0.0.0-20190603091049-60506f45cf65/go.mod h1:HSz+uSET+XFnRR8LxR5pz3Of3rY3CfYBVs4xY44aLks=
-golang.org/x/net v0.59.0 h1:5zfYln+w5XCxwrnMMJPufRgNoXEaGxl0wo5GqPXyues=
-golang.org/x/net v0.59.0/go.mod h1:2DA/G1UfVbCpQPeWTmMPGY7Cs2PkBkwu743bVX5PIVg=
-golang.org/x/oauth2 v0.37.0 h1:JUlcxA8oAtauLfiH8FX2/FkAWHAdi0QtGCGc+hofE98=
-golang.org/x/oauth2 v0.37.0/go.mod h1:IxwZNxUULJmpBFf9K/9NTMSIfZZuvuTy1gGxhigP/58=
-golang.org/x/sync v0.23.0 h1:KameEIfc1IkluZyXWLn39Wd4tURc6GbCiISGiZm2bQk=
-golang.org/x/sync v0.23.0/go.mod h1:sUUOizhqBxiL6pEWpqNLUiaJn1ShEbZ6BBqskPbjZm0=
+golang.org/x/net v0.58.0 h1:ynWG7rqYi4ccpTEuPZ2QGWHktVEM9DMCj9yzDE0Q7To=
+golang.org/x/net v0.58.0/go.mod h1:YwCddHnFlT7eLQqVprV19OnhLGtc5xOKgE0RyqgfWAU=
+golang.org/x/oauth2 v0.36.0 h1:peZ/1z27fi9hUOFCAZaHyrpWG5lwe0RJEEEeH0ThlIs=
+golang.org/x/oauth2 v0.36.0/go.mod h1:YDBUJMTkDnJS+A4BP4eZBjCqtokkg1hODuPjwiGPO7Q=
+golang.org/x/sync v0.22.0 h1:SZjpbeLmrCk4xhRSZFNZW5gFUeCeFgjekvI/+gfScek=
+golang.org/x/sync v0.22.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
 golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20190916202348-b4ddaad3f8a3/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20201204225414-ed752295db88/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.1.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.48.0 h1:bbX/i/6MgT9BVLM9RT1thmxL04yeTAhbEz4SyadbXoo=
-golang.org/x/sys v0.48.0/go.mod h1:hNLxWAXmnKAxqDtdwIYC4bM9oQPEecfsnNMuSxOs3og=
-golang.org/x/term v0.46.0 h1:3+OXuTbaKDgwk8jTi3aSLHRlmWqHEUDUtxnbFigO4YE=
-golang.org/x/term v0.46.0/go.mod h1:+K02xbkittuwc0Am4abfA3Fc+XRGXkvBXNO88NCXPoc=
+golang.org/x/sys v0.47.0 h1:o7XGOvZQCADBQQ4Y7VNq2dRWQR7JmOUW8Kxx4ZsNgWs=
+golang.org/x/sys v0.47.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
+golang.org/x/term v0.45.0 h1:NwWyBmoJCbfTHpxrWoZ9C6/VxOf7ic219I8xZZFdrf0=
+golang.org/x/term v0.45.0/go.mod h1:9aqxs0blBcrm/n0L9QW0aRVD+ktan8ssZromtqJC43w=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
-golang.org/x/text v0.42.0 h1:JbOZXgfeCPU9gacVtYliJqOhD+zhrEqK4LfdpmlUZqI=
-golang.org/x/text v0.42.0/go.mod h1:ojzP1Z+2QtioaF8DTtO8K5q7JWVVYwZKenzujK0Zd0E=
+golang.org/x/text v0.41.0 h1:vz/seA0lnX87Othu2f/0L24RcgrXD9/YFTSuGjj3rH8=
+golang.org/x/text v0.41.0/go.mod h1:jvf1O8ajNzZqhSrQBPbutR/EB83Cc0CFrezNQIwbb5M=
 golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
-golang.org/x/tools v0.50.0 h1:c2ifzfcuY7L90lZ2aKd8S4K2NpASF08SZx9ZuJkHmSU=
-golang.org/x/tools v0.50.0/go.mod h1:7ulVMw3831Mwi5EZD6RomGyffr4VFjuNYXf2BbCEAV0=
+golang.org/x/tools v0.49.0 h1:3NI7VXzL9+1WZD52Dx2ttoPwD5DWrFGpl9mFZDlmisI=
+golang.org/x/tools v0.49.0/go.mod h1:SJNXV9DBKT0UbdttsQjbfJlAE/q+y36++zo3uL3N0Oo=
 google.golang.org/appengine v1.6.5/go.mod h1:8WjMMxjGQR8xUklV/ARdw2HLXBOI7O7uCIDZVag1xfc=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
@@ -176,14 +172,6 @@ gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
 gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
 howett.net/plist v1.0.1 h1:37GdZ8tP09Q35o9ych3ehygcsL+HqKSwzctveSlarvM=
 howett.net/plist v1.0.1/go.mod h1:lqaXoTrLY4hg8tnEzNru53gicrbv7rrk+2xJA/7hw9g=
-mellium.im/reader v0.1.0 h1:UUEMev16gdvaxxZC7fC08j7IzuDKh310nB6BlwnxTww=
-mellium.im/reader v0.1.0/go.mod h1:F+X5HXpkIfJ9EE1zHQG9lM/hO946iYAmU7xjg5dsQHI=
-mellium.im/sasl v0.3.2 h1:PT6Xp7ccn9XaXAnJ03FcEjmAn7kK1x7aoXV6F+Vmrl0=
-mellium.im/sasl v0.3.2/go.mod h1:NKXDi1zkr+BlMHLQjY3ofYuU4KSPFxknb8mfEu6SveY=
-mellium.im/xmlstream v0.15.4 h1:gLKxcWl4rLMUpKgtzrTBvr4OexPeO/edYus+uK3F6ZI=
-mellium.im/xmlstream v0.15.4/go.mod h1:yXaCW2++fmVO4L9piKVkyLDqnCmictVYF7FDQW8prb4=
-mellium.im/xmpp v0.23.0 h1:rvKvOvMdIURCLaAWEJN8J0QpO3AJYCDKjxLsqtTPSjY=
-mellium.im/xmpp v0.23.0/go.mod h1:GHDKlKKQe0LNmD9YqExyxnFEEBiz84KGqnfiA2VNzb8=
 modernc.org/cc/v4 v4.29.1 h1:MKgdCV3WykTSPqpVrnxdEDS0HEd2FHpKZDzxzU5LyeI=
 modernc.org/cc/v4 v4.29.1/go.mod h1:OnovgIhbbMXMu1aISnJ0wvVD1KnW+cAUJkIrAWh+kVI=
 modernc.org/ccgo/v4 v4.34.6 h1:sBgfIwyN0TQ9C5hwIeuqyeAKyMWnbvj2fvpF4L11uzU=

internal/alerts/alerts.go

@@ -20,11 +20,10 @@ type hubLike interface {
 }
 
 type AlertManager struct {
-	hub             hubLike
-	stopOnce        sync.Once
-	pendingAlerts   sync.Map
-	alertsCache     *AlertsCache
-	networkMonitors *networkMonitorCache
+	hub           hubLike
+	stopOnce      sync.Once
+	pendingAlerts sync.Map
+	alertsCache   *AlertsCache
 }
 
 type AlertMessageData struct {
@@ -67,7 +66,6 @@ type SystemAlertGPUData struct {
 }
 
 type SystemAlertZfsPool struct {
-	Raw   bool    `json:"raw,omitempty"`
 	Total float64 `json:"d"`
 	Used  float64 `json:"du"`
 }
@@ -108,9 +106,8 @@ var supportsTitle = map[string]struct{}{
 // NewAlertManager creates a new AlertManager instance.
 func NewAlertManager(app hubLike) *AlertManager {
 	am := &AlertManager{
-		hub:             app,
-		alertsCache:     NewAlertsCache(app),
-		networkMonitors: newNetworkMonitorCache(app),
+		hub:         app,
+		alertsCache: NewAlertsCache(app),
 	}
 	am.bindEvents()
 	return am
@@ -118,7 +115,6 @@ func NewAlertManager(app hubLike) *AlertManager {
 
 // Bind events to the alerts collection lifecycle
 func (am *AlertManager) bindEvents() {
-	am.bindNetworkMonitorAlertEvents()
 	am.hub.OnRecordAfterUpdateSuccess("alerts").BindFunc(updateHistoryOnAlertUpdate)
 	am.hub.OnRecordAfterDeleteSuccess("alerts").BindFunc(resolveHistoryOnAlertDelete)
 	am.hub.OnRecordAfterUpdateSuccess("smart_devices").BindFunc(am.handleSmartDeviceAlert)

internal/alerts/alerts_api.go

@@ -29,13 +29,6 @@ func UpsertUserAlerts(e *core.RequestEvent) error {
 		return e.BadRequestError("Bad data", err)
 	}
 
-	if reqData.Name == alertNameNetworkMonitorLoss {
-		if reqData.Value < 0 || reqData.Value >= 100 {
-			return e.BadRequestError("Monitor loss threshold must be at least 0 and below 100", nil)
-		}
-		reqData.Min = 0
-	}
-
 	alertsCollection, err := e.App.FindCachedCollectionByNameOrId("alerts")
 	if err != nil {
 		return err

internal/alerts/alerts_api_test.go

@@ -11,7 +11,6 @@ import (
 	"strings"
 	"sync/atomic"
 	"testing"
-	"testing/synctest"
 
 	beszelTests "github.com/henrygd/beszel/internal/tests"
 	pbTests "github.com/pocketbase/pocketbase/tests"
@@ -534,20 +533,6 @@ func TestSendTestNotification(t *testing.T) {
 
 	for _, url := range []string{localURL, "smtp://user:pass@127.0.0.1/?fromAddress=sender@example.com&toAddresses=recipient@example.com", "mqtt://127.0.0.1/topic"} {
 		scenarios = append(scenarios, beszelTests.ApiScenario{
-			BeforeTestFunc: func(tb testing.TB, _ *pbTests.TestApp, e *core.ServeEvent) {
-				if !strings.HasPrefix(url, "mqtt://") {
-					return
-				}
-				// Keep the real MQTT rejection path, but advance its library's
-				// fixed timeout using virtual time instead of waiting 10 seconds.
-				e.Router.BindFunc(func(re *core.RequestEvent) error {
-					var err error
-					synctest.Test(tb.(*testing.T), func(t *testing.T) {
-						err = re.Next()
-					})
-					return err
-				})
-			},
 			Name:            "readonly cannot send to " + url,
 			Method:          http.MethodPost,
 			URL:             "/api/beszel/test-notification",

internal/alerts/alerts_cache.go

@@ -1,7 +1,6 @@
 package alerts
 
 import (
-	"sync"
 	"time"
 
 	"github.com/pocketbase/dbx"
@@ -19,9 +18,6 @@ type CachedAlertData struct {
 	Triggered    bool
 	Min          uint8
 	PendingSince time.Time
-	// Immutable after publication; decoded only when the alert record changes.
-	MonitorStates      map[string]string
-	MonitorStatesValid bool
 	// Created   types.DateTime
 }
 
@@ -34,18 +30,11 @@ func (a *CachedAlertData) PopulateFromRecord(record *core.Record) {
 	a.Triggered = record.GetBool("triggered")
 	a.Min = uint8(record.GetInt("min"))
 	a.PendingSince = record.GetDateTime("pending_since").Time()
-	if a.Name == alertNameNetworkMonitorLoss {
-		var state networkMonitorAlertState
-		a.MonitorStatesValid = record.UnmarshalJSONField("state", &state) == nil
-		a.MonitorStates = state.Monitors
-	}
 	// a.Created = record.GetDateTime("created")
 }
 
 // AlertsCache provides an in-memory cache for system alerts.
 type AlertsCache struct {
-	// Serialize lazy loads with updates so a late load cannot replace newer state.
-	loadMu    sync.Mutex
 	app       core.App
 	store     *store.Store[string, *store.Store[string, CachedAlertData]]
 	populated bool
@@ -80,8 +69,6 @@ func (c *AlertsCache) bindEvents() *AlertsCache {
 
 // PopulateFromDB clears current entries and loads all alerts from the database into the cache.
 func (c *AlertsCache) PopulateFromDB(force bool) error {
-	c.loadMu.Lock()
-	defer c.loadMu.Unlock()
 	if !force && c.populated {
 		return nil
 	}
@@ -91,7 +78,7 @@ func (c *AlertsCache) PopulateFromDB(force bool) error {
 	}
 	c.store.RemoveAll()
 	for _, record := range records {
-		c.update(record)
+		c.Update(record)
 	}
 	c.populated = true
 	return nil
@@ -99,12 +86,6 @@ func (c *AlertsCache) PopulateFromDB(force bool) error {
 
 // Update adds or updates an alert record in the cache.
 func (c *AlertsCache) Update(record *core.Record) {
-	c.loadMu.Lock()
-	defer c.loadMu.Unlock()
-	c.update(record)
-}
-
-func (c *AlertsCache) update(record *core.Record) {
 	systemID := record.GetString("system")
 	if systemID == "" {
 		return
@@ -121,8 +102,6 @@ func (c *AlertsCache) update(record *core.Record) {
 
 // Delete removes an alert record from the cache.
 func (c *AlertsCache) Delete(record *core.Record) {
-	c.loadMu.Lock()
-	defer c.loadMu.Unlock()
 	systemID := record.GetString("system")
 	if systemID == "" {
 		return
@@ -136,23 +115,18 @@ func (c *AlertsCache) Delete(record *core.Record) {
 func (c *AlertsCache) GetSystemAlerts(systemID string) []CachedAlertData {
 	systemStore, ok := c.store.GetOk(systemID)
 	if !ok {
-		c.loadMu.Lock()
-		defer c.loadMu.Unlock()
-		systemStore, ok = c.store.GetOk(systemID)
-		if !ok {
-			// Populate cache for this system
-			records, err := c.app.FindAllRecords("alerts", dbx.NewExp("system={:system}", dbx.Params{"system": systemID}))
-			if err != nil {
-				return nil
-			}
-			systemStore = store.New(map[string]CachedAlertData{})
-			for _, record := range records {
-				var ca CachedAlertData
-				ca.PopulateFromRecord(record)
-				systemStore.Set(record.Id, ca)
-			}
-			c.store.Set(systemID, systemStore)
+		// Populate cache for this system
+		records, err := c.app.FindAllRecords("alerts", dbx.NewExp("system={:system}", dbx.Params{"system": systemID}))
+		if err != nil {
+			return nil
 		}
+		systemStore = store.New(map[string]CachedAlertData{})
+		for _, record := range records {
+			var ca CachedAlertData
+			ca.PopulateFromRecord(record)
+			systemStore.Set(record.Id, ca)
+		}
+		c.store.Set(systemID, systemStore)
 	}
 	all := systemStore.GetAll()
 	alerts := make([]CachedAlertData, 0, len(all))

internal/alerts/alerts_history.go

@@ -9,12 +9,6 @@ import (
 
 // On triggered alert record delete, set matching alert history record to resolved
 func resolveHistoryOnAlertDelete(e *core.RecordEvent) error {
-	if e.Record.GetString("name") == alertNameNetworkMonitorLoss {
-		if err := resolveNetworkMonitorHistory(e.App, e.Record.Id); err != nil {
-			return err
-		}
-		return e.Next()
-	}
 	if !e.Record.GetBool("triggered") {
 		return e.Next()
 	}
@@ -24,10 +18,6 @@ func resolveHistoryOnAlertDelete(e *core.RecordEvent) error {
 
 // On alert record update, update alert history record
 func updateHistoryOnAlertUpdate(e *core.RecordEvent) error {
-	// Network monitor incidents have separate history entries per monitor.
-	if e.Record.GetString("name") == alertNameNetworkMonitorLoss {
-		return e.Next()
-	}
 	original := e.Record.Original()
 	new := e.Record
 

internal/alerts/alerts_network_monitors.go

@@ -1,269 +0,0 @@
-package alerts
-
-import (
-	"database/sql"
-	"errors"
-	"fmt"
-	"math"
-	"net"
-	"strconv"
-	"time"
-
-	"github.com/henrygd/beszel/internal/entities/monitor"
-	"github.com/pocketbase/dbx"
-	"github.com/pocketbase/pocketbase/core"
-)
-
-const alertNameNetworkMonitorLoss = "NetworkMonitorLoss"
-
-// networkMonitorAlertState is this alert type's persisted runtime state.
-// Monitor IDs map to their open history entries independently of history retention.
-type networkMonitorAlertState struct {
-	Monitors map[string]string `json:"monitors"`
-}
-
-func (am *AlertManager) bindNetworkMonitorAlertEvents() {
-	// Hidden fields are still writable through the record API unless protected.
-	protectState := func(e *core.RecordRequestEvent) error {
-		e.Record.Set("state", e.Record.Original().Get("state"))
-		oldName, newName := e.Record.Original().GetString("name"), e.Record.GetString("name")
-		if oldName != "" && (oldName == alertNameNetworkMonitorLoss || newName == alertNameNetworkMonitorLoss) &&
-			(oldName != newName || e.Record.GetString("system") != e.Record.Original().GetString("system")) {
-			return e.BadRequestError("Delete and recreate the alert to change its type or system", nil)
-		}
-		if e.Record.GetString("name") == alertNameNetworkMonitorLoss {
-			if !e.HasSuperuserAuth() && (e.Auth == nil || !userHasSystem(e.App, e.Auth.Id, e.Record.GetString("system"))) {
-				return e.ForbiddenError("You do not have access to this system", nil)
-			}
-			e.Record.Set("triggered", e.Record.Original().GetBool("triggered"))
-			value := e.Record.GetFloat("value")
-			if math.IsNaN(value) || math.IsInf(value, 0) || value < 0 || value >= 100 {
-				return e.BadRequestError("Monitor loss threshold must be at least 0 and below 100", nil)
-			}
-			e.Record.Set("min", 0)
-		}
-		return e.Next()
-	}
-	am.hub.OnRecordCreateRequest("alerts").BindFunc(protectState)
-	am.hub.OnRecordUpdateRequest("alerts").BindFunc(protectState)
-	cleanup := func(e *core.RecordEvent) error {
-		if err := e.Next(); err != nil {
-			return err
-		}
-		return am.evaluateNetworkMonitorAlerts(e.App, e.Record.GetString("system"), nil)
-	}
-	am.hub.OnRecordAfterDeleteSuccess("network_monitors").BindFunc(cleanup)
-	am.hub.OnRecordAfterUpdateSuccess("network_monitors").BindFunc(func(e *core.RecordEvent) error {
-		if e.Record.GetBool("enabled") || !e.Record.Original().GetBool("enabled") {
-			return e.Next()
-		}
-		return cleanup(e)
-	})
-}
-
-// HandleNetworkMonitorAlerts runs after the full monitoring transaction commits,
-// using its exact payload (dashboard requests can replace the cached payload).
-// Omitted results and disconnected systems never imply recovery.
-func (am *AlertManager) HandleNetworkMonitorAlerts(systemRecord *core.Record, results map[string]monitor.Result) error {
-	if systemRecord.GetString("status") != "up" {
-		return nil
-	}
-	alerts := am.alertsCache.GetAlertsByName(systemRecord.Id, alertNameNetworkMonitorLoss)
-	if len(alerts) == 0 {
-		return nil
-	}
-	monitors, err := am.networkMonitors.get(systemRecord.Id)
-	if err != nil {
-		return err
-	}
-	if !networkMonitorTransitionPending(alerts, monitors, results, time.Now()) {
-		return nil
-	}
-	// The cache only predicts a transition. Reload and recheck under the DB
-	// transaction before persisting, including current system/monitor status.
-	return am.evaluateNetworkMonitorAlerts(am.hub, systemRecord.Id, results)
-}
-
-// networkMonitorTransitionPending does no IO and never mutates cached maps.
-func networkMonitorTransitionPending(alerts []CachedAlertData, monitors map[string]int, results map[string]monitor.Result, now time.Time) bool {
-	for _, alert := range alerts {
-		if !alert.MonitorStatesValid || alert.Triggered != (len(alert.MonitorStates) > 0) {
-			return true
-		}
-		for id := range alert.MonitorStates {
-			if _, enabled := monitors[id]; !enabled {
-				return true
-			}
-		}
-		for id, result := range results {
-			interval, enabled := monitors[id]
-			if !enabled || !monitorResultReady(result, interval, now) {
-				continue
-			}
-			_, active := alert.MonitorStates[id]
-			if (result.PacketLoss1h > alert.Value) != active {
-				return true
-			}
-		}
-	}
-	return false
-}
-
-func (am *AlertManager) evaluateNetworkMonitorAlerts(app core.App, systemID string, results map[string]monitor.Result) error {
-	var messages []AlertMessageData
-	err := app.RunInTransaction(func(tx core.App) error {
-		// Read configuration inside the transaction so concurrent threshold changes,
-		// disabling, and evaluations cannot overwrite each other's incident state.
-		alerts, err := tx.FindAllRecords("alerts", dbx.HashExp{"system": systemID, "name": alertNameNetworkMonitorLoss})
-		if err != nil || len(alerts) == 0 {
-			return err
-		}
-		system, err := tx.FindRecordById("systems", systemID)
-		if errors.Is(err, sql.ErrNoRows) {
-			// System deletion cascades to its alerts.
-			return nil
-		}
-		if err != nil {
-			return err
-		}
-		monitors, err := tx.FindAllRecords("network_monitors", dbx.HashExp{"system": systemID, "enabled": true})
-		if err != nil {
-			return err
-		}
-		enabled := make(map[string]*core.Record, len(monitors))
-		for _, m := range monitors {
-			enabled[m.Id] = m
-		}
-		now := time.Now()
-		for _, alert := range alerts {
-			var state networkMonitorAlertState
-			if err := alert.UnmarshalJSONField("state", &state); err != nil {
-				return err
-			}
-			states := state.Monitors
-			if states == nil {
-				states = map[string]string{}
-			}
-			changed := false
-			// Removing or disabling a monitor closes its incident silently.
-			for id, historyID := range states {
-				if _, ok := enabled[id]; !ok {
-					if err := resolveMonitorIncident(tx, historyID, now); err != nil {
-						return err
-					}
-					delete(states, id)
-					changed = true
-				}
-			}
-			if system.GetString("status") == "up" {
-				for _, m := range monitors {
-					result, ok := results[m.Id]
-					if !ok || !monitorResultReady(result, m.GetInt("interval"), now) {
-						continue
-					}
-					historyID, active := states[m.Id]
-					triggered := result.PacketLoss1h > alert.GetFloat("value")
-					if triggered == active {
-						continue
-					}
-					label := m.GetString("target")
-					if m.GetString("protocol") == "tcp" {
-						label = net.JoinHostPort(label, strconv.Itoa(m.GetInt("port")))
-					}
-					if triggered {
-						collection, err := tx.FindCachedCollectionByNameOrId("alerts_history")
-						if err != nil {
-							return err
-						}
-						history := core.NewRecord(collection)
-						history.Load(map[string]any{
-							"alert_id": alert.Id, "user": alert.GetString("user"), "system": systemID,
-							"name": alertNameNetworkMonitorLoss, "monitor_name": label, "value": result.PacketLoss1h,
-						})
-						if err := tx.Save(history); err != nil {
-							return err
-						}
-						states[m.Id] = history.Id
-					} else {
-						if err := resolveMonitorIncident(tx, historyID, now); err != nil {
-							return err
-						}
-						delete(states, m.Id)
-					}
-					changed = true
-					state, comparison := "loss", "exceeds"
-					if !triggered {
-						state, comparison = "recovered", "is at or below"
-					}
-					messages = append(messages, AlertMessageData{
-						UserID: alert.GetString("user"), SystemID: systemID,
-						Title:   fmt.Sprintf("Network monitor %s on %s: %s", state, system.GetString("name"), label),
-						Message: fmt.Sprintf("%s on %s: loss over the past hour is %.2f%%, which %s the %.2f%% threshold.", label, system.GetString("name"), result.PacketLoss1h, comparison, alert.GetFloat("value")),
-						Link:    am.hub.MakeLink("system", systemID), LinkText: "View " + system.GetString("name"),
-					})
-				}
-			}
-			if changed || alert.GetBool("triggered") != (len(states) > 0) {
-				alert.Set("state", networkMonitorAlertState{Monitors: states})
-				alert.Set("triggered", len(states) > 0)
-				if err := tx.Save(alert); err != nil {
-					return err
-				}
-			}
-		}
-		return nil
-	})
-	if err != nil {
-		return err
-	}
-	// Match other alert types: persist transitions before delivery, and respect
-	// the user's existing notification destinations and quiet hours.
-	for _, message := range messages {
-		if err := am.SendAlert(message); err != nil {
-			app.Logger().Error("Failed to send network monitor alert", "err", err)
-		}
-	}
-	return nil
-}
-
-func monitorResultReady(result monitor.Result, interval int, now time.Time) bool {
-	// Three completed attempts provide a short warm-up, including after an agent
-	// restart.
-	if result.SampleCount < 3 || result.LastProbeAt <= 0 || math.IsNaN(result.PacketLoss1h) || math.IsInf(result.PacketLoss1h, 0) || result.PacketLoss1h < 0 || result.PacketLoss1h > 100 {
-		return false
-	}
-	// Never interpret an empty one-hour window as zero loss.
-	maxAge := min(time.Hour, max(3*time.Duration(interval)*time.Second, 3*time.Minute))
-	age := now.Sub(time.UnixMilli(result.LastProbeAt))
-	return age >= -time.Minute && age <= maxAge
-}
-
-func resolveMonitorIncident(app core.App, id string, now time.Time) error {
-	record, err := app.FindRecordById("alerts_history", id)
-	if errors.Is(err, sql.ErrNoRows) {
-		// History can be purged independently.
-		return nil
-	}
-	if err != nil {
-		return err
-	}
-	if !record.GetDateTime("resolved").IsZero() {
-		return nil
-	}
-	record.Set("resolved", now.UTC())
-	return app.Save(record)
-}
-
-func resolveNetworkMonitorHistory(app core.App, alertID string) error {
-	records, err := app.FindAllRecords("alerts_history", dbx.HashExp{"alert_id": alertID, "resolved": ""})
-	if err != nil {
-		return err
-	}
-	for _, record := range records {
-		record.Set("resolved", time.Now().UTC())
-		if err := app.Save(record); err != nil {
-			return err
-		}
-	}
-	return nil
-}

internal/alerts/alerts_network_monitors_test.go

@@ -1,513 +0,0 @@
-//go:build testing
-
-package alerts_test
-
-import (
-	"sync"
-	"sync/atomic"
-	"testing"
-	"time"
-
-	"github.com/henrygd/beszel/internal/alerts"
-	"github.com/henrygd/beszel/internal/entities/monitor"
-	beszelTests "github.com/henrygd/beszel/internal/tests"
-	"github.com/pocketbase/dbx"
-	"github.com/pocketbase/pocketbase/core"
-	pbTests "github.com/pocketbase/pocketbase/tests"
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func networkAlertSetup(t *testing.T) (*beszelTests.TestHub, *core.Record, *core.Record, []*core.Record) {
-	t.Helper()
-	hub, system, alert := systemdTestSetup(t, false)
-	t.Cleanup(hub.Cleanup)
-	alert.Set("name", "NetworkMonitorLoss")
-	alert.Set("value", 5)
-	require.NoError(t, hub.Save(alert))
-	var monitors []*core.Record
-	for _, name := range []string{"gateway", "website"} {
-		record, err := beszelTests.CreateRecord(hub, "network_monitors", map[string]any{
-			"system": system.Id, "target": name + ".example.com", "protocol": "icmp", "interval": 60, "enabled": true,
-		})
-		require.NoError(t, err)
-		monitors = append(monitors, record)
-	}
-	// Avoid starting a system update worker in tests.
-	_, err := hub.DB().Update("systems", dbx.Params{"status": "up"}, dbx.HashExp{"id": system.Id}).Execute()
-	require.NoError(t, err)
-	system.Set("status", "up")
-	return hub, system, alert, monitors
-}
-
-func monitorResult(loss float64) monitor.Result {
-	return monitor.Result{LastProbeAt: time.Now().UnixMilli(), SampleCount: 60, PacketLoss1h: loss}
-}
-
-func TestNetworkMonitorAlertIndependentIncidents(t *testing.T) {
-	hub, system, alert, monitors := networkAlertSetup(t)
-	am := alerts.NewTestAlertManagerWithoutWorker(hub)
-	count := hub.TestMailer.TotalSend()
-	results := map[string]monitor.Result{monitors[0].Id: monitorResult(10), monitors[1].Id: monitorResult(0)}
-	check := func(active bool, open, sent int) {
-		t.Helper()
-		record, err := hub.FindRecordById("alerts", alert.Id)
-		require.NoError(t, err)
-		assert.Equal(t, active, record.GetBool("triggered"))
-		total, err := hub.CountRecords("alerts_history", dbx.HashExp{"alert_id": alert.Id, "resolved": ""})
-		require.NoError(t, err)
-		assert.EqualValues(t, open, total)
-		assert.Equal(t, count+sent, hub.TestMailer.TotalSend())
-	}
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	check(true, 1, 1)
-	message := hub.TestMailer.Messages()[count]
-	assert.Contains(t, message.Text, "gateway.example.com")
-	assert.Contains(t, message.Text, "10.00%")
-	assert.Contains(t, message.Text, "5.00%")
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	check(true, 1, 1)
-	// Persisted monitor state prevents duplicate notifications after a hub restart.
-	am = alerts.NewTestAlertManagerWithoutWorker(hub)
-	results[monitors[1].Id] = monitorResult(20)
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	check(true, 2, 2)
-	results[monitors[0].Id] = monitorResult(5) // Equality is a recovery.
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	check(true, 1, 3)
-	results[monitors[1].Id] = monitorResult(0)
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	check(false, 0, 4)
-	histories, err := hub.FindAllRecords("alerts_history", dbx.HashExp{"alert_id": alert.Id})
-	require.NoError(t, err)
-	require.Len(t, histories, 2)
-	for _, history := range histories {
-		assert.NotEmpty(t, history.GetString("monitor_name"))
-	}
-}
-
-func TestNetworkMonitorAlertTargetLabel(t *testing.T) {
-	for _, tc := range []struct {
-		protocol, target, label string
-		port                    int
-	}{
-		{"icmp", "gateway.example.com", "gateway.example.com", 0},
-		{"http", "https://example.com/health", "https://example.com/health", 0},
-		{"tcp", "example.com", "example.com:8443", 8443},
-		{"tcp", "2001:db8::1", "[2001:db8::1]:443", 443},
-	} {
-		t.Run(tc.label, func(t *testing.T) {
-			hub, system, alert, monitors := networkAlertSetup(t)
-			m := monitors[0]
-			m.Set("protocol", tc.protocol)
-			m.Set("target", tc.target)
-			m.Set("port", tc.port)
-			require.NoError(t, hub.Save(m))
-			am := alerts.NewTestAlertManagerWithoutWorker(hub)
-			require.NoError(t, am.HandleNetworkMonitorAlerts(system, map[string]monitor.Result{m.Id: monitorResult(10)}))
-			histories, err := hub.FindAllRecords("alerts_history", dbx.HashExp{"alert_id": alert.Id})
-			require.NoError(t, err)
-			require.Len(t, histories, 1)
-			assert.Equal(t, tc.label, histories[0].GetString("monitor_name"))
-			assert.Contains(t, hub.TestMailer.Messages()[hub.TestMailer.TotalSend()-1].Text, tc.label)
-		})
-	}
-}
-
-func TestNetworkMonitorAlertIgnoresUnknownResults(t *testing.T) {
-	for _, scenario := range []string{"missing", "stale", "warmup", "no probes", "down", "paused", "future", "expired hourly window"} {
-		t.Run(scenario, func(t *testing.T) {
-			hub, system, alert, monitors := networkAlertSetup(t)
-			am := alerts.NewTestAlertManagerWithoutWorker(hub)
-			apply := func(loss float64) {
-				result := monitorResult(loss)
-				results := map[string]monitor.Result{monitors[0].Id: result}
-				switch scenario {
-				case "missing":
-					results = nil
-				case "stale":
-					result.LastProbeAt = time.Now().Add(-10 * time.Minute).UnixMilli()
-					results[monitors[0].Id] = result
-				case "expired hourly window":
-					monitors[0].Set("interval", 3600)
-					require.NoError(t, hub.Save(monitors[0]))
-					result.LastProbeAt = time.Now().Add(-2 * time.Hour).UnixMilli()
-					results[monitors[0].Id] = result
-				case "future":
-					result.LastProbeAt = time.Now().Add(time.Hour).UnixMilli()
-					results[monitors[0].Id] = result
-				case "warmup":
-					result.SampleCount = 2
-					results[monitors[0].Id] = result
-				case "no probes":
-					result.SampleCount = 0
-					results[monitors[0].Id] = result
-				case "down":
-					_, err := hub.DB().Update("systems", dbx.Params{"status": scenario}, dbx.HashExp{"id": system.Id}).Execute()
-					require.NoError(t, err)
-				case "paused":
-					record, err := hub.FindRecordById("systems", system.Id)
-					require.NoError(t, err)
-					record.Set("status", "paused")
-					require.NoError(t, hub.Save(record))
-				}
-				require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-			}
-			count := hub.TestMailer.TotalSend()
-			apply(100)
-			assert.Equal(t, count, hub.TestMailer.TotalSend())
-			_, err := hub.DB().Update("systems", dbx.Params{"status": "up"}, dbx.HashExp{"id": system.Id}).Execute()
-			require.NoError(t, err)
-			require.NoError(t, am.HandleNetworkMonitorAlerts(system, map[string]monitor.Result{monitors[0].Id: monitorResult(10)}))
-			apply(0)
-			assert.Equal(t, count+1, hub.TestMailer.TotalSend(), "unknown data must not recover an incident")
-			record, err := hub.FindRecordById("alerts", alert.Id)
-			require.NoError(t, err)
-			assert.True(t, record.GetBool("triggered"))
-		})
-	}
-}
-
-func TestNetworkMonitorAlertCleanup(t *testing.T) {
-	for _, scenario := range []string{"disable monitor", "delete monitor", "disable alert", "purge history", "delete system"} {
-		t.Run(scenario, func(t *testing.T) {
-			hub, system, alert, monitors := networkAlertSetup(t)
-			am := alerts.NewTestAlertManagerWithoutWorker(hub)
-			results := map[string]monitor.Result{monitors[0].Id: monitorResult(10), monitors[1].Id: monitorResult(20)}
-			require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-			count := hub.TestMailer.TotalSend()
-			switch scenario {
-			case "disable monitor":
-				monitors[0].Set("enabled", false)
-				require.NoError(t, hub.Save(monitors[0]))
-			case "delete monitor":
-				require.NoError(t, hub.Delete(monitors[0]))
-			case "disable alert":
-				require.NoError(t, hub.Delete(alert))
-			case "delete system":
-				require.NoError(t, hub.Delete(system))
-			case "purge history":
-				history, err := hub.FindAllRecords("alerts_history")
-				require.NoError(t, err)
-				for _, record := range history {
-					require.NoError(t, hub.Delete(record))
-				}
-			}
-			require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-			assert.Equal(t, count, hub.TestMailer.TotalSend())
-			open, err := hub.CountRecords("alerts_history", dbx.HashExp{"alert_id": alert.Id, "resolved": ""})
-			require.NoError(t, err)
-			if scenario == "disable monitor" || scenario == "delete monitor" {
-				assert.EqualValues(t, 1, open)
-				record, err := hub.FindRecordById("alerts", alert.Id)
-				require.NoError(t, err)
-				assert.True(t, record.GetBool("triggered"))
-				require.NoError(t, hub.Delete(monitors[1]))
-				record, err = hub.FindRecordById("alerts", alert.Id)
-				require.NoError(t, err)
-				assert.False(t, record.GetBool("triggered"))
-			} else {
-				assert.Zero(t, open)
-			}
-		})
-	}
-}
-
-func TestNetworkMonitorAlertPerUserThresholds(t *testing.T) {
-	hub, system, alert, monitors := networkAlertSetup(t)
-	user, err := beszelTests.CreateUser(hub, "monitor2@example.com", "password")
-	require.NoError(t, err)
-	other, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{"name": "NetworkMonitorLoss", "system": system.Id, "user": user.Id, "value": 20})
-	require.NoError(t, err)
-	am := alerts.NewTestAlertManagerWithoutWorker(hub)
-	results := map[string]monitor.Result{monitors[0].Id: monitorResult(10)}
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	other, err = hub.FindRecordById("alerts", other.Id)
-	require.NoError(t, err)
-	assert.False(t, other.GetBool("triggered"))
-	// Editing the threshold re-evaluates on the next batch, without losing state.
-	alert, err = hub.FindRecordById("alerts", alert.Id)
-	require.NoError(t, err)
-	alert.Set("value", 15)
-	require.NoError(t, hub.Save(alert))
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	alert, err = hub.FindRecordById("alerts", alert.Id)
-	require.NoError(t, err)
-	assert.False(t, alert.GetBool("triggered"))
-}
-
-func TestNetworkMonitorAlertAPI(t *testing.T) {
-	for _, tc := range []struct {
-		name                  string
-		value                 float64
-		direct, denied, patch bool
-		status                int
-	}{
-		{name: "zero threshold", value: 0, status: 200},
-		{name: "fractional threshold", value: 5.5, status: 200},
-		{name: "negative threshold", value: -1, status: 400},
-		{name: "unreachable threshold", value: 100, status: 400},
-		{name: "bulk inaccessible system", value: 5, denied: true, status: 200},
-		{name: "direct inaccessible system", value: 5, direct: true, denied: true, status: 403},
-		{name: "direct invalid threshold", value: -1, direct: true, status: 400},
-		{name: "direct private state", value: 5, direct: true, status: 200},
-		{name: "patch preserves state", value: 10, direct: true, patch: true, status: 200},
-	} {
-		t.Run(tc.name, func(t *testing.T) {
-			hub, user := beszelTests.GetHubWithUser(t)
-			defer hub.Cleanup()
-			owner := user.Id
-			if tc.denied {
-				other, err := beszelTests.CreateUser(hub, "other@example.com", "password")
-				require.NoError(t, err)
-				owner = other.Id
-			}
-			systems, err := beszelTests.CreateSystems(hub, 1, owner, "paused")
-			require.NoError(t, err)
-			token, err := user.NewAuthToken()
-			require.NoError(t, err)
-			body := map[string]any{"name": "NetworkMonitorLoss", "value": tc.value, "min": 60, "systems": []string{systems[0].Id}, "overwrite": true}
-			url, method := "/api/beszel/user-alerts", "POST"
-			if tc.direct {
-				url = "/api/collections/alerts/records"
-				body["system"], body["user"] = systems[0].Id, user.Id
-				body["state"], body["triggered"] = map[string]any{"monitors": map[string]string{"fake": "fake"}}, true
-			}
-			if tc.patch {
-				alert, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{"name": "NetworkMonitorLoss", "system": systems[0].Id, "user": user.Id, "value": 5, "triggered": true, "state": map[string]any{"monitors": map[string]string{"real": "history"}}})
-				require.NoError(t, err)
-				url += "/" + alert.Id
-				method = "PATCH"
-				body["triggered"] = false
-			}
-			content := `"success":true`
-			if tc.direct {
-				content = `"name":"NetworkMonitorLoss"`
-			}
-			if tc.status == 400 {
-				content = `"status":400`
-			}
-			if tc.status == 403 {
-				content = `"status":403`
-			}
-			scenario := beszelTests.ApiScenario{
-				Name: tc.name, Method: method, URL: url, Body: jsonReader(body),
-				Headers: map[string]string{"Authorization": token}, ExpectedStatus: tc.status, ExpectedContent: []string{content},
-				TestAppFactory: func(testing.TB) *pbTests.TestApp { return hub.TestApp },
-			}
-			scenario.Test(t)
-			records, err := hub.FindAllRecords("alerts")
-			require.NoError(t, err)
-			if tc.status != 200 || tc.denied {
-				assert.Empty(t, records)
-				return
-			}
-			require.Len(t, records, 1)
-			assert.Equal(t, tc.value, records[0].GetFloat("value"))
-			assert.Zero(t, records[0].GetInt("min"))
-			state := struct {
-				Monitors map[string]string `json:"monitors"`
-			}{}
-			require.NoError(t, records[0].UnmarshalJSONField("state", &state))
-			states := state.Monitors
-			if tc.patch {
-				assert.Equal(t, map[string]string{"real": "history"}, states)
-				assert.True(t, records[0].GetBool("triggered"))
-			} else {
-				assert.Empty(t, states)
-				assert.False(t, records[0].GetBool("triggered"))
-			}
-		})
-	}
-}
-
-type monitorCountingHub struct {
-	*beszelTests.TestHub
-	transactions      atomic.Int64
-	beforeTransaction func()
-}
-
-func (h *monitorCountingHub) RunInTransaction(fn func(core.App) error) error {
-	h.transactions.Add(1)
-	if h.beforeTransaction != nil {
-		h.beforeTransaction()
-	}
-	return h.App.RunInTransaction(fn)
-}
-
-// Count actual SQL on both DB connections, including queries through record APIs.
-func monitorSQLCounter(t *testing.T, app core.App) *atomic.Int64 {
-	t.Helper()
-	count := &atomic.Int64{}
-	for _, builder := range []dbx.Builder{app.ConcurrentDB(), app.NonconcurrentDB()} {
-		db := builder.(*dbx.DB)
-		old := db.LogFunc
-		db.LogFunc = func(string, ...any) { count.Add(1) }
-		t.Cleanup(func() { db.LogFunc = old })
-	}
-	return count
-}
-
-func TestNetworkMonitorAlertSteadyStateNoDatabaseWork(t *testing.T) {
-	hub, system, alert, monitors := networkAlertSetup(t)
-	counted := &monitorCountingHub{TestHub: hub}
-	am := alerts.NewTestAlertManagerWithoutWorker(counted)
-	sql := monitorSQLCounter(t, hub)
-	results := map[string]monitor.Result{monitors[0].Id: monitorResult(0)}
-	evaluate := func() { t.Helper(); require.NoError(t, am.HandleNetworkMonitorAlerts(system, results)) }
-	noWork := func() {
-		t.Helper()
-		sql.Store(0)
-		counted.transactions.Store(0)
-		for range 100 {
-			evaluate()
-		}
-		assert.Zero(t, sql.Load(), "steady state must not issue SQL")
-		assert.Zero(t, counted.transactions.Load(), "steady state must not open transactions")
-	}
-	// One-time lazy loads are permitted, including on hub restart.
-	evaluate()
-	assert.Positive(t, sql.Load())
-	noWork()
-	// Realtime metric saves invoke record hooks but must not invalidate config.
-	fresh, err := hub.FindRecordById("network_monitors", monitors[0].Id)
-	require.NoError(t, err)
-	monitors[0] = fresh
-	monitors[0].Set("loss1h", 0)
-	monitors[0].Set("res", 100)
-	require.NoError(t, hub.Save(monitors[0]))
-	noWork()
-	results[monitors[0].Id] = monitorResult(10)
-	evaluate()
-	assert.Positive(t, sql.Load(), "transitions must still be persisted")
-	assert.EqualValues(t, 1, counted.transactions.Load())
-	noWork()
-	// Missing and stale observations must not enter the transaction either.
-	results = nil
-	noWork()
-	results = map[string]monitor.Result{monitors[0].Id: {SampleCount: 60, LastProbeAt: time.Now().Add(-10 * time.Minute).UnixMilli()}}
-	noWork()
-	results[monitors[0].Id] = monitorResult(0)
-	evaluate()
-	noWork()
-	require.NoError(t, hub.Delete(alert))
-	noWork()
-}
-
-func TestNetworkMonitorAlertConfigCacheInvalidation(t *testing.T) {
-	hub, system, alert, monitors := networkAlertSetup(t)
-	am := alerts.NewTestAlertManagerWithoutWorker(hub)
-	results := map[string]monitor.Result{monitors[0].Id: monitorResult(0)}
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	count := hub.TestMailer.TotalSend()
-	// Widening the interval makes this observation fresh. A stale interval cache
-	// would miss the failure indefinitely, even though results keep arriving.
-	result := monitorResult(10)
-	result.LastProbeAt = time.Now().Add(-4 * time.Minute).UnixMilli()
-	results[monitors[0].Id] = result
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	assert.Equal(t, count, hub.TestMailer.TotalSend())
-	monitors[0].Set("interval", 120)
-	require.NoError(t, hub.Save(monitors[0]))
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	assert.Equal(t, count+1, hub.TestMailer.TotalSend())
-	// Disable, then re-enable the same ID: its new failure must be detected.
-	monitors[0].Set("enabled", false)
-	require.NoError(t, hub.Save(monitors[0]))
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	monitors[0].Set("enabled", true)
-	require.NoError(t, hub.Save(monitors[0]))
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	assert.Equal(t, count+2, hub.TestMailer.TotalSend())
-	// A new monitor must also become eligible without restarting the hub.
-	created, err := beszelTests.CreateRecord(hub, "network_monitors", map[string]any{
-		"system": system.Id, "name": "new", "target": "new.example.com", "protocol": "icmp", "interval": 60, "enabled": true,
-	})
-	require.NoError(t, err)
-	results[created.Id] = monitorResult(10)
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	assert.Equal(t, count+3, hub.TestMailer.TotalSend())
-	// Threshold changes refresh cached config and preserve the active incidents.
-	alert, err = hub.FindRecordById("alerts", alert.Id)
-	require.NoError(t, err)
-	alert.Set("value", 15)
-	require.NoError(t, hub.Save(alert))
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	assert.Equal(t, count+5, hub.TestMailer.TotalSend())
-}
-
-func TestNetworkMonitorAlertRevalidatesCandidate(t *testing.T) {
-	for _, change := range []string{"threshold", "disable alert", "disable monitor", "down"} {
-		t.Run(change, func(t *testing.T) {
-			hub, system, alert, monitors := networkAlertSetup(t)
-			counted := &monitorCountingHub{TestHub: hub}
-			am := alerts.NewTestAlertManagerWithoutWorker(counted)
-			results := map[string]monitor.Result{monitors[0].Id: monitorResult(0)}
-			require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-			count := hub.TestMailer.TotalSend()
-			// Change the DB after the cache predicts a transition, before its transaction.
-			counted.beforeTransaction = func() {
-				counted.beforeTransaction = nil
-				switch change {
-				case "threshold":
-					alert.Set("value", 20)
-					require.NoError(t, hub.Save(alert))
-				case "disable alert":
-					require.NoError(t, hub.Delete(alert))
-				case "disable monitor":
-					monitors[0].Set("enabled", false)
-					require.NoError(t, hub.Save(monitors[0]))
-				case "down":
-					_, err := hub.DB().Update("systems", dbx.Params{"status": "down"}, dbx.HashExp{"id": system.Id}).Execute()
-					require.NoError(t, err)
-				}
-			}
-			results[monitors[0].Id] = monitorResult(10)
-			require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-			assert.EqualValues(t, 1, counted.transactions.Load())
-			assert.Equal(t, count, hub.TestMailer.TotalSend())
-			histories, err := hub.CountRecords("alerts_history")
-			require.NoError(t, err)
-			assert.Zero(t, histories)
-		})
-	}
-}
-
-func TestNetworkMonitorAlertConcurrentEvaluations(t *testing.T) {
-	hub, system, _, monitors := networkAlertSetup(t)
-	am := alerts.NewTestAlertManagerWithoutWorker(hub)
-	results := map[string]monitor.Result{monitors[0].Id: monitorResult(10)}
-	count := hub.TestMailer.TotalSend()
-	var wg sync.WaitGroup
-	errs := make(chan error, 8)
-	for range 8 {
-		wg.Go(func() { errs <- am.HandleNetworkMonitorAlerts(system, results) })
-	}
-	wg.Wait()
-	close(errs)
-	for err := range errs {
-		require.NoError(t, err)
-	}
-	assert.Equal(t, count+1, hub.TestMailer.TotalSend())
-}
-
-func TestNetworkMonitorAlertCacheAfterRollback(t *testing.T) {
-	hub, system, _, monitors := networkAlertSetup(t)
-	am := alerts.NewTestAlertManagerWithoutWorker(hub)
-	results := map[string]monitor.Result{monitors[0].Id: monitorResult(0)}
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	_, err := hub.DB().NewQuery(`CREATE TRIGGER fail_alert BEFORE UPDATE ON alerts BEGIN SELECT RAISE(ABORT, 'test rollback'); END`).Execute()
-	require.NoError(t, err)
-	count := hub.TestMailer.TotalSend()
-	results[monitors[0].Id] = monitorResult(10)
-	require.Error(t, am.HandleNetworkMonitorAlerts(system, results))
-	assert.Equal(t, count, hub.TestMailer.TotalSend())
-	histories, err := hub.CountRecords("alerts_history")
-	require.NoError(t, err)
-	assert.Zero(t, histories)
-	_, err = hub.DB().NewQuery("DROP TRIGGER fail_alert").Execute()
-	require.NoError(t, err)
-	// A failed transition must not be published to the cache and mask the retry.
-	require.NoError(t, am.HandleNetworkMonitorAlerts(system, results))
-	assert.Equal(t, count+1, hub.TestMailer.TotalSend())
-}

internal/alerts/alerts_system.go

@@ -47,7 +47,7 @@ func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *syst
 		return nil
 	}
 
-	alerts := am.alertsCache.GetAlertsExcludingNames(systemRecord.Id, "Status", alertNameSystemdFailed, containerAlertName, alertNameNetworkMonitorLoss)
+	alerts := am.alertsCache.GetAlertsExcludingNames(systemRecord.Id, "Status", alertNameSystemdFailed, containerAlertName)
 	if len(alerts) == 0 {
 		return nil
 	}
@@ -78,7 +78,7 @@ func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *syst
 				}
 			}
 			for _, pool := range data.Stats.ZfsPools {
-				if pool != nil && !pool.Raw && pool.Total > 0 {
+				if pool != nil && pool.Total > 0 {
 					usedPct := pool.Used / pool.Total * 100
 					if usedPct > maxUsedPct {
 						maxUsedPct = usedPct
@@ -256,7 +256,7 @@ func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *syst
 				}
 				// add zfs pool usage from historical record
 				for key, pool := range stats.ZfsPools {
-					if !pool.Raw && pool.Total > 0 {
+					if pool.Total > 0 {
 						zfsKey := zfsDiskAlertKey(key)
 						if _, ok := alert.mapSums[zfsKey]; !ok {
 							alert.mapSums[zfsKey] = 0.0
@@ -319,11 +319,6 @@ func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *syst
 				if sumPct > maxPct {
 					maxPct = sumPct
 					alert.descriptor = diskAlertDescriptor(key)
-					if poolKey, ok := strings.CutPrefix(key, "zfs:"); ok {
-						if pool := data.Stats.ZfsPools[poolKey]; pool != nil && pool.DisplayName != "" {
-							alert.descriptor = diskAlertDescriptor(zfsDiskAlertKey(pool.DisplayName))
-						}
-					}
 				}
 			}
 			alert.val = float64(maxPct / float32(alert.count))
@@ -375,7 +370,7 @@ func zfsDiskAlertKey(poolName string) string {
 
 func diskAlertDescriptor(key string) string {
 	if poolName, ok := strings.CutPrefix(key, "zfs:"); ok {
-		return fmt.Sprintf("Usage of storage pool %s", poolName)
+		return fmt.Sprintf("Usage of ZFS pool %s", poolName)
 	}
 	return fmt.Sprintf("Usage of %s", key)
 }

internal/alerts/alerts_test_helpers.go

@@ -11,9 +11,8 @@ import (
 
 func NewTestAlertManagerWithoutWorker(app hubLike) *AlertManager {
 	return &AlertManager{
-		hub:             app,
-		alertsCache:     NewAlertsCache(app),
-		networkMonitors: newNetworkMonitorCache(app),
+		hub:         app,
+		alertsCache: NewAlertsCache(app),
 	}
 }
 

internal/alerts/alerts_zfs.go

@@ -46,15 +46,12 @@ func (am *AlertManager) handleZfsPoolHealthAlert(e *core.RecordEvent, oldHealth
 	}
 
 	systemName := systemRecord.GetString("name")
-	poolName := e.Record.GetString("display_name")
-	if poolName == "" {
-		poolName = e.Record.GetString("name")
-	}
+	poolName := e.Record.GetString("name")
 
-	title := fmt.Sprintf("Storage pool %s on %s: %s", newHealth, systemName, poolName)
-	message := fmt.Sprintf("Storage pool %s (%s) was first observed as %s", poolName, systemName, newHealth)
+	title := fmt.Sprintf("ZFS pool %s on %s: %s", newHealth, systemName, poolName)
+	message := fmt.Sprintf("ZFS pool %s (%s) was first observed as %s", poolName, systemName, newHealth)
 	if oldSeverity > 0 {
-		message = fmt.Sprintf("Storage pool %s (%s) health changed from %s to %s", poolName, systemName, oldHealth, newHealth)
+		message = fmt.Sprintf("ZFS pool %s (%s) health changed from %s to %s", poolName, systemName, oldHealth, newHealth)
 	}
 
 	userIDs := systemRecord.GetStringSlice("users")
@@ -119,7 +116,7 @@ func createZfsPoolHistoryRecord(app core.App, userID, systemID, alertID, poolNam
 	record.Set("user", userID)
 	record.Set("system", systemID)
 	record.Set("alert_id", alertID)
-	record.Set("name", "Storage Pool: "+poolName)
+	record.Set("name", "ZFS Pool: "+poolName)
 	return app.Save(record)
 }
 

internal/alerts/alerts_zfs_disk_test.go

@@ -143,37 +143,3 @@ func TestDiskAlertZfsPoolMultiMinute(t *testing.T) {
 	assert.False(t, diskAlert.GetBool("triggered"),
 		"Alert should be resolved when ZFS pool average (50%%) drops below threshold (80%%)")
 }
-
-func TestDiskAlertIgnoresRawPool(t *testing.T) {
-	for _, minutes := range []int{0, 2} {
-		hub, user := beszelTests.GetHubWithUser(t)
-		systems, err := beszelTests.CreateSystems(hub, 1, user.Id, "up")
-		require.NoError(t, err)
-		alert, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{"name": "Disk", "system": systems[0].Id, "user": user.Id, "value": 80, "min": minutes})
-		require.NoError(t, err)
-		pools := map[string]*system.ZfsPool{"btrfs": {Total: 100, Used: 99, Raw: true}}
-		for _, offset := range []time.Duration{-180, -90, -60, -30} {
-			data, err := json.Marshal(system.Stats{ZfsPools: pools})
-			require.NoError(t, err)
-			record, err := beszelTests.CreateRecord(hub, "system_stats", map[string]any{"system": systems[0].Id, "type": "1m", "stats": string(data)})
-			require.NoError(t, err)
-			record.SetRaw("created", time.Now().UTC().Add(offset*time.Second).Format(types.DefaultDateLayout))
-			require.NoError(t, hub.SaveNoValidate(record))
-		}
-		require.NoError(t, hub.GetAlertManager().HandleSystemAlerts(systems[0], &system.CombinedData{Stats: system.Stats{ZfsPools: pools}}))
-		time.Sleep(20 * time.Millisecond)
-		record, err := hub.FindRecordById("alerts", alert.Id)
-		require.NoError(t, err)
-		assert.False(t, record.GetBool("triggered"))
-		if minutes > 0 {
-			// A current usable sample must not make raw historical values eligible.
-			pools["btrfs"].Raw = false
-			require.NoError(t, hub.GetAlertManager().HandleSystemAlerts(systems[0], &system.CombinedData{Stats: system.Stats{ZfsPools: pools}}))
-			time.Sleep(20 * time.Millisecond)
-			record, err = hub.FindRecordById("alerts", alert.Id)
-			require.NoError(t, err)
-			assert.False(t, record.GetBool("triggered"))
-		}
-		hub.Cleanup()
-	}
-}

internal/alerts/alerts_zfs_key_test.go

@@ -10,6 +10,6 @@ import (
 
 func TestZfsDiskAlertKeyIsNamespaced(t *testing.T) {
 	assert.Equal(t, "zfs:tank", zfsDiskAlertKey("tank"))
-	assert.Equal(t, "Usage of storage pool tank", diskAlertDescriptor(zfsDiskAlertKey("tank")))
+	assert.Equal(t, "Usage of ZFS pool tank", diskAlertDescriptor(zfsDiskAlertKey("tank")))
 	assert.Equal(t, "Usage of tank", diskAlertDescriptor("tank"))
 }

internal/alerts/alerts_zfs_test.go

@@ -42,7 +42,7 @@ func TestZfsPoolAlertOnlineToDegraded(t *testing.T) {
 
 	assert.EqualValues(t, 1, hub.TestMailer.TotalSend(), "should have 1 email sent after pool became DEGRADED")
 	lastMessage := hub.TestMailer.LastMessage()
-	assert.Contains(t, lastMessage.Subject, "Storage pool DEGRADED on test-system")
+	assert.Contains(t, lastMessage.Subject, "ZFS pool DEGRADED on test-system")
 	assert.Contains(t, lastMessage.Subject, "tank")
 	assert.Contains(t, lastMessage.Text, "ONLINE to DEGRADED")
 }
@@ -76,7 +76,7 @@ func TestZfsPoolAlertDegradedToFaulted(t *testing.T) {
 
 	assert.EqualValues(t, 2, hub.TestMailer.TotalSend(), "should alert on initial DEGRADED state and later FAULTED transition")
 	lastMessage := hub.TestMailer.LastMessage()
-	assert.Contains(t, lastMessage.Subject, "Storage pool FAULTED on test-system")
+	assert.Contains(t, lastMessage.Subject, "ZFS pool FAULTED on test-system")
 }
 
 func TestZfsPoolAlertNoAlertOnRecovery(t *testing.T) {
@@ -239,7 +239,7 @@ func TestZfsPoolAlertWritesHistory(t *testing.T) {
 	history, err := hub.FindRecordsByFilter("alerts_history", "alert_id={:alert_id}", "", 0, 0, map[string]any{"alert_id": pool.Id})
 	assert.NoError(t, err)
 	require.Len(t, history, 1, "expected one history entry per user")
-	assert.Equal(t, "Storage Pool: tank", history[0].GetString("name"))
+	assert.Equal(t, "ZFS Pool: tank", history[0].GetString("name"))
 	assert.Equal(t, system.Id, history[0].GetString("system"))
 }
 

internal/alerts/network_monitor_cache.go

@@ -1,79 +0,0 @@
-package alerts
-
-import (
-	"sync"
-
-	"github.com/pocketbase/dbx"
-	"github.com/pocketbase/pocketbase/core"
-)
-
-// networkMonitorCache keeps just the enabled monitor IDs and probe intervals
-// needed for the alert fast path. Names and targets are read only on transitions.
-// Returned maps are immutable; configuration changes invalidate the whole entry.
-type networkMonitorCache struct {
-	app     core.App
-	mu      sync.RWMutex
-	systems map[string]map[string]int
-}
-
-func newNetworkMonitorCache(app core.App) *networkMonitorCache {
-	c := &networkMonitorCache{app: app, systems: make(map[string]map[string]int)}
-	invalidate := func(e *core.RecordEvent) error {
-		c.invalidate(e.Record.GetString("system"))
-		return e.Next()
-	}
-	app.OnRecordAfterCreateSuccess("network_monitors").BindFunc(invalidate)
-	app.OnRecordAfterDeleteSuccess("network_monitors").BindFunc(invalidate)
-	app.OnRecordAfterUpdateSuccess("network_monitors").BindFunc(func(e *core.RecordEvent) error {
-		old := e.Record.Original()
-		// Realtime metric saves also invoke this hook. They must not evict config.
-		if old.GetString("system") != e.Record.GetString("system") ||
-			old.GetBool("enabled") != e.Record.GetBool("enabled") ||
-			old.GetInt("interval") != e.Record.GetInt("interval") {
-			c.invalidate(old.GetString("system"))
-			c.invalidate(e.Record.GetString("system"))
-		}
-		return e.Next()
-	})
-	app.OnRecordAfterDeleteSuccess("systems").BindFunc(func(e *core.RecordEvent) error {
-		c.invalidate(e.Record.Id)
-		return e.Next()
-	})
-	return c
-}
-
-func (c *networkMonitorCache) invalidate(systemID string) {
-	c.mu.Lock()
-	delete(c.systems, systemID)
-	c.mu.Unlock()
-}
-
-func (c *networkMonitorCache) get(systemID string) (map[string]int, error) {
-	c.mu.RLock()
-	monitors, ok := c.systems[systemID]
-	c.mu.RUnlock()
-	if ok {
-		return monitors, nil
-	}
-	c.mu.Lock()
-	defer c.mu.Unlock()
-	if monitors, ok := c.systems[systemID]; ok {
-		return monitors, nil
-	}
-	// Keep the lock through the load so a concurrent config change cannot be
-	// invalidated first and then overwritten by the older query result.
-	var rows []struct {
-		ID       string `db:"id"`
-		Interval int    `db:"interval"`
-	}
-	if err := c.app.DB().Select("id", "interval").From("network_monitors").
-		Where(dbx.HashExp{"system": systemID, "enabled": true}).All(&rows); err != nil {
-		return nil, err
-	}
-	monitors = make(map[string]int, len(rows))
-	for _, row := range rows {
-		monitors[row.ID] = row.Interval
-	}
-	c.systems[systemID] = monitors
-	return monitors, nil
-}

internal/alerts/notification_client_test.go

@@ -11,7 +11,6 @@ import (
 	"strings"
 	"sync/atomic"
 	"testing"
-	"testing/synctest"
 
 	"github.com/nicholas-fedor/shoutrrr/pkg/types"
 	"golang.org/x/net/dns/dnsmessage"
@@ -179,45 +178,39 @@ func TestPublicNotificationTCP(t *testing.T) {
 	} {
 		t.Run(rawURL, func(t *testing.T) {
 			t.Parallel()
-			// MQTT waits for a fixed library timeout even after a dial failure.
-			// Virtual time preserves the full send/cleanup path without that delay.
 			t.Run("internal destination", func(t *testing.T) {
-				synctest.Test(t, func(t *testing.T) {
-					err := sendPublicNotification(strings.ReplaceAll(rawURL, "HOST", "127.0.0.1"), "test")
-					if !errors.Is(err, errInternalDestination) {
-						t.Fatalf("expected blocked destination, got %v", err)
-					}
-				})
+				err := sendPublicNotification(strings.ReplaceAll(rawURL, "HOST", "127.0.0.1"), "test")
+				if !errors.Is(err, errInternalDestination) {
+					t.Fatalf("expected blocked destination, got %v", err)
+				}
 			})
 			t.Run("public destination uses injected dialer", func(t *testing.T) {
-				synctest.Test(t, func(t *testing.T) {
-					var calls atomic.Int32
-					stopped := errors.New("test dial stopped")
-					service, err := newPublicNotificationService(strings.ReplaceAll(rawURL, "HOST", "8.8.8.8"), types.SenderOptions{
-						DialContext: func(ctx context.Context, network, address string) (net.Conn, error) {
-							calls.Add(1)
-							if network != "tcp" || !strings.HasPrefix(address, "8.8.8.8:") {
-								t.Errorf("unexpected dial: %s %s", network, address)
-							}
-							if err := checkNotificationAddress(address); err != nil {
-								t.Error(err)
-							}
-							return nil, stopped
-						},
-					})
-					if err != nil {
-						t.Fatal(err)
-					}
-					if closer, ok := service.(io.Closer); ok {
-						defer closer.Close()
-					}
-					if err := service.Send("test", &types.Params{}); err == nil {
-						t.Fatal("expected dial failure")
-					}
-					if calls.Load() == 0 {
-						t.Fatal("custom dialer was not used")
-					}
+				var calls atomic.Int32
+				stopped := errors.New("test dial stopped")
+				service, err := newPublicNotificationService(strings.ReplaceAll(rawURL, "HOST", "8.8.8.8"), types.SenderOptions{
+					DialContext: func(ctx context.Context, network, address string) (net.Conn, error) {
+						calls.Add(1)
+						if network != "tcp" || !strings.HasPrefix(address, "8.8.8.8:") {
+							t.Errorf("unexpected dial: %s %s", network, address)
+						}
+						if err := checkNotificationAddress(address); err != nil {
+							t.Error(err)
+						}
+						return nil, stopped
+					},
 				})
+				if err != nil {
+					t.Fatal(err)
+				}
+				if closer, ok := service.(io.Closer); ok {
+					defer closer.Close()
+				}
+				if err := service.Send("test", &types.Params{}); err == nil {
+					t.Fatal("expected dial failure")
+				}
+				if calls.Load() == 0 {
+					t.Fatal("custom dialer was not used")
+				}
 			})
 		})
 	}

internal/cmd/agent/agent.go

@@ -1,11 +1,9 @@
 package main
 
 import (
-	"errors"
 	"fmt"
 	"log"
 	"os"
-	"runtime"
 	"strings"
 
 	"github.com/henrygd/beszel"
@@ -16,12 +14,6 @@ import (
 	"golang.org/x/crypto/ssh"
 )
 
-type noKeyProvidedError struct{}
-
-func (noKeyProvidedError) Error() string {
-	return "no key provided: must set -key flag, KEY env var, or KEY_FILE env var. Use 'beszel-agent help' for usage"
-}
-
 // cli options
 type cmdOptions struct {
 	key    string // key is the public key(s) for SSH authentication.
@@ -132,7 +124,7 @@ func (opts *cmdOptions) loadPublicKeys() ([]ssh.PublicKey, error) {
 	// Try key file
 	keyFile, ok := utils.GetEnv("KEY_FILE")
 	if !ok {
-		return nil, noKeyProvidedError{}
+		return nil, fmt.Errorf("no key provided: must set -key flag, KEY env var, or KEY_FILE env var. Use 'beszel-agent help' for usage")
 	}
 
 	pubKey, err := os.ReadFile(keyFile)
@@ -146,14 +138,6 @@ func (opts *cmdOptions) getAddress() string {
 	return agent.GetAddress(opts.listen)
 }
 
-func isBenignStartupError(err error, goos string) bool {
-	if goos != "windows" {
-		return false
-	}
-	var noKeyErr noKeyProvidedError
-	return errors.As(err, &noKeyErr)
-}
-
 // handleFingerprint handles the "fingerprint" command with subcommands "view" and "reset".
 func handleFingerprint() {
 	subCmd := ""
@@ -198,12 +182,6 @@ func main() {
 	var err error
 	serverConfig.Keys, err = opts.loadPublicKeys()
 	if err != nil {
-		if isBenignStartupError(err, runtime.GOOS) {
-			// WinGet launches the executable without configuration during validation.
-			// Exit successfully in that case while retaining the error on other platforms.
-			log.Print("Failed to load public keys:", err)
-			return
-		}
 		log.Fatal("Failed to load public keys:", err)
 	}
 

internal/cmd/agent/agent_test.go

@@ -2,7 +2,6 @@ package main
 
 import (
 	"crypto/ed25519"
-	"errors"
 	"os"
 	"path/filepath"
 	"testing"
@@ -188,26 +187,6 @@ func TestLoadPublicKeys(t *testing.T) {
 	}
 }
 
-func TestIsBenignStartupError(t *testing.T) {
-	tests := []struct {
-		name string
-		err  error
-		goos string
-		want bool
-	}{
-		{name: "missing key on windows", err: noKeyProvidedError{}, goos: "windows", want: true},
-		{name: "wrapped missing key on windows", err: errors.Join(errors.New("startup failed"), noKeyProvidedError{}), goos: "windows", want: true},
-		{name: "missing key on linux", err: noKeyProvidedError{}, goos: "linux", want: false},
-		{name: "different error on windows", err: errors.New("invalid key"), goos: "windows", want: false},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			assert.Equal(t, tt.want, isBenignStartupError(tt.err, tt.goos))
-		})
-	}
-}
-
 func TestGetNetwork(t *testing.T) {
 	tests := []struct {
 		name     string

internal/common/common-ws.go

@@ -24,8 +24,6 @@ const (
 	GetSystemdInfo
 	// Request ZFS detail data from agent
 	GetZfsData
-	// Sync network monitor configuration to agent
-	SyncNetworkMonitors
 	// Add new actions here...
 )
 

internal/dockerfile_agent

@@ -12,7 +12,7 @@ RUN apk add --no-cache ca-certificates && update-ca-certificates
 
 # Build
 ARG TARGETOS TARGETARCH
-RUN CGO_ENABLED=0 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /agent ./internal/cmd/agent
+RUN CGO_ENABLED=0 GOGC=75 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /agent ./internal/cmd/agent
 
 RUN rm -rf /tmp/*
 

internal/dockerfile_agent_alpine

@@ -10,14 +10,14 @@ COPY . ./
 
 # Build
 ARG TARGETOS TARGETARCH
-RUN CGO_ENABLED=0 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /agent ./internal/cmd/agent
+RUN CGO_ENABLED=0 GOGC=75 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /agent ./internal/cmd/agent
 
 RUN rm -rf /tmp/*
 
 # --------------------------
 # Final image: default scratch-based agent
 # --------------------------
-FROM alpine:3.24
+FROM alpine:3.23
 COPY --from=builder /agent /agent
 
 # AMD GPU name lookup (used by agent on Linux when /usr/share/libdrm/amdgpu.ids is read)
@@ -28,4 +28,4 @@ RUN apk add --no-cache smartmontools zfs
 # Ensure data persistence across container recreations
 VOLUME ["/var/lib/beszel-agent"]
 
-ENTRYPOINT ["/agent"]
+ENTRYPOINT ["/agent"]

internal/dockerfile_agent_intel

@@ -10,13 +10,13 @@ COPY . ./
 
 # Build
 ARG TARGETOS TARGETARCH
-RUN CGO_ENABLED=0 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /agent ./internal/cmd/agent
+RUN CGO_ENABLED=0 GOGC=75 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /agent ./internal/cmd/agent
 
 # --------------------------
 # Final image
 # Note: must cap_add: [CAP_PERFMON] and mount /dev/dri/ as volume
 # --------------------------
-FROM alpine:3.24
+FROM alpine:3.23
 
 COPY --from=builder /agent /agent
 

internal/dockerfile_agent_nvidia

@@ -10,7 +10,7 @@ COPY . ./
 
 # Build
 ARG TARGETOS TARGETARCH
-RUN CGO_ENABLED=0 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -tags glibc -ldflags "-w -s" -o /agent ./internal/cmd/agent
+RUN CGO_ENABLED=0 GOGC=75 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -tags glibc -ldflags "-w -s" -o /agent ./internal/cmd/agent
 
 # --------------------------
 # Smartmontools builder stage
@@ -43,11 +43,6 @@ COPY --from=builder /app/agent/test-data/amdgpu.ids /usr/share/libdrm/amdgpu.ids
 # Copy smartmontools binaries and config files
 COPY --from=smartmontools-builder /usr/sbin/smartctl /usr/sbin/smartctl
 
-# Install ZFS userspace utilities (zpool, zfs) for pool/dataset monitoring
-RUN apt-get update && apt-get install -y --no-install-recommends \
-  zfsutils-linux \
-  && apt-get clean && rm -rf /var/lib/apt/lists/*
-
 # Ensure data persistence across container recreations
 VOLUME ["/var/lib/beszel-agent"]
 

internal/dockerfile_agent_nvidia_slim

@@ -17,7 +17,7 @@ RUN set -eux; \
   if [ "$TARGETARCH" = "arm" ] && [ -n "$TARGETVARIANT" ]; then \
     export GOARM="${TARGETVARIANT#v}"; \
   fi; \
-  CGO_ENABLED=0 GOOS=$TARGETOS GOARCH=$TARGETARCH \
+  CGO_ENABLED=0 GOGC=75 GOOS=$TARGETOS GOARCH=$TARGETARCH \
     go build -tags glibc -ldflags "-w -s" -o /agent ./internal/cmd/agent
 
 # --------------------------
@@ -65,34 +65,6 @@ RUN set -eux; \
     cp -v "$interp" "/out/rootfs$interp"; \
   fi
 
-# --------------------------
-# ZFS utilities builder stage
-# --------------------------
-FROM --platform=$TARGETPLATFORM debian:bookworm-slim AS zfsutils-builder
-
-# zfsutils-linux is distributed in Debian's contrib component.
-RUN sed -i 's/Components: main/Components: main contrib/' /etc/apt/sources.list.d/debian.sources \
-  && apt-get update && apt-get install -y --no-install-recommends \
-  zfsutils-linux \
-  && rm -rf /var/lib/apt/lists/*
-
-# Copy the zpool/zfs binaries and their required runtime libraries
-RUN set -eux; \
-  mkdir -p /out/rootfs/lib /out/rootfs/lib64 /out/rootfs/usr/lib; \
-  for bin in /usr/sbin/zpool /usr/sbin/zfs; do \
-    mkdir -p "/out/rootfs$(dirname "$bin")"; \
-    cp -v "$bin" "/out/rootfs$bin"; \
-    ldd "$bin" \
-      | awk '{print $3}' \
-      | grep '^/' \
-      | xargs -r -I '{}' sh -c 'mkdir -p "/out/rootfs$(dirname "{}")"; cp -v "{}" "/out/rootfs{}"'; \
-    interp="$(ldd "$bin" | awk "/ld-linux/ {print \$1}")"; \
-    if [ -n "$interp" ] && [ -e "$interp" ]; then \
-      mkdir -p "/out/rootfs$(dirname "$interp")"; \
-      cp -v "$interp" "/out/rootfs$interp"; \
-    fi; \
-  done
-
 # --------------------------
 # Final image: lightweight multi-arch NVIDIA agent (slim)
 # --------------------------
@@ -106,9 +78,6 @@ COPY --from=builder /app/agent/test-data/amdgpu.ids /usr/share/libdrm/amdgpu.ids
 COPY --from=smartmontools-builder /usr/sbin/smartctl /usr/sbin/smartctl
 COPY --from=smartmontools-builder /out/rootfs/ /
 
-# Copy ZFS utilities (zpool, zfs) binaries and required runtime libraries
-COPY --from=zfsutils-builder /out/rootfs/ /
-
 # nvidia-smi is intentionally not bundled.
 # Mount the host binary instead, for example:
 # - /usr/bin/nvidia-smi:/usr/bin/nvidia-smi:ro

internal/dockerfile_hub

@@ -17,7 +17,7 @@ RUN update-ca-certificates
 
 # Build
 ARG TARGETOS TARGETARCH
-RUN CGO_ENABLED=0 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /beszel ./internal/cmd/hub
+RUN CGO_ENABLED=0 GOGC=75 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /beszel ./internal/cmd/hub
 
 # ? -------------------------
 FROM scratch
@@ -31,4 +31,4 @@ VOLUME ["/beszel_data"]
 EXPOSE 8090
 
 ENTRYPOINT [ "/beszel" ]
-CMD ["serve", "--http=0.0.0.0:8090"]
+CMD ["serve", "--http=0.0.0.0:8090"]

internal/entities/container/container.go

@@ -186,12 +186,11 @@ type Stats struct {
 	NetworkRecv float64   `json:"nr,omitzero" cbor:"4,keyasint,omitzero"` // deprecated 0.18.3 (MB) - keep field for old agents/records
 	Bandwidth   [2]uint64 `json:"b,omitzero" cbor:"9,keyasint,omitzero"`  // [sent bytes, recv bytes]
 
-	Health          DockerHealth `json:"-" cbor:"5,keyasint"`
-	Status          string       `json:"-" cbor:"6,keyasint"`
-	Id              string       `json:"-" cbor:"7,keyasint"`
-	Image           string       `json:"-" cbor:"8,keyasint"`
-	Ports           string       `json:"-" cbor:"10,keyasint"`
-	UpdateAvailable bool         `json:"u,omitzero" cbor:"11,keyasint,omitzero"`
+	Health DockerHealth `json:"-" cbor:"5,keyasint"`
+	Status string       `json:"-" cbor:"6,keyasint"`
+	Id     string       `json:"-" cbor:"7,keyasint"`
+	Image  string       `json:"-" cbor:"8,keyasint"`
+	Ports  string       `json:"-" cbor:"10,keyasint"`
 	// PrevCpu     [2]uint64    `json:"-"`
 	CpuSystem    uint64       `json:"-"`
 	CpuContainer uint64       `json:"-"`

internal/entities/monitor/monitor.go

@@ -1,114 +0,0 @@
-package monitor
-
-import "time"
-
-// MaxProbeTimeout is the longest agent probe timeout (currently HTTP).
-// Hub requests that run a probe must allow this time in addition to transport overhead.
-const MaxProbeTimeout = 10 * time.Second
-
-type SyncAction uint8
-
-const (
-	// SyncActionReplace indicates a full sync where the provided configs should replace all existing monitors for the system.
-	SyncActionReplace SyncAction = iota
-	// SyncActionUpsert indicates an incremental sync where the provided config should be added or updated.
-	SyncActionUpsert
-	// SyncActionDelete indicates an incremental sync where the provided config should be removed.
-	SyncActionDelete
-)
-
-// Config defines a network monitor task sent from hub to agent.
-type Config struct {
-	// ID is the stable network_monitors record ID generated by the hub.
-	ID       string `cbor:"0,keyasint"`
-	Target   string `cbor:"1,keyasint"`
-	Protocol string `cbor:"2,keyasint"` // "icmp", "tcp", "http", or "dns"
-	Port     uint16 `cbor:"3,keyasint,omitempty"`
-	Interval uint16 `cbor:"4,keyasint"` // seconds
-	// Server is the DNS server to query (host or host:port, default port 53).
-	// Only used when Protocol is "dns"; empty means use the system resolver.
-	Server string `cbor:"5,keyasint,omitempty"`
-}
-
-// CertInfo holds details of the leaf TLS certificate presented by a target.
-type CertInfo struct {
-	// Expires is the certificate's NotAfter Unix timestamp in milliseconds.
-	Expires int64  `cbor:"0,keyasint" json:"expires"`
-	Issuer  string `cbor:"1,keyasint,omitempty" json:"issuer,omitempty"`
-}
-
-// SyncRequest defines an incremental or full monitor sync request sent to the agent.
-type SyncRequest struct {
-	Action  SyncAction `cbor:"0,keyasint"`
-	Config  Config     `cbor:"1,keyasint,omitempty"`
-	Configs []Config   `cbor:"2,keyasint,omitempty"`
-	RunNow  bool       `cbor:"3,keyasint,omitempty"`
-}
-
-// SyncResponse returns the immediate result for an upsert when requested.
-type SyncResponse struct {
-	Result Result `cbor:"0,keyasint,omitempty"`
-}
-
-// Result holds aggregated monitor results for a single target.
-//
-// 0: avg response in microseconds
-//
-// 1: 1h average response in microseconds
-//
-// 2: min response in microseconds
-//
-// 3: 1h min response in microseconds
-//
-// 4: max response in microseconds
-//
-// 5: 1h max response in microseconds
-//
-// 6: packet loss percentage (0-100)
-//
-// 7: 1h packet loss percentage (0-100)
-type Result struct {
-	AvgResponse   int64   `cbor:"0,keyasint,omitempty"`
-	AvgResponse1h int64   `cbor:"1,keyasint,omitempty"`
-	MinResponse   int64   `cbor:"2,keyasint,omitempty"`
-	MinResponse1h int64   `cbor:"3,keyasint,omitempty"`
-	MaxResponse   int64   `cbor:"4,keyasint,omitempty"`
-	MaxResponse1h int64   `cbor:"5,keyasint,omitempty"`
-	PacketLoss    float64 `cbor:"6,keyasint,omitempty"`
-	PacketLoss1h  float64 `cbor:"7,keyasint,omitempty"`
-	// LastProbeAt is the latest completed probe's Unix timestamp in milliseconds.
-	LastProbeAt int64 `cbor:"8,keyasint"`
-	// SampleCount includes all completed probes since this monitor started.
-	// Used for alert warm-up even when the interval is longer than 20 minutes.
-	SampleCount int64 `cbor:"9,keyasint,omitempty"`
-	// Counts and sum cover the current response window (or latest-sample
-	// fallback), not the hourly window or lifetime SampleCount.
-	TotalCount   int64 `cbor:"10,keyasint"`
-	SuccessCount int64 `cbor:"11,keyasint"`
-	ResponseSum  int64 `cbor:"12,keyasint"`
-	// Cert is set for HTTPS targets when a certificate check has new info the hub has not stored yet.
-	Cert *CertInfo `cbor:"13,keyasint,omitempty"`
-}
-
-// Stats holds response times in microseconds and packet loss percentage (0-100).
-type Stats struct {
-	ResAvg       float64 `json:"res_avg" db:"-"` // Derived for display; not stored.
-	ResMin       float64 `json:"res_min" db:"res_min"`
-	ResMax       float64 `json:"res_max" db:"res_max"`
-	Loss         float64 `json:"loss" db:"-"` // Derived for display; not stored.
-	TotalCount   int64   `json:"-" db:"total_count"`
-	SuccessCount int64   `json:"-" db:"success_count"`
-	ResponseSum  int64   `json:"-" db:"res_sum"`
-}
-
-func (s Stats) FromResult(result Result) Stats {
-	return Stats{
-		ResAvg:       float64(result.AvgResponse),
-		ResMin:       float64(result.MinResponse),
-		ResMax:       float64(result.MaxResponse),
-		Loss:         result.PacketLoss,
-		TotalCount:   result.TotalCount,
-		SuccessCount: result.SuccessCount,
-		ResponseSum:  result.ResponseSum,
-	}
-}

internal/entities/system/system.go

@@ -7,7 +7,6 @@ import (
 	"time"
 
 	"github.com/henrygd/beszel/internal/entities/container"
-	"github.com/henrygd/beszel/internal/entities/monitor"
 	"github.com/henrygd/beszel/internal/entities/systemd"
 )
 
@@ -60,15 +59,11 @@ type Stats struct {
 
 // ZfsPool holds per-pool ZFS metrics for a single collection interval.
 type ZfsPool struct {
-	DisplayName string  `json:"n,omitempty" cbor:"8,keyasint,omitempty"`
-	HideUsage   bool    `json:"hu,omitempty" cbor:"6,keyasint,omitempty"` // equivalent filesystem usage chart exists
-	HideIO      bool    `json:"hi,omitempty" cbor:"7,keyasint,omitempty"` // equivalent filesystem I/O chart exists
-	Raw         bool    `json:"raw,omitempty" cbor:"5,keyasint,omitempty"`
-	Total       float64 `json:"d" cbor:"0,keyasint"`                     // total capacity in GiB
-	Used        float64 `json:"du" cbor:"1,keyasint"`                    // allocated in GiB
-	ReadBytes   uint64  `json:"rb,omitzero" cbor:"2,keyasint,omitzero"`  // read throughput in bytes/s
-	WriteBytes  uint64  `json:"wb,omitzero" cbor:"3,keyasint,omitzero"`  // write throughput in bytes/s
-	Health      string  `json:"h,omitempty" cbor:"4,keyasint,omitempty"` // ONLINE, DEGRADED, FAULTED, ...
+	Total      float64 `json:"d" cbor:"0,keyasint"`                     // total capacity in GiB
+	Used       float64 `json:"du" cbor:"1,keyasint"`                    // allocated in GiB
+	ReadBytes  uint64  `json:"rb,omitzero" cbor:"2,keyasint,omitzero"`  // read throughput in bytes/s
+	WriteBytes uint64  `json:"wb,omitzero" cbor:"3,keyasint,omitzero"`  // write throughput in bytes/s
+	Health     string  `json:"h,omitempty" cbor:"4,keyasint,omitempty"` // ONLINE, DEGRADED, FAULTED, ...
 }
 
 // Uint8Slice wraps []uint8 to customize JSON encoding while keeping CBOR efficient.
@@ -184,7 +179,6 @@ type Info struct {
 	Services       []uint16           `json:"sv,omitempty" cbor:"22,keyasint,omitempty"`  // [totalServices, numFailedServices]
 	Battery        Battery            `json:"bat,omitzero" cbor:"23,keyasint,omitzero"`   // [percent, charge state]
 	RootDiskName   string             `json:"rdn,omitempty" cbor:"24,keyasint,omitempty"` // custom name for root disk (set via FILESYSTEM=device__name)
-	PackageUpdates []uint16           `json:"pu,omitempty" cbor:"25,keyasint,omitempty"`  // [totalUpdates, securityUpdates] (security omitted if unknown)
 }
 
 // Data that does not change during process lifetime and is not needed in All Systems table
@@ -212,6 +206,5 @@ type CombinedData struct {
 	Details         *Details           `cbor:"4,keyasint,omitempty"`
 	// SystemdServicesUpdated distinguishes a fresh empty snapshot from a response
 	// that omitted systemd data (for example, a short-cache dashboard request).
-	SystemdServicesUpdated bool                      `json:"systemdUpdated,omitempty" cbor:"5,keyasint,omitempty"`
-	Monitors               map[string]monitor.Result `cbor:"6,keyasint"`
+	SystemdServicesUpdated bool `json:"systemdUpdated,omitempty" cbor:"5,keyasint,omitempty"`
 }