Compare commits

..

11 Commits

Author SHA1 Message Date
hank
24feffcc92 New translations en.po (German) 2026-03-14 16:37:21 -04:00
hank
9586b461d8 New translations en.po (Turkish) 2026-03-02 08:16:29 -05:00
hank
78936b1481 New translations en.po (Chinese Simplified) 2026-03-01 12:14:20 -05:00
hank
5b6af761f1 New translations en.po (German) 2026-02-27 10:20:45 -05:00
hank
29b914a43b New translations en.po (Bulgarian) 2026-02-25 06:49:42 -05:00
hank
4a86357d64 New translations en.po (Bulgarian) 2026-02-25 03:59:03 -05:00
hank
b65f011222 New translations en.po (Dutch) 2026-02-24 06:37:18 -05:00
hank
f836609552 New translations en.po (Serbian (Cyrillic)) 2026-02-23 10:11:15 -05:00
hank
b2a3c52005 New translations en.po (Russian) 2026-02-21 04:46:18 -05:00
hank
6e2277ead1 New translations en.po (Chinese Traditional, Hong Kong) 2026-02-19 21:25:17 -05:00
hank
cffc3d8569 New translations en.po (Chinese Simplified) 2026-02-19 21:25:16 -05:00
104 changed files with 1064 additions and 4035 deletions

View File

@@ -6,6 +6,7 @@ package agent
import (
"log/slog"
"os"
"strings"
"sync"
"time"
@@ -13,7 +14,6 @@ import (
"github.com/gliderlabs/ssh"
"github.com/henrygd/beszel"
"github.com/henrygd/beszel/agent/deltatracker"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/common"
"github.com/henrygd/beszel/internal/entities/system"
gossh "golang.org/x/crypto/ssh"
@@ -68,11 +68,11 @@ func NewAgent(dataDir ...string) (agent *Agent, err error) {
slog.Info("Data directory", "path", agent.dataDir)
}
agent.memCalc, _ = utils.GetEnv("MEM_CALC")
agent.memCalc, _ = GetEnv("MEM_CALC")
agent.sensorConfig = agent.newSensorConfig()
// Parse disk usage cache duration (e.g., "15m", "1h") to avoid waking sleeping disks
if diskUsageCache, exists := utils.GetEnv("DISK_USAGE_CACHE"); exists {
if diskUsageCache, exists := GetEnv("DISK_USAGE_CACHE"); exists {
if duration, err := time.ParseDuration(diskUsageCache); err == nil {
agent.diskUsageCacheDuration = duration
slog.Info("DISK_USAGE_CACHE", "duration", duration)
@@ -82,7 +82,7 @@ func NewAgent(dataDir ...string) (agent *Agent, err error) {
}
// Set up slog with a log level determined by the LOG_LEVEL env var
if logLevelStr, exists := utils.GetEnv("LOG_LEVEL"); exists {
if logLevelStr, exists := GetEnv("LOG_LEVEL"); exists {
switch strings.ToLower(logLevelStr) {
case "debug":
agent.debug = true
@@ -103,7 +103,7 @@ func NewAgent(dataDir ...string) (agent *Agent, err error) {
agent.refreshSystemDetails()
// SMART_INTERVAL env var to update smart data at this interval
if smartIntervalEnv, exists := utils.GetEnv("SMART_INTERVAL"); exists {
if smartIntervalEnv, exists := GetEnv("SMART_INTERVAL"); exists {
if duration, err := time.ParseDuration(smartIntervalEnv); err == nil && duration > 0 {
agent.systemDetails.SmartInterval = duration
slog.Info("SMART_INTERVAL", "duration", duration)
@@ -148,6 +148,15 @@ func NewAgent(dataDir ...string) (agent *Agent, err error) {
return agent, nil
}
// GetEnv retrieves an environment variable with a "BESZEL_AGENT_" prefix, or falls back to the unprefixed key.
func GetEnv(key string) (value string, exists bool) {
if value, exists = os.LookupEnv("BESZEL_AGENT_" + key); exists {
return value, exists
}
// Fallback to the old unprefixed key
return os.LookupEnv(key)
}
func (a *Agent) gatherStats(options common.DataRequestOptions) *system.CombinedData {
a.Lock()
defer a.Unlock()
@@ -204,7 +213,7 @@ func (a *Agent) gatherStats(options common.DataRequestOptions) *system.CombinedD
data.Stats.ExtraFs[key] = stats
// Add percentages to Info struct for dashboard
if stats.DiskTotal > 0 {
pct := utils.TwoDecimals((stats.DiskUsed / stats.DiskTotal) * 100)
pct := twoDecimals((stats.DiskUsed / stats.DiskTotal) * 100)
data.Info.ExtraFsPct[key] = pct
}
}

View File

@@ -14,7 +14,6 @@ import (
"time"
"github.com/henrygd/beszel"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/common"
"github.com/fxamacker/cbor/v2"
@@ -44,7 +43,7 @@ type WebSocketClient struct {
// newWebSocketClient creates a new WebSocket client for the given agent.
// It reads configuration from environment variables and validates the hub URL.
func newWebSocketClient(agent *Agent) (client *WebSocketClient, err error) {
hubURLStr, exists := utils.GetEnv("HUB_URL")
hubURLStr, exists := GetEnv("HUB_URL")
if !exists {
return nil, errors.New("HUB_URL environment variable not set")
}
@@ -73,12 +72,12 @@ func newWebSocketClient(agent *Agent) (client *WebSocketClient, err error) {
// If neither is set, it returns an error.
func getToken() (string, error) {
// get token from env var
token, _ := utils.GetEnv("TOKEN")
token, _ := GetEnv("TOKEN")
if token != "" {
return token, nil
}
// get token from file
tokenFile, _ := utils.GetEnv("TOKEN_FILE")
tokenFile, _ := GetEnv("TOKEN_FILE")
if tokenFile == "" {
return "", errors.New("must set TOKEN or TOKEN_FILE")
}
@@ -198,7 +197,7 @@ func (client *WebSocketClient) handleAuthChallenge(msg *common.HubRequest[cbor.R
}
if authRequest.NeedSysInfo {
response.Name, _ = utils.GetEnv("SYSTEM_NAME")
response.Name, _ = GetEnv("SYSTEM_NAME")
response.Hostname = client.agent.systemDetails.Hostname
serverAddr := client.agent.connectionManager.serverOptions.Addr
_, response.Port, _ = net.SplitHostPort(serverAddr)

View File

@@ -14,10 +14,10 @@ var lastPerCoreCpuTimes = make(map[uint16][]cpu.TimesStat)
// init initializes the CPU monitoring by storing the initial CPU times
// for the default 60-second cache interval.
func init() {
if times, err := cpu.Times(false); err == nil && len(times) > 0 {
if times, err := cpu.Times(false); err == nil {
lastCpuTimes[60000] = times[0]
}
if perCoreTimes, err := cpu.Times(true); err == nil && len(perCoreTimes) > 0 {
if perCoreTimes, err := cpu.Times(true); err == nil {
lastPerCoreCpuTimes[60000] = perCoreTimes
}
}

View File

@@ -6,8 +6,6 @@ import (
"os"
"path/filepath"
"runtime"
"github.com/henrygd/beszel/agent/utils"
)
// GetDataDir returns the path to the data directory for the agent and an error
@@ -18,7 +16,7 @@ func GetDataDir(dataDirs ...string) (string, error) {
return testDataDirs(dataDirs)
}
dataDir, _ := utils.GetEnv("DATA_DIR")
dataDir, _ := GetEnv("DATA_DIR")
if dataDir != "" {
dataDirs = append(dataDirs, dataDir)
}

View File

@@ -8,31 +8,11 @@ import (
"strings"
"time"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/system"
"github.com/shirou/gopsutil/v4/disk"
)
// fsRegistrationContext holds the shared lookup state needed to resolve a
// filesystem into the tracked fsStats key and metadata.
type fsRegistrationContext struct {
filesystem string // value of optional FILESYSTEM env var
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
// helper methods can share the same partitions, mount paths, and lookup functions
type diskDiscovery struct {
agent *Agent
rootMountPoint string
partitions []disk.PartitionStat
usageFn func(string) (*disk.UsageStat, error)
ctx fsRegistrationContext
}
// parseFilesystemEntry parses a filesystem entry in the format "device__customname"
// Returns the device/filesystem part and the custom name part
func parseFilesystemEntry(entry string) (device, customName string) {
@@ -46,230 +26,19 @@ func parseFilesystemEntry(entry string) (device, customName string) {
return device, customName
}
// extraFilesystemPartitionInfo derives the I/O device and optional display name
// for a mounted /extra-filesystems partition. Prefer the partition device reported
// by the system and only use the folder name for custom naming metadata.
func extraFilesystemPartitionInfo(p disk.PartitionStat) (device, customName string) {
device = strings.TrimSpace(p.Device)
folderDevice, customName := parseFilesystemEntry(filepath.Base(p.Mountpoint))
if device == "" {
device = folderDevice
}
return device, customName
}
func isDockerSpecialMountpoint(mountpoint string) bool {
switch mountpoint {
case "/etc/hosts", "/etc/resolv.conf", "/etc/hostname":
return true
}
return false
}
// registerFilesystemStats resolves the tracked key and stats payload for a
// filesystem before it is inserted into fsStats.
func registerFilesystemStats(existing map[string]*system.FsStats, device, mountpoint string, root bool, customName string, ctx fsRegistrationContext) (string, *system.FsStats, bool) {
key := device
if !ctx.isWindows {
key = filepath.Base(device)
}
if root {
// Try to map root device to a diskIoCounters entry. First checks for an
// exact key match, then uses findIoDevice for normalized / prefix-based
// matching (e.g. nda0p2 -> nda0), and finally falls back to FILESYSTEM.
if _, ioMatch := ctx.diskIoCounters[key]; !ioMatch {
if matchedKey, match := findIoDevice(key, ctx.diskIoCounters); match {
key = matchedKey
} else if ctx.filesystem != "" {
if matchedKey, match := findIoDevice(ctx.filesystem, ctx.diskIoCounters); match {
key = matchedKey
}
}
if _, ioMatch = ctx.diskIoCounters[key]; !ioMatch {
slog.Warn("Root I/O unmapped; set FILESYSTEM", "device", device, "mountpoint", mountpoint)
}
}
} else {
// Check if non-root has diskstats and prefer the folder device for
// /extra-filesystems mounts when the discovered partition device is a
// mapper path (e.g. luks UUID) that obscures the underlying block device.
if _, ioMatch := ctx.diskIoCounters[key]; !ioMatch {
if strings.HasPrefix(mountpoint, ctx.efPath) {
folderDevice, _ := parseFilesystemEntry(filepath.Base(mountpoint))
if folderDevice != "" {
if matchedKey, match := findIoDevice(folderDevice, ctx.diskIoCounters); match {
key = matchedKey
}
}
}
if _, ioMatch = ctx.diskIoCounters[key]; !ioMatch {
if matchedKey, match := findIoDevice(key, ctx.diskIoCounters); match {
key = matchedKey
}
}
}
}
if _, exists := existing[key]; exists {
return "", nil, false
}
fsStats := &system.FsStats{Root: root, Mountpoint: mountpoint}
if customName != "" {
fsStats.Name = customName
}
return key, fsStats, true
}
// addFsStat inserts a discovered filesystem if it resolves to a new tracking
// key. The key selection itself lives in buildFsStatRegistration so that logic
// can stay directly unit-tested.
func (d *diskDiscovery) addFsStat(device, mountpoint string, root bool, customName string) {
key, fsStats, ok := registerFilesystemStats(d.agent.fsStats, device, mountpoint, root, customName, d.ctx)
if !ok {
return
}
d.agent.fsStats[key] = fsStats
name := key
if customName != "" {
name = customName
}
slog.Info("Detected disk", "name", name, "device", device, "mount", mountpoint, "io", key, "root", root)
}
// addConfiguredRootFs resolves FILESYSTEM against partitions first, then falls
// back to direct diskstats matching for setups like ZFS where partitions do not
// expose the physical device name.
func (d *diskDiscovery) addConfiguredRootFs() bool {
if d.ctx.filesystem == "" {
default:
return false
}
for _, p := range d.partitions {
if filesystemMatchesPartitionSetting(d.ctx.filesystem, p) {
d.addFsStat(p.Device, p.Mountpoint, true, "")
return true
}
}
// FILESYSTEM may name a physical disk absent from partitions (e.g. ZFS lists
// dataset paths like zroot/ROOT/default, not block devices).
if ioKey, match := findIoDevice(d.ctx.filesystem, d.ctx.diskIoCounters); match {
d.agent.fsStats[ioKey] = &system.FsStats{Root: true, Mountpoint: d.rootMountPoint}
return true
}
slog.Warn("Partition details not found", "filesystem", d.ctx.filesystem)
return false
}
func isRootFallbackPartition(p disk.PartitionStat, rootMountPoint string) bool {
return p.Mountpoint == rootMountPoint ||
(isDockerSpecialMountpoint(p.Mountpoint) && strings.HasPrefix(p.Device, "/dev"))
}
// addPartitionRootFs handles the non-configured root fallback path when a
// partition looks like the active root mount but still needs translating to an
// I/O device key.
func (d *diskDiscovery) addPartitionRootFs(device, mountpoint string) bool {
fs, match := findIoDevice(filepath.Base(device), d.ctx.diskIoCounters)
if !match {
return false
}
// The resolved I/O device is already known here, so use it directly to avoid
// a second fallback search inside buildFsStatRegistration.
d.addFsStat(fs, mountpoint, true, "")
return true
}
// addLastResortRootFs is only used when neither FILESYSTEM nor partition-based
// heuristics can identify root, so it picks the busiest I/O device as a final
// fallback and preserves the root mountpoint for usage collection.
func (d *diskDiscovery) addLastResortRootFs() {
rootKey := mostActiveIoDevice(d.ctx.diskIoCounters)
if rootKey != "" {
slog.Warn("Using most active device for root I/O; set FILESYSTEM to override", "device", rootKey)
} else {
rootKey = filepath.Base(d.rootMountPoint)
if _, exists := d.agent.fsStats[rootKey]; exists {
rootKey = "root"
}
slog.Warn("Root I/O device not detected; set FILESYSTEM to override")
}
d.agent.fsStats[rootKey] = &system.FsStats{Root: true, Mountpoint: d.rootMountPoint}
}
// findPartitionByFilesystemSetting matches an EXTRA_FILESYSTEMS entry against a
// discovered partition either by mountpoint or by device suffix.
func findPartitionByFilesystemSetting(filesystem string, partitions []disk.PartitionStat) (disk.PartitionStat, bool) {
for _, p := range partitions {
if strings.HasSuffix(p.Device, filesystem) || p.Mountpoint == filesystem {
return p, true
}
}
return disk.PartitionStat{}, false
}
// addConfiguredExtraFsEntry resolves one EXTRA_FILESYSTEMS entry, preferring a
// discovered partition and falling back to any path that disk.Usage accepts.
func (d *diskDiscovery) addConfiguredExtraFsEntry(filesystem, customName string) {
if p, found := findPartitionByFilesystemSetting(filesystem, d.partitions); found {
d.addFsStat(p.Device, p.Mountpoint, false, customName)
return
}
if _, err := d.usageFn(filesystem); err == nil {
d.addFsStat(filepath.Base(filesystem), filesystem, false, customName)
return
} else {
slog.Error("Invalid filesystem", "name", filesystem, "err", err)
}
}
// addConfiguredExtraFilesystems parses and registers the comma-separated
// EXTRA_FILESYSTEMS env var entries.
func (d *diskDiscovery) addConfiguredExtraFilesystems(extraFilesystems string) {
for fsEntry := range strings.SplitSeq(extraFilesystems, ",") {
filesystem, customName := parseFilesystemEntry(fsEntry)
d.addConfiguredExtraFsEntry(filesystem, customName)
}
}
// addPartitionExtraFs registers partitions mounted under /extra-filesystems so
// their display names can come from the folder name while their I/O keys still
// prefer the underlying partition device.
func (d *diskDiscovery) addPartitionExtraFs(p disk.PartitionStat) {
if !strings.HasPrefix(p.Mountpoint, d.ctx.efPath) {
return
}
device, customName := extraFilesystemPartitionInfo(p)
d.addFsStat(device, p.Mountpoint, false, customName)
}
// addExtraFilesystemFolders handles bare directories under /extra-filesystems
// that may not appear in partition discovery, while skipping mountpoints that
// were already registered from higher-fidelity sources.
func (d *diskDiscovery) addExtraFilesystemFolders(folderNames []string) {
existingMountpoints := make(map[string]bool, len(d.agent.fsStats))
for _, stats := range d.agent.fsStats {
existingMountpoints[stats.Mountpoint] = true
}
for _, folderName := range folderNames {
mountpoint := filepath.Join(d.ctx.efPath, folderName)
slog.Debug("/extra-filesystems", "mountpoint", mountpoint)
if existingMountpoints[mountpoint] {
continue
}
device, customName := parseFilesystemEntry(folderName)
d.addFsStat(device, mountpoint, false, customName)
}
}
// Sets up the filesystems to monitor for disk usage and I/O.
func (a *Agent) initializeDiskInfo() {
filesystem, _ := utils.GetEnv("FILESYSTEM")
filesystem, _ := GetEnv("FILESYSTEM")
efPath := "/extra-filesystems"
hasRoot := false
isWindows := runtime.GOOS == "windows"
@@ -286,223 +55,164 @@ func (a *Agent) initializeDiskInfo() {
}
}
// ioContext := context.WithValue(a.sensorsContext,
// common.EnvKey, common.EnvMap{common.HostProcEnvKey: "/tmp/testproc"},
// )
// diskIoCounters, err := disk.IOCountersWithContext(ioContext)
diskIoCounters, err := disk.IOCounters()
if err != nil {
slog.Error("Error getting diskstats", "err", err)
}
slog.Debug("Disk I/O", "diskstats", diskIoCounters)
ctx := fsRegistrationContext{
filesystem: filesystem,
isWindows: isWindows,
diskIoCounters: diskIoCounters,
efPath: "/extra-filesystems",
// Helper function to add a filesystem to fsStats if it doesn't exist
addFsStat := func(device, mountpoint string, root bool, customName ...string) {
var key string
if isWindows {
key = device
} else {
key = filepath.Base(device)
}
var ioMatch bool
if _, exists := a.fsStats[key]; !exists {
if root {
slog.Info("Detected root device", "name", key)
// Check if root device is in /proc/diskstats. Do not guess a
// fallback device for root: that can misattribute root I/O to a
// different disk while usage remains tied to root mountpoint.
if _, ioMatch = diskIoCounters[key]; !ioMatch {
if matchedKey, match := findIoDevice(filesystem, diskIoCounters); match {
key = matchedKey
ioMatch = true
} else {
slog.Warn("Root I/O unmapped; set FILESYSTEM", "device", device, "mountpoint", mountpoint)
}
}
} else {
// Check if non-root has diskstats and fall back to folder name if not
// Scenario: device is encrypted and named luks-2bcb02be-999d-4417-8d18-5c61e660fb6e - not in /proc/diskstats.
// However, the device can be specified by mounting folder from luks device at /extra-filesystems/sda1
if _, ioMatch = diskIoCounters[key]; !ioMatch {
efBase := filepath.Base(mountpoint)
if _, ioMatch = diskIoCounters[efBase]; ioMatch {
key = efBase
}
}
}
fsStats := &system.FsStats{Root: root, Mountpoint: mountpoint}
if len(customName) > 0 && customName[0] != "" {
fsStats.Name = customName[0]
}
a.fsStats[key] = fsStats
}
}
// Get the appropriate root mount point for this system
discovery := diskDiscovery{
agent: a,
rootMountPoint: a.getRootMountPoint(),
partitions: partitions,
usageFn: disk.Usage,
ctx: ctx,
rootMountPoint := a.getRootMountPoint()
// Use FILESYSTEM env var to find root filesystem
if filesystem != "" {
for _, p := range partitions {
if strings.HasSuffix(p.Device, filesystem) || p.Mountpoint == filesystem {
addFsStat(p.Device, p.Mountpoint, true)
hasRoot = true
break
}
}
if !hasRoot {
slog.Warn("Partition details not found", "filesystem", filesystem)
}
}
hasRoot = discovery.addConfiguredRootFs()
// Add EXTRA_FILESYSTEMS env var values to fsStats
if extraFilesystems, exists := utils.GetEnv("EXTRA_FILESYSTEMS"); exists {
discovery.addConfiguredExtraFilesystems(extraFilesystems)
if extraFilesystems, exists := GetEnv("EXTRA_FILESYSTEMS"); exists {
for fsEntry := range strings.SplitSeq(extraFilesystems, ",") {
// Parse custom name from format: device__customname
fs, customName := parseFilesystemEntry(fsEntry)
found := false
for _, p := range partitions {
if strings.HasSuffix(p.Device, fs) || p.Mountpoint == fs {
addFsStat(p.Device, p.Mountpoint, false, customName)
found = true
break
}
}
// if not in partitions, test if we can get disk usage
if !found {
if _, err := disk.Usage(fs); err == nil {
addFsStat(filepath.Base(fs), fs, false, customName)
} else {
slog.Error("Invalid filesystem", "name", fs, "err", err)
}
}
}
}
// Process partitions for various mount points
for _, p := range partitions {
if !hasRoot && isRootFallbackPartition(p, discovery.rootMountPoint) {
hasRoot = discovery.addPartitionRootFs(p.Device, p.Mountpoint)
// fmt.Println(p.Device, p.Mountpoint)
// Binary root fallback or docker root fallback
if !hasRoot && (p.Mountpoint == rootMountPoint || (isDockerSpecialMountpoint(p.Mountpoint) && strings.HasPrefix(p.Device, "/dev"))) {
fs, match := findIoDevice(filepath.Base(p.Device), diskIoCounters)
if match {
addFsStat(fs, p.Mountpoint, true)
hasRoot = true
}
}
// Check if device is in /extra-filesystems
if strings.HasPrefix(p.Mountpoint, efPath) {
device, customName := parseFilesystemEntry(p.Mountpoint)
addFsStat(device, p.Mountpoint, false, customName)
}
discovery.addPartitionExtraFs(p)
}
// Check all folders in /extra-filesystems and add them if not already present
if folders, err := os.ReadDir(discovery.ctx.efPath); err == nil {
folderNames := make([]string, 0, len(folders))
if folders, err := os.ReadDir(efPath); err == nil {
existingMountpoints := make(map[string]bool)
for _, stats := range a.fsStats {
existingMountpoints[stats.Mountpoint] = true
}
for _, folder := range folders {
if folder.IsDir() {
folderNames = append(folderNames, folder.Name())
mountpoint := filepath.Join(efPath, folder.Name())
slog.Debug("/extra-filesystems", "mountpoint", mountpoint)
if !existingMountpoints[mountpoint] {
device, customName := parseFilesystemEntry(folder.Name())
addFsStat(device, mountpoint, false, customName)
}
}
}
discovery.addExtraFilesystemFolders(folderNames)
}
// If no root filesystem set, try the most active I/O device as a last
// resort (e.g. ZFS where dataset names are unrelated to disk names).
// If no root filesystem set, use fallback
if !hasRoot {
discovery.addLastResortRootFs()
rootKey := filepath.Base(rootMountPoint)
if _, exists := a.fsStats[rootKey]; exists {
rootKey = "root"
}
slog.Warn("Root device not detected; root I/O disabled", "mountpoint", rootMountPoint)
a.fsStats[rootKey] = &system.FsStats{Root: true, Mountpoint: rootMountPoint}
}
a.pruneDuplicateRootExtraFilesystems()
a.initializeDiskIoStats(diskIoCounters)
}
// Removes extra filesystems that mirror root usage (https://github.com/henrygd/beszel/issues/1428).
func (a *Agent) pruneDuplicateRootExtraFilesystems() {
var rootMountpoint string
for _, stats := range a.fsStats {
if stats != nil && stats.Root {
rootMountpoint = stats.Mountpoint
break
}
}
if rootMountpoint == "" {
return
}
rootUsage, err := disk.Usage(rootMountpoint)
if err != nil {
return
}
for name, stats := range a.fsStats {
if stats == nil || stats.Root {
continue
}
extraUsage, err := disk.Usage(stats.Mountpoint)
if err != nil {
continue
}
if hasSameDiskUsage(rootUsage, extraUsage) {
slog.Info("Ignoring duplicate FS", "name", name, "mount", stats.Mountpoint)
delete(a.fsStats, name)
}
}
}
// hasSameDiskUsage compares root/extra usage with a small byte tolerance.
func hasSameDiskUsage(a, b *disk.UsageStat) bool {
if a == nil || b == nil || a.Total == 0 || b.Total == 0 {
return false
}
// Allow minor drift between sequential disk usage calls.
const toleranceBytes uint64 = 16 * 1024 * 1024
return withinUsageTolerance(a.Total, b.Total, toleranceBytes) &&
withinUsageTolerance(a.Used, b.Used, toleranceBytes)
}
// withinUsageTolerance reports whether two byte values differ by at most tolerance.
func withinUsageTolerance(a, b, tolerance uint64) bool {
if a >= b {
return a-b <= tolerance
}
return b-a <= tolerance
}
type ioMatchCandidate struct {
name string
bytes uint64
ops uint64
}
// findIoDevice prefers exact device/label matches, then falls back to a
// prefix-related candidate with the highest recent activity.
// Returns matching device from /proc/diskstats.
// bool is true if a match was found.
func findIoDevice(filesystem string, diskIoCounters map[string]disk.IOCountersStat) (string, bool) {
filesystem = normalizeDeviceName(filesystem)
if filesystem == "" {
return "", false
}
candidates := []ioMatchCandidate{}
for _, d := range diskIoCounters {
if normalizeDeviceName(d.Name) == filesystem || (d.Label != "" && normalizeDeviceName(d.Label) == filesystem) {
if d.Name == filesystem || (d.Label != "" && d.Label == filesystem) {
return d.Name, true
}
if prefixRelated(normalizeDeviceName(d.Name), filesystem) ||
(d.Label != "" && prefixRelated(normalizeDeviceName(d.Label), filesystem)) {
candidates = append(candidates, ioMatchCandidate{
name: d.Name,
bytes: d.ReadBytes + d.WriteBytes,
ops: d.ReadCount + d.WriteCount,
})
}
}
if len(candidates) == 0 {
return "", false
}
best := candidates[0]
for _, c := range candidates[1:] {
if c.bytes > best.bytes ||
(c.bytes == best.bytes && c.ops > best.ops) ||
(c.bytes == best.bytes && c.ops == best.ops && c.name < best.name) {
best = c
}
}
slog.Info("Using disk I/O fallback", "requested", filesystem, "selected", best.name)
return best.name, true
}
// mostActiveIoDevice returns the device with the highest I/O activity,
// or "" if diskIoCounters is empty.
func mostActiveIoDevice(diskIoCounters map[string]disk.IOCountersStat) string {
var best ioMatchCandidate
for _, d := range diskIoCounters {
c := ioMatchCandidate{
name: d.Name,
bytes: d.ReadBytes + d.WriteBytes,
ops: d.ReadCount + d.WriteCount,
}
if best.name == "" || c.bytes > best.bytes ||
(c.bytes == best.bytes && c.ops > best.ops) ||
(c.bytes == best.bytes && c.ops == best.ops && c.name < best.name) {
best = c
}
}
return best.name
}
// prefixRelated reports whether either identifier is a prefix of the other.
func prefixRelated(a, b string) bool {
if a == "" || b == "" || a == b {
return false
}
return strings.HasPrefix(a, b) || strings.HasPrefix(b, a)
}
// filesystemMatchesPartitionSetting checks whether a FILESYSTEM env var value
// matches a partition by mountpoint, exact device name, or prefix relationship
// (e.g. FILESYSTEM=ada0 matches partition /dev/ada0p2).
func filesystemMatchesPartitionSetting(filesystem string, p disk.PartitionStat) bool {
filesystem = strings.TrimSpace(filesystem)
if filesystem == "" {
return false
}
if p.Mountpoint == filesystem {
return true
}
fsName := normalizeDeviceName(filesystem)
partName := normalizeDeviceName(p.Device)
if fsName == "" || partName == "" {
return false
}
if fsName == partName {
return true
}
return prefixRelated(partName, fsName)
}
// normalizeDeviceName canonicalizes device strings for comparisons.
func normalizeDeviceName(value string) string {
name := filepath.Base(strings.TrimSpace(value))
if name == "." {
return ""
}
return name
return "", false
}
// Sets start values for disk I/O stats.
func (a *Agent) initializeDiskIoStats(diskIoCounters map[string]disk.IOCountersStat) {
a.fsNames = a.fsNames[:0]
now := time.Now()
for device, stats := range a.fsStats {
// skip if not in diskIoCounters
d, exists := diskIoCounters[device]
@@ -511,7 +221,7 @@ func (a *Agent) initializeDiskIoStats(diskIoCounters map[string]disk.IOCountersS
continue
}
// populate initial values
stats.Time = now
stats.Time = time.Now()
stats.TotalRead = d.ReadBytes
stats.TotalWrite = d.WriteBytes
// add to list of valid io device names
@@ -535,12 +245,12 @@ func (a *Agent) updateDiskUsage(systemStats *system.Stats) {
continue
}
if d, err := disk.Usage(stats.Mountpoint); err == nil {
stats.DiskTotal = utils.BytesToGigabytes(d.Total)
stats.DiskUsed = utils.BytesToGigabytes(d.Used)
stats.DiskTotal = bytesToGigabytes(d.Total)
stats.DiskUsed = bytesToGigabytes(d.Used)
if stats.Root {
systemStats.DiskTotal = utils.BytesToGigabytes(d.Total)
systemStats.DiskUsed = utils.BytesToGigabytes(d.Used)
systemStats.DiskPct = utils.TwoDecimals(d.UsedPercent)
systemStats.DiskTotal = bytesToGigabytes(d.Total)
systemStats.DiskUsed = bytesToGigabytes(d.Used)
systemStats.DiskPct = twoDecimals(d.UsedPercent)
}
} else {
// reset stats if error (likely unmounted)
@@ -593,8 +303,8 @@ func (a *Agent) updateDiskIo(cacheTimeMs uint16, systemStats *system.Stats) {
diskIORead := (d.ReadBytes - prev.readBytes) * 1000 / msElapsed
diskIOWrite := (d.WriteBytes - prev.writeBytes) * 1000 / msElapsed
readMbPerSecond := utils.BytesToMegabytes(float64(diskIORead))
writeMbPerSecond := utils.BytesToMegabytes(float64(diskIOWrite))
readMbPerSecond := bytesToMegabytes(float64(diskIORead))
writeMbPerSecond := bytesToMegabytes(float64(diskIOWrite))
// validate values
if readMbPerSecond > 50_000 || writeMbPerSecond > 50_000 {

View File

@@ -93,443 +93,6 @@ func TestParseFilesystemEntry(t *testing.T) {
}
}
func TestExtraFilesystemPartitionInfo(t *testing.T) {
t.Run("uses partition device for label-only mountpoint", func(t *testing.T) {
device, customName := extraFilesystemPartitionInfo(disk.PartitionStat{
Device: "/dev/sdc",
Mountpoint: "/extra-filesystems/Share",
})
assert.Equal(t, "/dev/sdc", device)
assert.Equal(t, "", customName)
})
t.Run("uses custom name from mountpoint suffix", func(t *testing.T) {
device, customName := extraFilesystemPartitionInfo(disk.PartitionStat{
Device: "/dev/sdc",
Mountpoint: "/extra-filesystems/sdc__Share",
})
assert.Equal(t, "/dev/sdc", device)
assert.Equal(t, "Share", customName)
})
t.Run("falls back to folder device when partition device is unavailable", func(t *testing.T) {
device, customName := extraFilesystemPartitionInfo(disk.PartitionStat{
Mountpoint: "/extra-filesystems/sdc__Share",
})
assert.Equal(t, "sdc", device)
assert.Equal(t, "Share", customName)
})
t.Run("supports custom name without folder device prefix", func(t *testing.T) {
device, customName := extraFilesystemPartitionInfo(disk.PartitionStat{
Device: "/dev/sdc",
Mountpoint: "/extra-filesystems/__Share",
})
assert.Equal(t, "/dev/sdc", device)
assert.Equal(t, "Share", customName)
})
}
func TestBuildFsStatRegistration(t *testing.T) {
t.Run("uses basename for non-windows exact io match", func(t *testing.T) {
key, stats, ok := registerFilesystemStats(
map[string]*system.FsStats{},
"/dev/sda1",
"/mnt/data",
false,
"archive",
fsRegistrationContext{
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"sda1": {Name: "sda1"},
},
},
)
assert.True(t, ok)
assert.Equal(t, "sda1", key)
assert.Equal(t, "/mnt/data", stats.Mountpoint)
assert.Equal(t, "archive", stats.Name)
assert.False(t, stats.Root)
})
t.Run("maps root partition to io device by prefix", func(t *testing.T) {
key, stats, ok := registerFilesystemStats(
map[string]*system.FsStats{},
"/dev/ada0p2",
"/",
true,
"",
fsRegistrationContext{
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"ada0": {Name: "ada0", ReadBytes: 1000, WriteBytes: 1000},
},
},
)
assert.True(t, ok)
assert.Equal(t, "ada0", key)
assert.True(t, stats.Root)
assert.Equal(t, "/", stats.Mountpoint)
})
t.Run("uses filesystem setting as root fallback", func(t *testing.T) {
key, _, ok := registerFilesystemStats(
map[string]*system.FsStats{},
"overlay",
"/",
true,
"",
fsRegistrationContext{
filesystem: "nvme0n1p2",
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"nvme0n1": {Name: "nvme0n1", ReadBytes: 1000, WriteBytes: 1000},
},
},
)
assert.True(t, ok)
assert.Equal(t, "nvme0n1", key)
})
t.Run("prefers parsed extra-filesystems device over mapper device", func(t *testing.T) {
key, stats, ok := registerFilesystemStats(
map[string]*system.FsStats{},
"/dev/mapper/luks-2bcb02be-999d-4417-8d18-5c61e660fb6e",
"/extra-filesystems/nvme0n1p2__Archive",
false,
"Archive",
fsRegistrationContext{
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"dm-1": {Name: "dm-1", Label: "luks-2bcb02be-999d-4417-8d18-5c61e660fb6e"},
"nvme0n1p2": {Name: "nvme0n1p2"},
},
},
)
assert.True(t, ok)
assert.Equal(t, "nvme0n1p2", key)
assert.Equal(t, "Archive", stats.Name)
})
t.Run("falls back to mapper io device when folder device cannot be resolved", func(t *testing.T) {
key, stats, ok := registerFilesystemStats(
map[string]*system.FsStats{},
"/dev/mapper/luks-2bcb02be-999d-4417-8d18-5c61e660fb6e",
"/extra-filesystems/Archive",
false,
"Archive",
fsRegistrationContext{
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"dm-1": {Name: "dm-1", Label: "luks-2bcb02be-999d-4417-8d18-5c61e660fb6e"},
},
},
)
assert.True(t, ok)
assert.Equal(t, "dm-1", key)
assert.Equal(t, "Archive", stats.Name)
})
t.Run("uses full device name on windows", func(t *testing.T) {
key, _, ok := registerFilesystemStats(
map[string]*system.FsStats{},
`C:`,
`C:\\`,
false,
"",
fsRegistrationContext{
isWindows: true,
diskIoCounters: map[string]disk.IOCountersStat{
`C:`: {Name: `C:`},
},
},
)
assert.True(t, ok)
assert.Equal(t, `C:`, key)
})
t.Run("skips existing key", func(t *testing.T) {
key, stats, ok := registerFilesystemStats(
map[string]*system.FsStats{"sda1": {Mountpoint: "/existing"}},
"/dev/sda1",
"/mnt/data",
false,
"",
fsRegistrationContext{
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"sda1": {Name: "sda1"},
},
},
)
assert.False(t, ok)
assert.Empty(t, key)
assert.Nil(t, stats)
})
}
func TestAddConfiguredRootFs(t *testing.T) {
t.Run("adds root from matching partition", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{
agent: agent,
rootMountPoint: "/",
partitions: []disk.PartitionStat{{Device: "/dev/ada0p2", Mountpoint: "/"}},
ctx: fsRegistrationContext{
filesystem: "/dev/ada0p2",
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"ada0": {Name: "ada0", ReadBytes: 1000, WriteBytes: 1000},
},
},
}
ok := discovery.addConfiguredRootFs()
assert.True(t, ok)
stats, exists := agent.fsStats["ada0"]
assert.True(t, exists)
assert.True(t, stats.Root)
assert.Equal(t, "/", stats.Mountpoint)
})
t.Run("adds root from io device when partition is missing", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{
agent: agent,
rootMountPoint: "/sysroot",
ctx: fsRegistrationContext{
filesystem: "zroot",
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"nda0": {Name: "nda0", Label: "zroot", ReadBytes: 1000, WriteBytes: 1000},
},
},
}
ok := discovery.addConfiguredRootFs()
assert.True(t, ok)
stats, exists := agent.fsStats["nda0"]
assert.True(t, exists)
assert.True(t, stats.Root)
assert.Equal(t, "/sysroot", stats.Mountpoint)
})
t.Run("returns false when filesystem cannot be resolved", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{
agent: agent,
rootMountPoint: "/",
ctx: fsRegistrationContext{
filesystem: "missing-disk",
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{},
},
}
ok := discovery.addConfiguredRootFs()
assert.False(t, ok)
assert.Empty(t, agent.fsStats)
})
}
func TestAddPartitionRootFs(t *testing.T) {
t.Run("adds root from fallback partition candidate", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{
agent: agent,
ctx: fsRegistrationContext{
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"nvme0n1": {Name: "nvme0n1", ReadBytes: 1000, WriteBytes: 1000},
},
},
}
ok := discovery.addPartitionRootFs("/dev/nvme0n1p2", "/")
assert.True(t, ok)
stats, exists := agent.fsStats["nvme0n1"]
assert.True(t, exists)
assert.True(t, stats.Root)
assert.Equal(t, "/", stats.Mountpoint)
})
t.Run("returns false when no io device matches", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{agent: agent, ctx: fsRegistrationContext{diskIoCounters: map[string]disk.IOCountersStat{}}}
ok := discovery.addPartitionRootFs("/dev/mapper/root", "/")
assert.False(t, ok)
assert.Empty(t, agent.fsStats)
})
}
func TestAddLastResortRootFs(t *testing.T) {
t.Run("uses most active io device when available", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{agent: agent, rootMountPoint: "/", ctx: fsRegistrationContext{diskIoCounters: map[string]disk.IOCountersStat{
"sda": {Name: "sda", ReadBytes: 5000, WriteBytes: 5000},
"sdb": {Name: "sdb", ReadBytes: 1000, WriteBytes: 1000},
}}}
discovery.addLastResortRootFs()
stats, exists := agent.fsStats["sda"]
assert.True(t, exists)
assert.True(t, stats.Root)
})
t.Run("falls back to root key when mountpoint basename collides", func(t *testing.T) {
agent := &Agent{fsStats: map[string]*system.FsStats{
"sysroot": {Mountpoint: "/extra-filesystems/sysroot"},
}}
discovery := diskDiscovery{agent: agent, rootMountPoint: "/sysroot", ctx: fsRegistrationContext{diskIoCounters: map[string]disk.IOCountersStat{}}}
discovery.addLastResortRootFs()
stats, exists := agent.fsStats["root"]
assert.True(t, exists)
assert.True(t, stats.Root)
assert.Equal(t, "/sysroot", stats.Mountpoint)
})
}
func TestAddConfiguredExtraFsEntry(t *testing.T) {
t.Run("uses matching partition when present", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{
agent: agent,
partitions: []disk.PartitionStat{{Device: "/dev/sdb1", Mountpoint: "/mnt/backup"}},
usageFn: func(string) (*disk.UsageStat, error) {
t.Fatal("usage fallback should not be called when partition matches")
return nil, nil
},
ctx: fsRegistrationContext{
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"sdb1": {Name: "sdb1"},
},
},
}
discovery.addConfiguredExtraFsEntry("sdb1", "backup")
stats, exists := agent.fsStats["sdb1"]
assert.True(t, exists)
assert.Equal(t, "/mnt/backup", stats.Mountpoint)
assert.Equal(t, "backup", stats.Name)
})
t.Run("falls back to usage-validated path", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{
agent: agent,
usageFn: func(path string) (*disk.UsageStat, error) {
assert.Equal(t, "/srv/archive", path)
return &disk.UsageStat{}, nil
},
ctx: fsRegistrationContext{
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"archive": {Name: "archive"},
},
},
}
discovery.addConfiguredExtraFsEntry("/srv/archive", "archive")
stats, exists := agent.fsStats["archive"]
assert.True(t, exists)
assert.Equal(t, "/srv/archive", stats.Mountpoint)
assert.Equal(t, "archive", stats.Name)
})
t.Run("ignores invalid filesystem entry", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{
agent: agent,
usageFn: func(string) (*disk.UsageStat, error) {
return nil, os.ErrNotExist
},
}
discovery.addConfiguredExtraFsEntry("/missing/archive", "")
assert.Empty(t, agent.fsStats)
})
}
func TestAddConfiguredExtraFilesystems(t *testing.T) {
t.Run("parses and registers multiple configured filesystems", func(t *testing.T) {
agent := &Agent{fsStats: make(map[string]*system.FsStats)}
discovery := diskDiscovery{
agent: agent,
partitions: []disk.PartitionStat{{Device: "/dev/sda1", Mountpoint: "/mnt/fast"}},
usageFn: func(path string) (*disk.UsageStat, error) {
if path == "/srv/archive" {
return &disk.UsageStat{}, nil
}
return nil, os.ErrNotExist
},
ctx: fsRegistrationContext{
isWindows: false,
diskIoCounters: map[string]disk.IOCountersStat{
"sda1": {Name: "sda1"},
"archive": {Name: "archive"},
},
},
}
discovery.addConfiguredExtraFilesystems("sda1__fast,/srv/archive__cold")
assert.Contains(t, agent.fsStats, "sda1")
assert.Equal(t, "fast", agent.fsStats["sda1"].Name)
assert.Contains(t, agent.fsStats, "archive")
assert.Equal(t, "cold", agent.fsStats["archive"].Name)
})
}
func TestAddExtraFilesystemFolders(t *testing.T) {
t.Run("adds missing folders and skips existing mountpoints", func(t *testing.T) {
agent := &Agent{fsStats: map[string]*system.FsStats{
"existing": {Mountpoint: "/extra-filesystems/existing"},
}}
discovery := diskDiscovery{
agent: agent,
ctx: fsRegistrationContext{
isWindows: false,
efPath: "/extra-filesystems",
diskIoCounters: map[string]disk.IOCountersStat{
"newdisk": {Name: "newdisk"},
},
},
}
discovery.addExtraFilesystemFolders([]string{"existing", "newdisk__Archive"})
assert.Len(t, agent.fsStats, 2)
stats, exists := agent.fsStats["newdisk"]
assert.True(t, exists)
assert.Equal(t, "/extra-filesystems/newdisk__Archive", stats.Mountpoint)
assert.Equal(t, "Archive", stats.Name)
})
}
func TestFindIoDevice(t *testing.T) {
t.Run("matches by device name", func(t *testing.T) {
ioCounters := map[string]disk.IOCountersStat{
@@ -553,7 +116,7 @@ func TestFindIoDevice(t *testing.T) {
assert.Equal(t, "sda", device)
})
t.Run("returns no match when not found", func(t *testing.T) {
t.Run("returns no fallback when not found", func(t *testing.T) {
ioCounters := map[string]disk.IOCountersStat{
"sda": {Name: "sda"},
"sdb": {Name: "sdb"},
@@ -563,106 +126,6 @@ func TestFindIoDevice(t *testing.T) {
assert.False(t, ok)
assert.Equal(t, "", device)
})
t.Run("uses uncertain unique prefix fallback", func(t *testing.T) {
ioCounters := map[string]disk.IOCountersStat{
"nvme0n1": {Name: "nvme0n1"},
"sda": {Name: "sda"},
}
device, ok := findIoDevice("nvme0n1p2", ioCounters)
assert.True(t, ok)
assert.Equal(t, "nvme0n1", device)
})
t.Run("uses dominant activity when prefix matches are ambiguous", func(t *testing.T) {
ioCounters := map[string]disk.IOCountersStat{
"sda": {Name: "sda", ReadBytes: 5000, WriteBytes: 5000, ReadCount: 100, WriteCount: 100},
"sdb": {Name: "sdb", ReadBytes: 1000, WriteBytes: 1000, ReadCount: 50, WriteCount: 50},
}
device, ok := findIoDevice("sd", ioCounters)
assert.True(t, ok)
assert.Equal(t, "sda", device)
})
t.Run("uses highest activity when ambiguous without dominance", func(t *testing.T) {
ioCounters := map[string]disk.IOCountersStat{
"sda": {Name: "sda", ReadBytes: 3000, WriteBytes: 3000, ReadCount: 50, WriteCount: 50},
"sdb": {Name: "sdb", ReadBytes: 2500, WriteBytes: 2500, ReadCount: 40, WriteCount: 40},
}
device, ok := findIoDevice("sd", ioCounters)
assert.True(t, ok)
assert.Equal(t, "sda", device)
})
t.Run("matches /dev/-prefixed partition to parent disk", func(t *testing.T) {
ioCounters := map[string]disk.IOCountersStat{
"nda0": {Name: "nda0", ReadBytes: 1000, WriteBytes: 1000},
}
device, ok := findIoDevice("/dev/nda0p2", ioCounters)
assert.True(t, ok)
assert.Equal(t, "nda0", device)
})
t.Run("uses deterministic name tie-breaker", func(t *testing.T) {
ioCounters := map[string]disk.IOCountersStat{
"sdb": {Name: "sdb", ReadBytes: 2000, WriteBytes: 2000, ReadCount: 10, WriteCount: 10},
"sda": {Name: "sda", ReadBytes: 2000, WriteBytes: 2000, ReadCount: 10, WriteCount: 10},
}
device, ok := findIoDevice("sd", ioCounters)
assert.True(t, ok)
assert.Equal(t, "sda", device)
})
}
func TestFilesystemMatchesPartitionSetting(t *testing.T) {
p := disk.PartitionStat{Device: "/dev/ada0p2", Mountpoint: "/"}
t.Run("matches mountpoint setting", func(t *testing.T) {
assert.True(t, filesystemMatchesPartitionSetting("/", p))
})
t.Run("matches exact partition setting", func(t *testing.T) {
assert.True(t, filesystemMatchesPartitionSetting("ada0p2", p))
assert.True(t, filesystemMatchesPartitionSetting("/dev/ada0p2", p))
})
t.Run("matches prefix-style parent setting", func(t *testing.T) {
assert.True(t, filesystemMatchesPartitionSetting("ada0", p))
assert.True(t, filesystemMatchesPartitionSetting("/dev/ada0", p))
})
t.Run("does not match unrelated device", func(t *testing.T) {
assert.False(t, filesystemMatchesPartitionSetting("sda", p))
assert.False(t, filesystemMatchesPartitionSetting("nvme0n1", p))
assert.False(t, filesystemMatchesPartitionSetting("", p))
})
}
func TestMostActiveIoDevice(t *testing.T) {
t.Run("returns most active device", func(t *testing.T) {
ioCounters := map[string]disk.IOCountersStat{
"nda0": {Name: "nda0", ReadBytes: 5000, WriteBytes: 5000, ReadCount: 100, WriteCount: 100},
"nda1": {Name: "nda1", ReadBytes: 1000, WriteBytes: 1000, ReadCount: 50, WriteCount: 50},
}
assert.Equal(t, "nda0", mostActiveIoDevice(ioCounters))
})
t.Run("uses deterministic tie-breaker", func(t *testing.T) {
ioCounters := map[string]disk.IOCountersStat{
"sdb": {Name: "sdb", ReadBytes: 1000, WriteBytes: 1000, ReadCount: 10, WriteCount: 10},
"sda": {Name: "sda", ReadBytes: 1000, WriteBytes: 1000, ReadCount: 10, WriteCount: 10},
}
assert.Equal(t, "sda", mostActiveIoDevice(ioCounters))
})
t.Run("returns empty for empty map", func(t *testing.T) {
assert.Equal(t, "", mostActiveIoDevice(map[string]disk.IOCountersStat{}))
})
}
func TestIsDockerSpecialMountpoint(t *testing.T) {
@@ -747,7 +210,7 @@ func TestInitializeDiskInfoWithCustomNames(t *testing.T) {
// Test the parsing logic by calling the relevant part
// We'll create a simplified version to test just the parsing
extraFilesystems := tc.envValue
for fsEntry := range strings.SplitSeq(extraFilesystems, ",") {
for _, fsEntry := range strings.Split(extraFilesystems, ",") {
// Parse the entry
fsEntry = strings.TrimSpace(fsEntry)
var fs, customName string
@@ -909,67 +372,3 @@ func TestDiskUsageCaching(t *testing.T) {
"lastDiskUsageUpdate should be refreshed when cache expires")
})
}
func TestHasSameDiskUsage(t *testing.T) {
const toleranceBytes uint64 = 16 * 1024 * 1024
t.Run("returns true when totals and usage are equal", func(t *testing.T) {
a := &disk.UsageStat{Total: 100 * 1024 * 1024 * 1024, Used: 42 * 1024 * 1024 * 1024}
b := &disk.UsageStat{Total: 100 * 1024 * 1024 * 1024, Used: 42 * 1024 * 1024 * 1024}
assert.True(t, hasSameDiskUsage(a, b))
})
t.Run("returns true within tolerance", func(t *testing.T) {
a := &disk.UsageStat{Total: 100 * 1024 * 1024 * 1024, Used: 42 * 1024 * 1024 * 1024}
b := &disk.UsageStat{
Total: a.Total + toleranceBytes - 1,
Used: a.Used - toleranceBytes + 1,
}
assert.True(t, hasSameDiskUsage(a, b))
})
t.Run("returns false when total exceeds tolerance", func(t *testing.T) {
a := &disk.UsageStat{Total: 100 * 1024 * 1024 * 1024, Used: 42 * 1024 * 1024 * 1024}
b := &disk.UsageStat{
Total: a.Total + toleranceBytes + 1,
Used: a.Used,
}
assert.False(t, hasSameDiskUsage(a, b))
})
t.Run("returns false for nil or zero total", func(t *testing.T) {
assert.False(t, hasSameDiskUsage(nil, &disk.UsageStat{Total: 1, Used: 1}))
assert.False(t, hasSameDiskUsage(&disk.UsageStat{Total: 1, Used: 1}, nil))
assert.False(t, hasSameDiskUsage(&disk.UsageStat{Total: 0, Used: 0}, &disk.UsageStat{Total: 1, Used: 1}))
})
}
func TestInitializeDiskIoStatsResetsTrackedDevices(t *testing.T) {
agent := &Agent{
fsStats: map[string]*system.FsStats{
"sda": {},
"sdb": {},
},
fsNames: []string{"stale", "sda"},
}
agent.initializeDiskIoStats(map[string]disk.IOCountersStat{
"sda": {Name: "sda", ReadBytes: 10, WriteBytes: 20},
"sdb": {Name: "sdb", ReadBytes: 30, WriteBytes: 40},
})
assert.ElementsMatch(t, []string{"sda", "sdb"}, agent.fsNames)
assert.Len(t, agent.fsNames, 2)
assert.Equal(t, uint64(10), agent.fsStats["sda"].TotalRead)
assert.Equal(t, uint64(20), agent.fsStats["sda"].TotalWrite)
assert.False(t, agent.fsStats["sda"].Time.IsZero())
assert.False(t, agent.fsStats["sdb"].Time.IsZero())
agent.initializeDiskIoStats(map[string]disk.IOCountersStat{
"sdb": {Name: "sdb", ReadBytes: 50, WriteBytes: 60},
})
assert.Equal(t, []string{"sdb"}, agent.fsNames)
assert.Equal(t, uint64(50), agent.fsStats["sdb"].TotalRead)
assert.Equal(t, uint64(60), agent.fsStats["sdb"].TotalWrite)
}

View File

@@ -16,14 +16,11 @@ import (
"os"
"path"
"regexp"
"sort"
"strconv"
"strings"
"sync"
"time"
"github.com/henrygd/beszel/agent/deltatracker"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/container"
"github.com/blang/semver"
@@ -339,42 +336,15 @@ func validateCpuPercentage(cpuPct float64, containerName string) error {
// updateContainerStatsValues updates the final stats values
func updateContainerStatsValues(stats *container.Stats, cpuPct float64, usedMemory uint64, sent_delta, recv_delta uint64, readTime time.Time) {
stats.Cpu = utils.TwoDecimals(cpuPct)
stats.Mem = utils.BytesToMegabytes(float64(usedMemory))
stats.Cpu = twoDecimals(cpuPct)
stats.Mem = bytesToMegabytes(float64(usedMemory))
stats.Bandwidth = [2]uint64{sent_delta, recv_delta}
// TODO(0.19+): stop populating NetworkSent/NetworkRecv (deprecated in 0.18.3)
stats.NetworkSent = utils.BytesToMegabytes(float64(sent_delta))
stats.NetworkRecv = utils.BytesToMegabytes(float64(recv_delta))
stats.NetworkSent = bytesToMegabytes(float64(sent_delta))
stats.NetworkRecv = bytesToMegabytes(float64(recv_delta))
stats.PrevReadTime = readTime
}
// convertContainerPortsToString formats the ports of a container into a sorted, deduplicated string.
// ctr.Ports is nilled out after processing so the slice is not accidentally reused.
func convertContainerPortsToString(ctr *container.ApiInfo) string {
if len(ctr.Ports) == 0 {
return ""
}
sort.Slice(ctr.Ports, func(i, j int) bool {
return ctr.Ports[i].PublicPort < ctr.Ports[j].PublicPort
})
var builder strings.Builder
seenPorts := make(map[uint16]struct{})
for _, p := range ctr.Ports {
_, ok := seenPorts[p.PublicPort]
if p.PublicPort == 0 || ok {
continue
}
seenPorts[p.PublicPort] = struct{}{}
if builder.Len() > 0 {
builder.WriteString(", ")
}
builder.WriteString(strconv.Itoa(int(p.PublicPort)))
}
// clear ports slice so it doesn't get reused and blend into next response
ctr.Ports = nil
return builder.String()
}
func parseDockerStatus(status string) (string, container.DockerHealth) {
trimmed := strings.TrimSpace(status)
if trimmed == "" {
@@ -432,21 +402,9 @@ func (dm *dockerManager) updateContainerStats(ctr *container.ApiInfo, cacheTimeM
stats.Id = ctr.IdShort
statusText, health := parseDockerStatus(ctr.Status)
// Use Health.Status if it's available (Docker API 1.52+; Podman TBD - https://github.com/containers/podman/issues/27786)
if ctr.Health.Status != "" {
if h, ok := container.DockerHealthStrings[ctr.Health.Status]; ok {
health = h
}
}
stats.Status = statusText
stats.Health = health
if len(ctr.Ports) > 0 {
stats.Ports = convertContainerPortsToString(ctr)
}
// reset current stats
stats.Cpu = 0
stats.Mem = 0
@@ -529,7 +487,7 @@ func (dm *dockerManager) deleteContainerStatsSync(id string) {
// Creates a new http client for Docker or Podman API
func newDockerManager() *dockerManager {
dockerHost, exists := utils.GetEnv("DOCKER_HOST")
dockerHost, exists := GetEnv("DOCKER_HOST")
if exists {
// return nil if set to empty string
if dockerHost == "" {
@@ -565,7 +523,7 @@ func newDockerManager() *dockerManager {
// configurable timeout
timeout := time.Millisecond * time.Duration(dockerTimeoutMs)
if t, set := utils.GetEnv("DOCKER_TIMEOUT"); set {
if t, set := GetEnv("DOCKER_TIMEOUT"); set {
timeout, err = time.ParseDuration(t)
if err != nil {
slog.Error(err.Error())
@@ -582,7 +540,7 @@ func newDockerManager() *dockerManager {
// Read container exclusion patterns from environment variable
var excludeContainers []string
if excludeStr, set := utils.GetEnv("EXCLUDE_CONTAINERS"); set && excludeStr != "" {
if excludeStr, set := GetEnv("EXCLUDE_CONTAINERS"); set && excludeStr != "" {
parts := strings.SplitSeq(excludeStr, ",")
for part := range parts {
trimmed := strings.TrimSpace(part)

View File

@@ -18,7 +18,6 @@ import (
"time"
"github.com/henrygd/beszel/agent/deltatracker"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/container"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
@@ -299,6 +298,48 @@ func TestUpdateContainerStatsValues(t *testing.T) {
assert.Equal(t, testTime, stats.PrevReadTime)
}
func TestTwoDecimals(t *testing.T) {
tests := []struct {
name string
input float64
expected float64
}{
{"round down", 1.234, 1.23},
{"round half up", 1.235, 1.24}, // math.Round rounds half up
{"no rounding needed", 1.23, 1.23},
{"negative number", -1.235, -1.24}, // math.Round rounds half up (more negative)
{"zero", 0.0, 0.0},
{"large number", 123.456, 123.46}, // rounds 5 up
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result := twoDecimals(tt.input)
assert.Equal(t, tt.expected, result)
})
}
}
func TestBytesToMegabytes(t *testing.T) {
tests := []struct {
name string
input float64
expected float64
}{
{"1 MB", 1048576, 1.0},
{"512 KB", 524288, 0.5},
{"zero", 0, 0},
{"large value", 1073741824, 1024}, // 1 GB = 1024 MB
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result := bytesToMegabytes(tt.input)
assert.Equal(t, tt.expected, result)
})
}
}
func TestInitializeCpuTracking(t *testing.T) {
dm := &dockerManager{
lastCpuContainer: make(map[uint16]map[string]uint64),
@@ -864,11 +905,11 @@ func TestContainerStatsEndToEndWithRealData(t *testing.T) {
updateContainerStatsValues(testStats, cpuPct, usedMemory, 1000000, 500000, testTime)
assert.Equal(t, cpuPct, testStats.Cpu)
assert.Equal(t, utils.BytesToMegabytes(float64(usedMemory)), testStats.Mem)
assert.Equal(t, bytesToMegabytes(float64(usedMemory)), testStats.Mem)
assert.Equal(t, [2]uint64{1000000, 500000}, testStats.Bandwidth)
// Deprecated fields still populated for backward compatibility with older hubs
assert.Equal(t, utils.BytesToMegabytes(1000000), testStats.NetworkSent)
assert.Equal(t, utils.BytesToMegabytes(500000), testStats.NetworkRecv)
assert.Equal(t, bytesToMegabytes(1000000), testStats.NetworkSent)
assert.Equal(t, bytesToMegabytes(500000), testStats.NetworkRecv)
assert.Equal(t, testTime, testStats.PrevReadTime)
}
@@ -1149,13 +1190,13 @@ func TestConstantsAndUtilityFunctions(t *testing.T) {
assert.Equal(t, 5*1024*1024, maxTotalLogSize) // 5MB
// Test utility functions
assert.Equal(t, 1.5, utils.TwoDecimals(1.499))
assert.Equal(t, 1.5, utils.TwoDecimals(1.5))
assert.Equal(t, 1.5, utils.TwoDecimals(1.501))
assert.Equal(t, 1.5, twoDecimals(1.499))
assert.Equal(t, 1.5, twoDecimals(1.5))
assert.Equal(t, 1.5, twoDecimals(1.501))
assert.Equal(t, 1.0, utils.BytesToMegabytes(1048576)) // 1 MB
assert.Equal(t, 0.5, utils.BytesToMegabytes(524288)) // 512 KB
assert.Equal(t, 0.0, utils.BytesToMegabytes(0))
assert.Equal(t, 1.0, bytesToMegabytes(1048576)) // 1 MB
assert.Equal(t, 0.5, bytesToMegabytes(524288)) // 512 KB
assert.Equal(t, 0.0, bytesToMegabytes(0))
}
func TestDecodeDockerLogStream(t *testing.T) {
@@ -1455,109 +1496,3 @@ func TestAnsiEscapePattern(t *testing.T) {
})
}
}
func TestConvertContainerPortsToString(t *testing.T) {
type port = struct {
PrivatePort uint16
PublicPort uint16
Type string
}
tests := []struct {
name string
ports []port
expected string
}{
{
name: "empty ports",
ports: nil,
expected: "",
},
{
name: "single port public==private",
ports: []port{
{PublicPort: 80, PrivatePort: 80},
},
expected: "80",
},
{
name: "single port public!=private",
ports: []port{
{PublicPort: 443, PrivatePort: 2019},
},
// expected: "443:2019",
expected: "443",
},
{
name: "zero PublicPort is skipped",
ports: []port{
{PublicPort: 0, PrivatePort: 8080},
{PublicPort: 80, PrivatePort: 80},
},
expected: "80",
},
{
name: "ports sorted ascending by PublicPort",
ports: []port{
{PublicPort: 443, PrivatePort: 443},
{PublicPort: 80, PrivatePort: 80},
{PublicPort: 8080, PrivatePort: 8080},
},
expected: "80, 443, 8080",
},
{
name: "same PublicPort sorted by PrivatePort",
ports: []port{
{PublicPort: 443, PrivatePort: 9000},
{PublicPort: 443, PrivatePort: 2019},
},
// expected: "443:2019,443:9000",
expected: "443",
},
{
name: "duplicates are deduplicated",
ports: []port{
{PublicPort: 80, PrivatePort: 80},
{PublicPort: 80, PrivatePort: 80},
{PublicPort: 443, PrivatePort: 2019},
{PublicPort: 443, PrivatePort: 2019},
},
// expected: "80,443:2019",
expected: "80, 443",
},
{
name: "mixed zero and non-zero ports",
ports: []port{
{PublicPort: 0, PrivatePort: 5432},
{PublicPort: 443, PrivatePort: 2019},
{PublicPort: 80, PrivatePort: 80},
{PublicPort: 0, PrivatePort: 9000},
},
// expected: "80,443:2019",
expected: "80, 443",
},
{
name: "ports slice is nilled after call",
ports: []port{
{PublicPort: 8080, PrivatePort: 8080},
},
expected: "8080",
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
ctr := &container.ApiInfo{}
for _, p := range tt.ports {
ctr.Ports = append(ctr.Ports, struct {
// PrivatePort uint16
PublicPort uint16
// Type string
}{PublicPort: p.PublicPort})
}
result := convertContainerPortsToString(ctr)
assert.Equal(t, tt.expected, result)
// Ports slice must be cleared to prevent bleed-over into the next response
assert.Nil(t, ctr.Ports, "ctr.Ports should be nil after formatContainerPorts")
})
}
}

View File

@@ -8,7 +8,6 @@ import (
"strconv"
"strings"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/smart"
)
@@ -142,9 +141,9 @@ func readEmmcHealth(blockName string) (emmcHealth, bool) {
out.lifeA = lifeA
out.lifeB = lifeB
out.model = utils.ReadStringFile(filepath.Join(deviceDir, "name"))
out.serial = utils.ReadStringFile(filepath.Join(deviceDir, "serial"))
out.revision = utils.ReadStringFile(filepath.Join(deviceDir, "prv"))
out.model = readStringFile(filepath.Join(deviceDir, "name"))
out.serial = readStringFile(filepath.Join(deviceDir, "serial"))
out.revision = readStringFile(filepath.Join(deviceDir, "prv"))
if capBytes, ok := readBlockCapacityBytes(blockName); ok {
out.capacity = capBytes
@@ -154,7 +153,7 @@ func readEmmcHealth(blockName string) (emmcHealth, bool) {
}
func readLifeTime(deviceDir string) (uint8, uint8, bool) {
if content, ok := utils.ReadStringFileOK(filepath.Join(deviceDir, "life_time")); ok {
if content, ok := readStringFileOK(filepath.Join(deviceDir, "life_time")); ok {
a, b, ok := parseHexBytePair(content)
return a, b, ok
}
@@ -171,7 +170,7 @@ func readBlockCapacityBytes(blockName string) (uint64, bool) {
sizePath := filepath.Join(emmcSysfsRoot, "class", "block", blockName, "size")
lbsPath := filepath.Join(emmcSysfsRoot, "class", "block", blockName, "queue", "logical_block_size")
sizeStr, ok := utils.ReadStringFileOK(sizePath)
sizeStr, ok := readStringFileOK(sizePath)
if !ok {
return 0, false
}
@@ -180,7 +179,7 @@ func readBlockCapacityBytes(blockName string) (uint64, bool) {
return 0, false
}
lbsStr, ok := utils.ReadStringFileOK(lbsPath)
lbsStr, ok := readStringFileOK(lbsPath)
logicalBlockSize := uint64(512)
if ok {
if parsed, err := strconv.ParseUint(lbsStr, 10, 64); err == nil && parsed > 0 {
@@ -192,7 +191,7 @@ func readBlockCapacityBytes(blockName string) (uint64, bool) {
}
func readHexByteFile(path string) (uint8, bool) {
content, ok := utils.ReadStringFileOK(path)
content, ok := readStringFileOK(path)
if !ok {
return 0, false
}
@@ -200,6 +199,19 @@ func readHexByteFile(path string) (uint8, bool) {
return b, ok
}
func readStringFile(path string) string {
content, _ := readStringFileOK(path)
return content
}
func readStringFileOK(path string) (string, bool) {
b, err := os.ReadFile(path)
if err != nil {
return "", false
}
return strings.TrimSpace(string(b)), true
}
func hasEmmcHealthFiles(deviceDir string) bool {
entries, err := os.ReadDir(deviceDir)
if err != nil {

View File

@@ -15,7 +15,6 @@ import (
"sync"
"time"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/system"
)
@@ -292,8 +291,8 @@ func (gm *GPUManager) parseAmdData(output []byte) bool {
}
gpu := gm.GpuDataMap[id]
gpu.Temperature, _ = strconv.ParseFloat(v.Temperature, 64)
gpu.MemoryUsed = utils.BytesToMegabytes(memoryUsage)
gpu.MemoryTotal = utils.BytesToMegabytes(totalMemory)
gpu.MemoryUsed = bytesToMegabytes(memoryUsage)
gpu.MemoryTotal = bytesToMegabytes(totalMemory)
gpu.Usage += usage
gpu.Power += power
gpu.Count++
@@ -367,16 +366,16 @@ func (gm *GPUManager) calculateGPUAverage(id string, gpu *system.GPUData, cacheK
gpuAvg := *gpu
deltaUsage, deltaPower, deltaPowerPkg := gm.calculateDeltas(gpu, lastSnapshot)
gpuAvg.Power = utils.TwoDecimals(deltaPower / float64(deltaCount))
gpuAvg.Power = twoDecimals(deltaPower / float64(deltaCount))
if gpu.Engines != nil {
// make fresh map for averaged engine metrics to avoid mutating
// the accumulator map stored in gm.GpuDataMap
gpuAvg.Engines = make(map[string]float64, len(gpu.Engines))
gpuAvg.Usage = gm.calculateIntelGPUUsage(&gpuAvg, gpu, lastSnapshot, deltaCount)
gpuAvg.PowerPkg = utils.TwoDecimals(deltaPowerPkg / float64(deltaCount))
gpuAvg.PowerPkg = twoDecimals(deltaPowerPkg / float64(deltaCount))
} else {
gpuAvg.Usage = utils.TwoDecimals(deltaUsage / float64(deltaCount))
gpuAvg.Usage = twoDecimals(deltaUsage / float64(deltaCount))
}
gm.lastAvgData[id] = gpuAvg
@@ -411,17 +410,17 @@ func (gm *GPUManager) calculateIntelGPUUsage(gpuAvg, gpu *system.GPUData, lastSn
} else {
deltaEngine = engine
}
gpuAvg.Engines[name] = utils.TwoDecimals(deltaEngine / float64(deltaCount))
gpuAvg.Engines[name] = twoDecimals(deltaEngine / float64(deltaCount))
maxEngineUsage = max(maxEngineUsage, deltaEngine/float64(deltaCount))
}
return utils.TwoDecimals(maxEngineUsage)
return twoDecimals(maxEngineUsage)
}
// updateInstantaneousValues updates values that should reflect current state, not averages
func (gm *GPUManager) updateInstantaneousValues(gpuAvg *system.GPUData, gpu *system.GPUData) {
gpuAvg.Temperature = utils.TwoDecimals(gpu.Temperature)
gpuAvg.MemoryUsed = utils.TwoDecimals(gpu.MemoryUsed)
gpuAvg.MemoryTotal = utils.TwoDecimals(gpu.MemoryTotal)
gpuAvg.Temperature = twoDecimals(gpu.Temperature)
gpuAvg.MemoryUsed = twoDecimals(gpu.MemoryUsed)
gpuAvg.MemoryTotal = twoDecimals(gpu.MemoryTotal)
}
// storeSnapshot saves the current GPU state for this cache key
@@ -688,7 +687,7 @@ func (gm *GPUManager) resolveLegacyCollectorPriority(caps gpuCapabilities) []col
priorities := make([]collectorSource, 0, 4)
if caps.hasNvidiaSmi && !caps.hasTegrastats {
if nvml, _ := utils.GetEnv("NVML"); nvml == "true" {
if nvml, _ := GetEnv("NVML"); nvml == "true" {
priorities = append(priorities, collectorSourceNVML, collectorSourceNvidiaSMI)
} else {
priorities = append(priorities, collectorSourceNvidiaSMI)
@@ -696,7 +695,7 @@ func (gm *GPUManager) resolveLegacyCollectorPriority(caps gpuCapabilities) []col
}
if caps.hasRocmSmi {
if val, _ := utils.GetEnv("AMD_SYSFS"); val == "true" {
if val, _ := GetEnv("AMD_SYSFS"); val == "true" {
priorities = append(priorities, collectorSourceAmdSysfs)
} else {
priorities = append(priorities, collectorSourceRocmSMI)
@@ -729,7 +728,7 @@ func (gm *GPUManager) resolveLegacyCollectorPriority(caps gpuCapabilities) []col
// NewGPUManager creates and initializes a new GPUManager
func NewGPUManager() (*GPUManager, error) {
if skipGPU, _ := utils.GetEnv("SKIP_GPU"); skipGPU == "true" {
if skipGPU, _ := GetEnv("SKIP_GPU"); skipGPU == "true" {
return nil, nil
}
var gm GPUManager
@@ -746,7 +745,7 @@ func NewGPUManager() (*GPUManager, error) {
}
// if GPU_COLLECTOR is set, start user-defined collectors.
if collectorConfig, ok := utils.GetEnv("GPU_COLLECTOR"); ok && strings.TrimSpace(collectorConfig) != "" {
if collectorConfig, ok := GetEnv("GPU_COLLECTOR"); ok && strings.TrimSpace(collectorConfig) != "" {
priorities := parseCollectorPriority(collectorConfig)
if gm.startCollectorsByPriority(priorities, caps) == 0 {
return nil, fmt.Errorf("no configured GPU collectors are available")

View File

@@ -13,7 +13,6 @@ import (
"sync"
"time"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/system"
)
@@ -33,8 +32,8 @@ func (gm *GPUManager) hasAmdSysfs() bool {
return false
}
for _, vendorPath := range cards {
vendor, err := utils.ReadStringFileLimited(vendorPath, 64)
if err == nil && vendor == "0x1002" {
vendor, err := os.ReadFile(vendorPath)
if err == nil && strings.TrimSpace(string(vendor)) == "0x1002" {
return true
}
}
@@ -88,11 +87,12 @@ func (gm *GPUManager) collectAmdStats() error {
// isAmdGpu checks whether a DRM card path belongs to AMD vendor ID 0x1002.
func isAmdGpu(cardPath string) bool {
vendor, err := utils.ReadStringFileLimited(filepath.Join(cardPath, "device/vendor"), 64)
vendorPath := filepath.Join(cardPath, "device/vendor")
vendor, err := os.ReadFile(vendorPath)
if err != nil {
return false
}
return vendor == "0x1002"
return strings.TrimSpace(string(vendor)) == "0x1002"
}
// updateAmdGpuData reads GPU metrics from sysfs and updates the GPU data map.
@@ -144,8 +144,8 @@ func (gm *GPUManager) updateAmdGpuData(cardPath string) bool {
if usageErr == nil {
gpu.Usage += usage
}
gpu.MemoryUsed = utils.BytesToMegabytes(memUsed)
gpu.MemoryTotal = utils.BytesToMegabytes(memTotal)
gpu.MemoryUsed = bytesToMegabytes(memUsed)
gpu.MemoryTotal = bytesToMegabytes(memTotal)
gpu.Temperature = temp
gpu.Power += power
gpu.Count++
@@ -154,11 +154,11 @@ func (gm *GPUManager) updateAmdGpuData(cardPath string) bool {
// readSysfsFloat reads and parses a numeric value from a sysfs file.
func readSysfsFloat(path string) (float64, error) {
val, err := utils.ReadStringFileLimited(path, 64)
val, err := os.ReadFile(path)
if err != nil {
return 0, err
}
return strconv.ParseFloat(val, 64)
return strconv.ParseFloat(strings.TrimSpace(string(val)), 64)
}
// normalizeHexID normalizes hex IDs by trimming spaces, lowercasing, and dropping 0x.
@@ -273,16 +273,16 @@ func cacheMissingAmdgpuName(deviceID, revisionID string) {
// Falls back to showing the raw device ID if not found in the lookup table.
func getAmdGpuName(devicePath string) string {
// Try product_name first (works for some enterprise GPUs)
if prod, err := utils.ReadStringFileLimited(filepath.Join(devicePath, "product_name"), 128); err == nil {
return prod
if prod, err := os.ReadFile(filepath.Join(devicePath, "product_name")); err == nil {
return strings.TrimSpace(string(prod))
}
// Read PCI device ID and look it up
if deviceID, err := utils.ReadStringFileLimited(filepath.Join(devicePath, "device"), 64); err == nil {
id := normalizeHexID(deviceID)
if deviceID, err := os.ReadFile(filepath.Join(devicePath, "device")); err == nil {
id := normalizeHexID(string(deviceID))
revision := ""
if rev, revErr := utils.ReadStringFileLimited(filepath.Join(devicePath, "revision"), 64); revErr == nil {
revision = normalizeHexID(rev)
if revBytes, revErr := os.ReadFile(filepath.Join(devicePath, "revision")); revErr == nil {
revision = normalizeHexID(string(revBytes))
}
if name, found, done := getCachedAmdgpuName(id, revision); found {

View File

@@ -7,7 +7,6 @@ import (
"path/filepath"
"testing"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/system"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
@@ -129,14 +128,14 @@ func TestUpdateAmdGpuDataWithFakeSysfs(t *testing.T) {
{
name: "sums vram and gtt when gtt is present",
writeGTT: true,
wantMemoryUsed: utils.BytesToMegabytes(1073741824 + 536870912),
wantMemoryTotal: utils.BytesToMegabytes(2147483648 + 4294967296),
wantMemoryUsed: bytesToMegabytes(1073741824 + 536870912),
wantMemoryTotal: bytesToMegabytes(2147483648 + 4294967296),
},
{
name: "falls back to vram when gtt is missing",
writeGTT: false,
wantMemoryUsed: utils.BytesToMegabytes(1073741824),
wantMemoryTotal: utils.BytesToMegabytes(2147483648),
wantMemoryUsed: bytesToMegabytes(1073741824),
wantMemoryTotal: bytesToMegabytes(2147483648),
},
}

View File

@@ -7,7 +7,6 @@ import (
"strconv"
"strings"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/system"
)
@@ -53,7 +52,7 @@ func (gm *GPUManager) updateIntelFromStats(sample *intelGpuStats) bool {
func (gm *GPUManager) collectIntelStats() (err error) {
// Build command arguments, optionally selecting a device via -d
args := []string{"-s", intelGpuStatsInterval, "-l"}
if dev, ok := utils.GetEnv("INTEL_GPU_DEVICE"); ok && dev != "" {
if dev, ok := GetEnv("INTEL_GPU_DEVICE"); ok && dev != "" {
args = append(args, "-d", dev)
}
cmd := exec.Command(intelGpuStatsCmd, args...)

View File

@@ -9,7 +9,6 @@ import (
"strings"
"time"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/system"
)
@@ -81,10 +80,10 @@ func (gm *GPUManager) updateNvtopSnapshots(snapshots []nvtopSnapshot) bool {
gpu.Temperature = parseNvtopNumber(*sample.Temp)
}
if sample.MemUsed != nil {
gpu.MemoryUsed = utils.BytesToMegabytes(parseNvtopNumber(*sample.MemUsed))
gpu.MemoryUsed = bytesToMegabytes(parseNvtopNumber(*sample.MemUsed))
}
if sample.MemTotal != nil {
gpu.MemoryTotal = utils.BytesToMegabytes(parseNvtopNumber(*sample.MemTotal))
gpu.MemoryTotal = bytesToMegabytes(parseNvtopNumber(*sample.MemTotal))
}
if sample.GpuUtil != nil {
gpu.Usage += parseNvtopNumber(*sample.GpuUtil)

View File

@@ -10,7 +10,6 @@ import (
"testing"
"time"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/system"
"github.com/stretchr/testify/assert"
@@ -266,8 +265,8 @@ func TestParseNvtopData(t *testing.T) {
assert.Equal(t, 48.0, g0.Temperature)
assert.Equal(t, 5.0, g0.Usage)
assert.Equal(t, 13.0, g0.Power)
assert.Equal(t, utils.BytesToMegabytes(349372416), g0.MemoryUsed)
assert.Equal(t, utils.BytesToMegabytes(4294967296), g0.MemoryTotal)
assert.Equal(t, bytesToMegabytes(349372416), g0.MemoryUsed)
assert.Equal(t, bytesToMegabytes(4294967296), g0.MemoryTotal)
assert.Equal(t, 1.0, g0.Count)
g1, ok := gm.GpuDataMap["n1"]
@@ -276,8 +275,8 @@ func TestParseNvtopData(t *testing.T) {
assert.Equal(t, 48.0, g1.Temperature)
assert.Equal(t, 12.0, g1.Usage)
assert.Equal(t, 9.0, g1.Power)
assert.Equal(t, utils.BytesToMegabytes(1213784064), g1.MemoryUsed)
assert.Equal(t, utils.BytesToMegabytes(16929173504), g1.MemoryTotal)
assert.Equal(t, bytesToMegabytes(1213784064), g1.MemoryUsed)
assert.Equal(t, bytesToMegabytes(16929173504), g1.MemoryTotal)
assert.Equal(t, 1.0, g1.Count)
}

View File

@@ -1,233 +0,0 @@
//go:build linux
package agent
import (
"fmt"
"os"
"path/filepath"
"strconv"
"strings"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/smart"
)
// mdraidSysfsRoot is a test hook; production value is "/sys".
var mdraidSysfsRoot = "/sys"
type mdraidHealth struct {
level string
arrayState string
degraded uint64
raidDisks uint64
syncAction string
syncCompleted string
syncSpeed string
mismatchCnt uint64
capacity uint64
}
// scanMdraidDevices discovers Linux md arrays exposed in sysfs.
func scanMdraidDevices() []*DeviceInfo {
blockDir := filepath.Join(mdraidSysfsRoot, "block")
entries, err := os.ReadDir(blockDir)
if err != nil {
return nil
}
devices := make([]*DeviceInfo, 0, 2)
for _, ent := range entries {
name := ent.Name()
if !isMdraidBlockName(name) {
continue
}
mdDir := filepath.Join(blockDir, name, "md")
if !utils.FileExists(filepath.Join(mdDir, "array_state")) {
continue
}
devPath := filepath.Join("/dev", name)
devices = append(devices, &DeviceInfo{
Name: devPath,
Type: "mdraid",
InfoName: devPath + " [mdraid]",
Protocol: "MD",
})
}
return devices
}
// collectMdraidHealth reads mdraid health and stores it in SmartDataMap.
func (sm *SmartManager) collectMdraidHealth(deviceInfo *DeviceInfo) (bool, error) {
if deviceInfo == nil || deviceInfo.Name == "" {
return false, nil
}
base := filepath.Base(deviceInfo.Name)
if !isMdraidBlockName(base) && !strings.EqualFold(deviceInfo.Type, "mdraid") {
return false, nil
}
health, ok := readMdraidHealth(base)
if !ok {
return false, nil
}
deviceInfo.Type = "mdraid"
key := fmt.Sprintf("mdraid:%s", base)
status := mdraidSmartStatus(health)
attrs := make([]*smart.SmartAttribute, 0, 10)
if health.arrayState != "" {
attrs = append(attrs, &smart.SmartAttribute{Name: "ArrayState", RawString: health.arrayState})
}
if health.level != "" {
attrs = append(attrs, &smart.SmartAttribute{Name: "RaidLevel", RawString: health.level})
}
if health.raidDisks > 0 {
attrs = append(attrs, &smart.SmartAttribute{Name: "RaidDisks", RawValue: health.raidDisks})
}
if health.degraded > 0 {
attrs = append(attrs, &smart.SmartAttribute{Name: "Degraded", RawValue: health.degraded})
}
if health.syncAction != "" {
attrs = append(attrs, &smart.SmartAttribute{Name: "SyncAction", RawString: health.syncAction})
}
if health.syncCompleted != "" {
attrs = append(attrs, &smart.SmartAttribute{Name: "SyncCompleted", RawString: health.syncCompleted})
}
if health.syncSpeed != "" {
attrs = append(attrs, &smart.SmartAttribute{Name: "SyncSpeed", RawString: health.syncSpeed})
}
if health.mismatchCnt > 0 {
attrs = append(attrs, &smart.SmartAttribute{Name: "MismatchCount", RawValue: health.mismatchCnt})
}
sm.Lock()
defer sm.Unlock()
if _, exists := sm.SmartDataMap[key]; !exists {
sm.SmartDataMap[key] = &smart.SmartData{}
}
data := sm.SmartDataMap[key]
data.ModelName = "Linux MD RAID"
if health.level != "" {
data.ModelName = "Linux MD RAID (" + health.level + ")"
}
data.Capacity = health.capacity
data.SmartStatus = status
data.DiskName = filepath.Join("/dev", base)
data.DiskType = "mdraid"
data.Attributes = attrs
return true, nil
}
// readMdraidHealth reads md array health fields from sysfs.
func readMdraidHealth(blockName string) (mdraidHealth, bool) {
var out mdraidHealth
if !isMdraidBlockName(blockName) {
return out, false
}
mdDir := filepath.Join(mdraidSysfsRoot, "block", blockName, "md")
arrayState, okState := utils.ReadStringFileOK(filepath.Join(mdDir, "array_state"))
if !okState {
return out, false
}
out.arrayState = arrayState
out.level = utils.ReadStringFile(filepath.Join(mdDir, "level"))
out.syncAction = utils.ReadStringFile(filepath.Join(mdDir, "sync_action"))
out.syncCompleted = utils.ReadStringFile(filepath.Join(mdDir, "sync_completed"))
out.syncSpeed = utils.ReadStringFile(filepath.Join(mdDir, "sync_speed"))
if val, ok := utils.ReadUintFile(filepath.Join(mdDir, "raid_disks")); ok {
out.raidDisks = val
}
if val, ok := utils.ReadUintFile(filepath.Join(mdDir, "degraded")); ok {
out.degraded = val
}
if val, ok := utils.ReadUintFile(filepath.Join(mdDir, "mismatch_cnt")); ok {
out.mismatchCnt = val
}
if capBytes, ok := readMdraidBlockCapacityBytes(blockName, mdraidSysfsRoot); ok {
out.capacity = capBytes
}
return out, true
}
// mdraidSmartStatus maps md state/sync signals to a SMART-like status.
func mdraidSmartStatus(health mdraidHealth) string {
state := strings.ToLower(strings.TrimSpace(health.arrayState))
switch state {
case "inactive", "faulty", "broken", "stopped":
return "FAILED"
}
// During rebuild/recovery, arrays are often temporarily degraded; report as
// warning instead of hard failure while synchronization is in progress.
syncAction := strings.ToLower(strings.TrimSpace(health.syncAction))
switch syncAction {
case "resync", "recover", "reshape":
return "WARNING"
}
if health.degraded > 0 {
return "FAILED"
}
switch syncAction {
case "check", "repair":
return "WARNING"
}
switch state {
case "clean", "active", "active-idle", "write-pending", "read-auto", "readonly":
return "PASSED"
}
return "UNKNOWN"
}
// isMdraidBlockName matches /dev/mdN-style block device names.
func isMdraidBlockName(name string) bool {
if !strings.HasPrefix(name, "md") {
return false
}
suffix := strings.TrimPrefix(name, "md")
if suffix == "" {
return false
}
for _, c := range suffix {
if c < '0' || c > '9' {
return false
}
}
return true
}
// readMdraidBlockCapacityBytes converts block size metadata into bytes.
func readMdraidBlockCapacityBytes(blockName, root string) (uint64, bool) {
sizePath := filepath.Join(root, "block", blockName, "size")
lbsPath := filepath.Join(root, "block", blockName, "queue", "logical_block_size")
sizeStr, ok := utils.ReadStringFileOK(sizePath)
if !ok {
return 0, false
}
sectors, err := strconv.ParseUint(sizeStr, 10, 64)
if err != nil || sectors == 0 {
return 0, false
}
logicalBlockSize := uint64(512)
if lbsStr, ok := utils.ReadStringFileOK(lbsPath); ok {
if parsed, err := strconv.ParseUint(lbsStr, 10, 64); err == nil && parsed > 0 {
logicalBlockSize = parsed
}
}
return sectors * logicalBlockSize, true
}

View File

@@ -1,103 +0,0 @@
//go:build linux
package agent
import (
"os"
"path/filepath"
"testing"
"github.com/henrygd/beszel/internal/entities/smart"
)
func TestMdraidMockSysfsScanAndCollect(t *testing.T) {
tmp := t.TempDir()
prev := mdraidSysfsRoot
mdraidSysfsRoot = tmp
t.Cleanup(func() { mdraidSysfsRoot = prev })
mdDir := filepath.Join(tmp, "block", "md0", "md")
queueDir := filepath.Join(tmp, "block", "md0", "queue")
if err := os.MkdirAll(mdDir, 0o755); err != nil {
t.Fatal(err)
}
if err := os.MkdirAll(queueDir, 0o755); err != nil {
t.Fatal(err)
}
write := func(path, content string) {
t.Helper()
if err := os.WriteFile(path, []byte(content), 0o644); err != nil {
t.Fatal(err)
}
}
write(filepath.Join(mdDir, "array_state"), "active\n")
write(filepath.Join(mdDir, "level"), "raid1\n")
write(filepath.Join(mdDir, "raid_disks"), "2\n")
write(filepath.Join(mdDir, "degraded"), "0\n")
write(filepath.Join(mdDir, "sync_action"), "resync\n")
write(filepath.Join(mdDir, "sync_completed"), "10%\n")
write(filepath.Join(mdDir, "sync_speed"), "100M\n")
write(filepath.Join(mdDir, "mismatch_cnt"), "0\n")
write(filepath.Join(queueDir, "logical_block_size"), "512\n")
write(filepath.Join(tmp, "block", "md0", "size"), "2048\n")
devs := scanMdraidDevices()
if len(devs) != 1 {
t.Fatalf("scanMdraidDevices() = %d devices, want 1", len(devs))
}
if devs[0].Name != "/dev/md0" || devs[0].Type != "mdraid" {
t.Fatalf("scanMdraidDevices()[0] = %+v, want Name=/dev/md0 Type=mdraid", devs[0])
}
sm := &SmartManager{SmartDataMap: map[string]*smart.SmartData{}}
ok, err := sm.collectMdraidHealth(devs[0])
if err != nil || !ok {
t.Fatalf("collectMdraidHealth() = (ok=%v, err=%v), want (true,nil)", ok, err)
}
if len(sm.SmartDataMap) != 1 {
t.Fatalf("SmartDataMap len=%d, want 1", len(sm.SmartDataMap))
}
var got *smart.SmartData
for _, v := range sm.SmartDataMap {
got = v
break
}
if got == nil {
t.Fatalf("SmartDataMap value nil")
}
if got.DiskType != "mdraid" || got.DiskName != "/dev/md0" {
t.Fatalf("disk fields = (type=%q name=%q), want (mdraid,/dev/md0)", got.DiskType, got.DiskName)
}
if got.SmartStatus != "WARNING" {
t.Fatalf("SmartStatus=%q, want WARNING", got.SmartStatus)
}
if got.ModelName == "" || got.Capacity == 0 {
t.Fatalf("identity fields = (model=%q cap=%d), want non-empty model and cap>0", got.ModelName, got.Capacity)
}
if len(got.Attributes) < 5 {
t.Fatalf("attributes len=%d, want >= 5", len(got.Attributes))
}
}
func TestMdraidSmartStatus(t *testing.T) {
if got := mdraidSmartStatus(mdraidHealth{arrayState: "inactive"}); got != "FAILED" {
t.Fatalf("mdraidSmartStatus(inactive) = %q, want FAILED", got)
}
if got := mdraidSmartStatus(mdraidHealth{arrayState: "active", degraded: 1, syncAction: "recover"}); got != "WARNING" {
t.Fatalf("mdraidSmartStatus(degraded+recover) = %q, want WARNING", got)
}
if got := mdraidSmartStatus(mdraidHealth{arrayState: "active", degraded: 1}); got != "FAILED" {
t.Fatalf("mdraidSmartStatus(degraded) = %q, want FAILED", got)
}
if got := mdraidSmartStatus(mdraidHealth{arrayState: "active", syncAction: "recover"}); got != "WARNING" {
t.Fatalf("mdraidSmartStatus(recover) = %q, want WARNING", got)
}
if got := mdraidSmartStatus(mdraidHealth{arrayState: "clean"}); got != "PASSED" {
t.Fatalf("mdraidSmartStatus(clean) = %q, want PASSED", got)
}
if got := mdraidSmartStatus(mdraidHealth{arrayState: "unknown"}); got != "UNKNOWN" {
t.Fatalf("mdraidSmartStatus(unknown) = %q, want UNKNOWN", got)
}
}

View File

@@ -1,11 +0,0 @@
//go:build !linux
package agent
func scanMdraidDevices() []*DeviceInfo {
return nil
}
func (sm *SmartManager) collectMdraidHealth(deviceInfo *DeviceInfo) (bool, error) {
return false, nil
}

View File

@@ -8,7 +8,6 @@ import (
"time"
"github.com/henrygd/beszel/agent/deltatracker"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/system"
psutilNet "github.com/shirou/gopsutil/v4/net"
)
@@ -95,7 +94,7 @@ func (a *Agent) initializeNetIoStats() {
a.netInterfaces = make(map[string]struct{}, 0)
// parse NICS env var for whitelist / blacklist
nicsEnvVal, nicsEnvExists := utils.GetEnv("NICS")
nicsEnvVal, nicsEnvExists := GetEnv("NICS")
var nicCfg *NicConfig
if nicsEnvExists {
nicCfg = newNicConfig(nicsEnvVal)
@@ -104,7 +103,10 @@ func (a *Agent) initializeNetIoStats() {
// get current network I/O stats and record valid interfaces
if netIO, err := psutilNet.IOCounters(true); err == nil {
for _, v := range netIO {
if skipNetworkInterface(v, nicCfg) {
if nicsEnvExists && !isValidNic(v.Name, nicCfg) {
continue
}
if a.skipNetworkInterface(v) {
continue
}
slog.Info("Detected network interface", "name", v.Name, "sent", v.BytesSent, "recv", v.BytesRecv)
@@ -213,8 +215,10 @@ func (a *Agent) applyNetworkTotals(
totalBytesSent, totalBytesRecv uint64,
bytesSentPerSecond, bytesRecvPerSecond uint64,
) {
if bytesSentPerSecond > 10_000_000_000 || bytesRecvPerSecond > 10_000_000_000 {
slog.Warn("Invalid net stats. Resetting.", "sent", bytesSentPerSecond, "recv", bytesRecvPerSecond)
networkSentPs := bytesToMegabytes(float64(bytesSentPerSecond))
networkRecvPs := bytesToMegabytes(float64(bytesRecvPerSecond))
if networkSentPs > 10_000 || networkRecvPs > 10_000 {
slog.Warn("Invalid net stats. Resetting.", "sent", networkSentPs, "recv", networkRecvPs)
for _, v := range netIO {
if _, exists := a.netInterfaces[v.Name]; !exists {
continue
@@ -224,29 +228,21 @@ func (a *Agent) applyNetworkTotals(
a.initializeNetIoStats()
delete(a.netIoStats, cacheTimeMs)
delete(a.netInterfaceDeltaTrackers, cacheTimeMs)
systemStats.NetworkSent = 0
systemStats.NetworkRecv = 0
systemStats.Bandwidth[0], systemStats.Bandwidth[1] = 0, 0
return
}
systemStats.NetworkSent = networkSentPs
systemStats.NetworkRecv = networkRecvPs
systemStats.Bandwidth[0], systemStats.Bandwidth[1] = bytesSentPerSecond, bytesRecvPerSecond
nis.BytesSent = totalBytesSent
nis.BytesRecv = totalBytesRecv
a.netIoStats[cacheTimeMs] = nis
}
// skipNetworkInterface returns true if the network interface should be ignored.
func skipNetworkInterface(v psutilNet.IOCountersStat, nicCfg *NicConfig) bool {
if nicCfg != nil {
if !isValidNic(v.Name, nicCfg) {
return true
}
// In whitelist mode, we honor explicit inclusion without auto-filtering.
if !nicCfg.isBlacklist {
return false
}
// In blacklist mode, still apply the auto-filter below.
}
func (a *Agent) skipNetworkInterface(v psutilNet.IOCountersStat) bool {
switch {
case strings.HasPrefix(v.Name, "lo"),
strings.HasPrefix(v.Name, "docker"),

View File

@@ -261,39 +261,6 @@ func TestNewNicConfig(t *testing.T) {
})
}
}
func TestSkipNetworkInterface(t *testing.T) {
tests := []struct {
name string
nic psutilNet.IOCountersStat
nicCfg *NicConfig
expectSkip bool
}{
{"loopback lo", psutilNet.IOCountersStat{Name: "lo", BytesSent: 100, BytesRecv: 100}, nil, true},
{"loopback lo0", psutilNet.IOCountersStat{Name: "lo0", BytesSent: 100, BytesRecv: 100}, nil, true},
{"docker prefix", psutilNet.IOCountersStat{Name: "docker0", BytesSent: 100, BytesRecv: 100}, nil, true},
{"br- prefix", psutilNet.IOCountersStat{Name: "br-lan", BytesSent: 100, BytesRecv: 100}, nil, true},
{"veth prefix", psutilNet.IOCountersStat{Name: "veth0abc", BytesSent: 100, BytesRecv: 100}, nil, true},
{"bond prefix", psutilNet.IOCountersStat{Name: "bond0", BytesSent: 100, BytesRecv: 100}, nil, true},
{"cali prefix", psutilNet.IOCountersStat{Name: "cali1234", BytesSent: 100, BytesRecv: 100}, nil, true},
{"zero BytesRecv", psutilNet.IOCountersStat{Name: "eth0", BytesSent: 100, BytesRecv: 0}, nil, true},
{"zero BytesSent", psutilNet.IOCountersStat{Name: "eth0", BytesSent: 0, BytesRecv: 100}, nil, true},
{"both zero", psutilNet.IOCountersStat{Name: "eth0", BytesSent: 0, BytesRecv: 0}, nil, true},
{"normal eth0", psutilNet.IOCountersStat{Name: "eth0", BytesSent: 100, BytesRecv: 200}, nil, false},
{"normal wlan0", psutilNet.IOCountersStat{Name: "wlan0", BytesSent: 1, BytesRecv: 1}, nil, false},
{"whitelist overrides skip (docker)", psutilNet.IOCountersStat{Name: "docker0", BytesSent: 100, BytesRecv: 100}, newNicConfig("docker0"), false},
{"whitelist overrides skip (lo)", psutilNet.IOCountersStat{Name: "lo", BytesSent: 100, BytesRecv: 100}, newNicConfig("lo"), false},
{"whitelist exclusion", psutilNet.IOCountersStat{Name: "eth1", BytesSent: 100, BytesRecv: 100}, newNicConfig("eth0"), true},
{"blacklist skip lo", psutilNet.IOCountersStat{Name: "lo", BytesSent: 100, BytesRecv: 100}, newNicConfig("-eth0"), true},
{"blacklist explicit eth0", psutilNet.IOCountersStat{Name: "eth0", BytesSent: 100, BytesRecv: 100}, newNicConfig("-eth0"), true},
{"blacklist allow eth1", psutilNet.IOCountersStat{Name: "eth1", BytesSent: 100, BytesRecv: 100}, newNicConfig("-eth0"), false},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
assert.Equal(t, tt.expectSkip, skipNetworkInterface(tt.nic, tt.nicCfg))
})
}
}
func TestEnsureNetworkInterfacesMap(t *testing.T) {
var a Agent
var stats system.Stats
@@ -416,6 +383,8 @@ func TestApplyNetworkTotals(t *testing.T) {
totalBytesSent uint64
totalBytesRecv uint64
expectReset bool
expectedNetworkSent float64
expectedNetworkRecv float64
expectedBandwidthSent uint64
expectedBandwidthRecv uint64
}{
@@ -426,6 +395,8 @@ func TestApplyNetworkTotals(t *testing.T) {
totalBytesSent: 10000000,
totalBytesRecv: 20000000,
expectReset: false,
expectedNetworkSent: 0.95, // ~1 MB/s rounded to 2 decimals
expectedNetworkRecv: 1.91, // ~2 MB/s rounded to 2 decimals
expectedBandwidthSent: 1000000,
expectedBandwidthRecv: 2000000,
},
@@ -453,6 +424,18 @@ func TestApplyNetworkTotals(t *testing.T) {
totalBytesRecv: 20000000,
expectReset: true,
},
{
name: "Valid network stats - at threshold boundary",
bytesSentPerSecond: 10485750000, // ~9999.99 MB/s (rounds to 9999.99)
bytesRecvPerSecond: 10485750000, // ~9999.99 MB/s (rounds to 9999.99)
totalBytesSent: 10000000,
totalBytesRecv: 20000000,
expectReset: false,
expectedNetworkSent: 9999.99,
expectedNetworkRecv: 9999.99,
expectedBandwidthSent: 10485750000,
expectedBandwidthRecv: 10485750000,
},
{
name: "Zero values",
bytesSentPerSecond: 0,
@@ -460,6 +443,8 @@ func TestApplyNetworkTotals(t *testing.T) {
totalBytesSent: 0,
totalBytesRecv: 0,
expectReset: false,
expectedNetworkSent: 0.0,
expectedNetworkRecv: 0.0,
expectedBandwidthSent: 0,
expectedBandwidthRecv: 0,
},
@@ -496,10 +481,14 @@ func TestApplyNetworkTotals(t *testing.T) {
// Should have reset network tracking state - maps cleared and stats zeroed
assert.NotContains(t, a.netIoStats, cacheTimeMs, "cache entry should be cleared after reset")
assert.NotContains(t, a.netInterfaceDeltaTrackers, cacheTimeMs, "tracker should be cleared on reset")
assert.Zero(t, systemStats.NetworkSent)
assert.Zero(t, systemStats.NetworkRecv)
assert.Zero(t, systemStats.Bandwidth[0])
assert.Zero(t, systemStats.Bandwidth[1])
} else {
// Should have applied stats
assert.Equal(t, tt.expectedNetworkSent, systemStats.NetworkSent)
assert.Equal(t, tt.expectedNetworkRecv, systemStats.NetworkRecv)
assert.Equal(t, tt.expectedBandwidthSent, systemStats.Bandwidth[0])
assert.Equal(t, tt.expectedBandwidthRecv, systemStats.Bandwidth[1])

View File

@@ -10,7 +10,6 @@ import (
"strings"
"unicode/utf8"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/system"
"github.com/shirou/gopsutil/v4/common"
@@ -27,9 +26,9 @@ type SensorConfig struct {
}
func (a *Agent) newSensorConfig() *SensorConfig {
primarySensor, _ := utils.GetEnv("PRIMARY_SENSOR")
sysSensors, _ := utils.GetEnv("SYS_SENSORS")
sensorsEnvVal, sensorsSet := utils.GetEnv("SENSORS")
primarySensor, _ := GetEnv("PRIMARY_SENSOR")
sysSensors, _ := GetEnv("SYS_SENSORS")
sensorsEnvVal, sensorsSet := GetEnv("SENSORS")
skipCollection := sensorsSet && sensorsEnvVal == ""
return a.newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal, skipCollection)
@@ -136,7 +135,7 @@ func (a *Agent) updateTemperatures(systemStats *system.Stats) {
case sensorName:
a.systemInfo.DashboardTemp = sensor.Temperature
}
systemStats.Temperatures[sensorName] = utils.TwoDecimals(sensor.Temperature)
systemStats.Temperatures[sensorName] = twoDecimals(sensor.Temperature)
}
}

View File

@@ -12,7 +12,6 @@ import (
"time"
"github.com/henrygd/beszel"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/common"
"github.com/henrygd/beszel/internal/entities/system"
@@ -37,7 +36,7 @@ var hubVersions map[string]semver.Version
// and begins listening for connections. Returns an error if the server
// is already running or if there's an issue starting the server.
func (a *Agent) StartServer(opts ServerOptions) error {
if disableSSH, _ := utils.GetEnv("DISABLE_SSH"); disableSSH == "true" {
if disableSSH, _ := GetEnv("DISABLE_SSH"); disableSSH == "true" {
return errors.New("SSH disabled")
}
if a.server != nil {
@@ -239,11 +238,11 @@ func ParseKeys(input string) ([]gossh.PublicKey, error) {
// and finally defaults to ":45876".
func GetAddress(addr string) string {
if addr == "" {
addr, _ = utils.GetEnv("LISTEN")
addr, _ = GetEnv("LISTEN")
}
if addr == "" {
// Legacy PORT environment variable support
addr, _ = utils.GetEnv("PORT")
addr, _ = GetEnv("PORT")
}
if addr == "" {
return ":45876"
@@ -259,7 +258,7 @@ func GetAddress(addr string) string {
// It checks the NETWORK environment variable first, then infers from
// the address format: addresses starting with "/" are "unix", others are "tcp".
func GetNetwork(addr string) string {
if network, ok := utils.GetEnv("NETWORK"); ok && network != "" {
if network, ok := GetEnv("NETWORK"); ok && network != "" {
return network
}
if strings.HasPrefix(addr, "/") {

View File

@@ -18,7 +18,6 @@ import (
"sync"
"time"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/smart"
)
@@ -157,7 +156,7 @@ func (sm *SmartManager) ScanDevices(force bool) error {
currentDevices := sm.devicesSnapshot()
var configuredDevices []*DeviceInfo
if configuredRaw, ok := utils.GetEnv("SMART_DEVICES"); ok {
if configuredRaw, ok := GetEnv("SMART_DEVICES"); ok {
slog.Info("SMART_DEVICES", "value", configuredRaw)
config := strings.TrimSpace(configuredRaw)
if config == "" {
@@ -200,13 +199,6 @@ func (sm *SmartManager) ScanDevices(force bool) error {
hasValidScan = true
}
// Add Linux mdraid arrays by reading sysfs health fields. This does not
// require smartctl and does not scan the whole device.
if raidDevices := scanMdraidDevices(); len(raidDevices) > 0 {
scannedDevices = append(scannedDevices, raidDevices...)
hasValidScan = true
}
finalDevices := mergeDeviceLists(currentDevices, scannedDevices, configuredDevices)
finalDevices = sm.filterExcludedDevices(finalDevices)
sm.updateSmartDevices(finalDevices)
@@ -223,7 +215,7 @@ func (sm *SmartManager) ScanDevices(force bool) error {
}
func (sm *SmartManager) parseConfiguredDevices(config string) ([]*DeviceInfo, error) {
splitChar, _ := utils.GetEnv("SMART_DEVICES_SEPARATOR")
splitChar := os.Getenv("SMART_DEVICES_SEPARATOR")
if splitChar == "" {
splitChar = ","
}
@@ -261,7 +253,7 @@ func (sm *SmartManager) parseConfiguredDevices(config string) ([]*DeviceInfo, er
}
func (sm *SmartManager) refreshExcludedDevices() {
rawValue, _ := utils.GetEnv("EXCLUDE_SMART")
rawValue, _ := GetEnv("EXCLUDE_SMART")
sm.excludedDevices = make(map[string]struct{})
for entry := range strings.SplitSeq(rawValue, ",") {
@@ -458,12 +450,6 @@ func (sm *SmartManager) CollectSmart(deviceInfo *DeviceInfo) error {
return errNoValidSmartData
}
// mdraid health is not exposed via SMART; Linux exposes array state in sysfs.
if deviceInfo != nil {
if ok, err := sm.collectMdraidHealth(deviceInfo); ok {
return err
}
}
// eMMC health is not exposed via SMART on Linux, but the kernel provides
// wear / EOL indicators via sysfs. Prefer that path when available.
if deviceInfo != nil {
@@ -871,18 +857,15 @@ func (sm *SmartManager) parseSmartForSata(output []byte) (bool, int) {
smartData.FirmwareVersion = data.FirmwareVersion
smartData.Capacity = data.UserCapacity.Bytes
smartData.Temperature = data.Temperature.Current
if smartData.Temperature == 0 {
if temp, ok := temperatureFromAtaDeviceStatistics(data.AtaDeviceStatistics); ok {
smartData.Temperature = temp
}
}
smartData.SmartStatus = getSmartStatus(smartData.Temperature, data.SmartStatus.Passed)
smartData.DiskName = data.Device.Name
smartData.DiskType = data.Device.Type
// get values from ata_device_statistics if necessary
var ataDeviceStats smart.AtaDeviceStatistics
if smartData.Temperature == 0 {
if temp := findAtaDeviceStatisticsValue(&data, &ataDeviceStats, 5, "Current Temperature", 0, 255); temp != nil {
smartData.Temperature = uint8(*temp)
}
}
// update SmartAttributes
smartData.Attributes = make([]*smart.SmartAttribute, 0, len(data.AtaSmartAttributes.Table))
for _, attr := range data.AtaSmartAttributes.Table {
@@ -917,20 +900,23 @@ func getSmartStatus(temperature uint8, passed bool) string {
}
}
func temperatureFromAtaDeviceStatistics(stats smart.AtaDeviceStatistics) (uint8, bool) {
entry := findAtaDeviceStatisticsEntry(stats, 5, "Current Temperature")
if entry == nil || entry.Value == nil {
return 0, false
}
if *entry.Value > 255 {
return 0, false
}
return uint8(*entry.Value), true
}
// findAtaDeviceStatisticsEntry centralizes ATA devstat lookups so additional
// metrics can be pulled from the same structure in the future.
func findAtaDeviceStatisticsValue(data *smart.SmartInfoForSata, ataDeviceStats *smart.AtaDeviceStatistics, entryNumber uint8, entryName string, minValue, maxValue int64) *int64 {
if len(ataDeviceStats.Pages) == 0 {
if len(data.AtaDeviceStatistics) == 0 {
return nil
}
if err := json.Unmarshal(data.AtaDeviceStatistics, ataDeviceStats); err != nil {
return nil
}
}
for pageIdx := range ataDeviceStats.Pages {
page := &ataDeviceStats.Pages[pageIdx]
if page.Number != entryNumber {
func findAtaDeviceStatisticsEntry(stats smart.AtaDeviceStatistics, pageNumber uint8, entryName string) *smart.AtaDeviceStatisticsEntry {
for pageIdx := range stats.Pages {
page := &stats.Pages[pageIdx]
if page.Number != pageNumber {
continue
}
for entryIdx := range page.Table {
@@ -938,10 +924,7 @@ func findAtaDeviceStatisticsValue(data *smart.SmartInfoForSata, ataDeviceStats *
if !strings.EqualFold(entry.Name, entryName) {
continue
}
if entry.Value == nil || *entry.Value < minValue || *entry.Value > maxValue {
return nil
}
return entry.Value
return entry
}
}
return nil
@@ -1163,11 +1146,9 @@ func NewSmartManager() (*SmartManager, error) {
slog.Debug("smartctl", "path", path, "err", err)
if err != nil {
// Keep the previous fail-fast behavior unless this Linux host exposes
// eMMC or mdraid health via sysfs, in which case smartctl is optional.
if runtime.GOOS == "linux" {
if len(scanEmmcDevices()) > 0 || len(scanMdraidDevices()) > 0 {
return sm, nil
}
// eMMC health via sysfs, in which case smartctl is optional.
if runtime.GOOS == "linux" && len(scanEmmcDevices()) > 0 {
return sm, nil
}
return nil, err
}

View File

@@ -121,78 +121,6 @@ func TestParseSmartForSataDeviceStatisticsTemperature(t *testing.T) {
assert.Equal(t, uint8(22), deviceData.Temperature)
}
func TestParseSmartForSataAtaDeviceStatistics(t *testing.T) {
// tests that ata_device_statistics values are parsed correctly
jsonPayload := []byte(`{
"smartctl": {"exit_status": 0},
"device": {"name": "/dev/sdb", "type": "sat"},
"model_name": "SanDisk SSD U110 16GB",
"serial_number": "lksjfh23lhj",
"firmware_version": "U21B001",
"user_capacity": {"bytes": 16013942784},
"smart_status": {"passed": true},
"ata_smart_attributes": {"table": []},
"ata_device_statistics": {
"pages": [
{
"number": 5,
"name": "Temperature Statistics",
"table": [
{"name": "Current Temperature", "value": 43, "flags": {"valid": true}},
{"name": "Specified Minimum Operating Temperature", "value": -20, "flags": {"valid": true}}
]
}
]
}
}`)
sm := &SmartManager{SmartDataMap: make(map[string]*smart.SmartData)}
hasData, exitStatus := sm.parseSmartForSata(jsonPayload)
require.True(t, hasData)
assert.Equal(t, 0, exitStatus)
deviceData, ok := sm.SmartDataMap["lksjfh23lhj"]
require.True(t, ok, "expected smart data entry for serial lksjfh23lhj")
assert.Equal(t, uint8(43), deviceData.Temperature)
}
func TestParseSmartForSataNegativeDeviceStatistics(t *testing.T) {
// Tests that negative values in ata_device_statistics (e.g. min operating temp)
// do not cause the entire SAT parser to fail.
jsonPayload := []byte(`{
"smartctl": {"exit_status": 0},
"device": {"name": "/dev/sdb", "type": "sat"},
"model_name": "SanDisk SSD U110 16GB",
"serial_number": "NEGATIVE123",
"firmware_version": "U21B001",
"user_capacity": {"bytes": 16013942784},
"smart_status": {"passed": true},
"temperature": {"current": 38},
"ata_smart_attributes": {"table": []},
"ata_device_statistics": {
"pages": [
{
"number": 5,
"name": "Temperature Statistics",
"table": [
{"name": "Current Temperature", "value": 38, "flags": {"valid": true}},
{"name": "Specified Minimum Operating Temperature", "value": -20, "flags": {"valid": true}}
]
}
]
}
}`)
sm := &SmartManager{SmartDataMap: make(map[string]*smart.SmartData)}
hasData, exitStatus := sm.parseSmartForSata(jsonPayload)
require.True(t, hasData)
assert.Equal(t, 0, exitStatus)
deviceData, ok := sm.SmartDataMap["NEGATIVE123"]
require.True(t, ok, "expected smart data entry for serial NEGATIVE123")
assert.Equal(t, uint8(38), deviceData.Temperature)
}
func TestParseSmartForSataParentheticalRawValue(t *testing.T) {
jsonPayload := []byte(`{
"smartctl": {"exit_status": 0},
@@ -799,182 +727,6 @@ func TestIsVirtualDeviceScsi(t *testing.T) {
}
}
func TestFindAtaDeviceStatisticsValue(t *testing.T) {
val42 := int64(42)
val100 := int64(100)
valMinus20 := int64(-20)
tests := []struct {
name string
data smart.SmartInfoForSata
ataDeviceStats smart.AtaDeviceStatistics
entryNumber uint8
entryName string
minValue int64
maxValue int64
expectedValue *int64
}{
{
name: "value in ataDeviceStats",
ataDeviceStats: smart.AtaDeviceStatistics{
Pages: []smart.AtaDeviceStatisticsPage{
{
Number: 5,
Table: []smart.AtaDeviceStatisticsEntry{
{Name: "Current Temperature", Value: &val42},
},
},
},
},
entryNumber: 5,
entryName: "Current Temperature",
minValue: 0,
maxValue: 100,
expectedValue: &val42,
},
{
name: "value unmarshaled from data",
data: smart.SmartInfoForSata{
AtaDeviceStatistics: []byte(`{"pages":[{"number":5,"table":[{"name":"Current Temperature","value":100}]}]}`),
},
entryNumber: 5,
entryName: "Current Temperature",
minValue: 0,
maxValue: 255,
expectedValue: &val100,
},
{
name: "value out of range (too high)",
ataDeviceStats: smart.AtaDeviceStatistics{
Pages: []smart.AtaDeviceStatisticsPage{
{
Number: 5,
Table: []smart.AtaDeviceStatisticsEntry{
{Name: "Current Temperature", Value: &val100},
},
},
},
},
entryNumber: 5,
entryName: "Current Temperature",
minValue: 0,
maxValue: 50,
expectedValue: nil,
},
{
name: "value out of range (too low)",
ataDeviceStats: smart.AtaDeviceStatistics{
Pages: []smart.AtaDeviceStatisticsPage{
{
Number: 5,
Table: []smart.AtaDeviceStatisticsEntry{
{Name: "Min Temp", Value: &valMinus20},
},
},
},
},
entryNumber: 5,
entryName: "Min Temp",
minValue: 0,
maxValue: 100,
expectedValue: nil,
},
{
name: "no statistics available",
data: smart.SmartInfoForSata{},
entryNumber: 5,
entryName: "Current Temperature",
minValue: 0,
maxValue: 255,
expectedValue: nil,
},
{
name: "wrong page number",
ataDeviceStats: smart.AtaDeviceStatistics{
Pages: []smart.AtaDeviceStatisticsPage{
{
Number: 1,
Table: []smart.AtaDeviceStatisticsEntry{
{Name: "Current Temperature", Value: &val42},
},
},
},
},
entryNumber: 5,
entryName: "Current Temperature",
minValue: 0,
maxValue: 100,
expectedValue: nil,
},
{
name: "wrong entry name",
ataDeviceStats: smart.AtaDeviceStatistics{
Pages: []smart.AtaDeviceStatisticsPage{
{
Number: 5,
Table: []smart.AtaDeviceStatisticsEntry{
{Name: "Other Stat", Value: &val42},
},
},
},
},
entryNumber: 5,
entryName: "Current Temperature",
minValue: 0,
maxValue: 100,
expectedValue: nil,
},
{
name: "case insensitive name match",
ataDeviceStats: smart.AtaDeviceStatistics{
Pages: []smart.AtaDeviceStatisticsPage{
{
Number: 5,
Table: []smart.AtaDeviceStatisticsEntry{
{Name: "CURRENT TEMPERATURE", Value: &val42},
},
},
},
},
entryNumber: 5,
entryName: "Current Temperature",
minValue: 0,
maxValue: 100,
expectedValue: &val42,
},
{
name: "entry value is nil",
ataDeviceStats: smart.AtaDeviceStatistics{
Pages: []smart.AtaDeviceStatisticsPage{
{
Number: 5,
Table: []smart.AtaDeviceStatisticsEntry{
{Name: "Current Temperature", Value: nil},
},
},
},
},
entryNumber: 5,
entryName: "Current Temperature",
minValue: 0,
maxValue: 100,
expectedValue: nil,
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result := findAtaDeviceStatisticsValue(&tt.data, &tt.ataDeviceStats, tt.entryNumber, tt.entryName, tt.minValue, tt.maxValue)
if tt.expectedValue == nil {
assert.Nil(t, result)
} else {
require.NotNil(t, result)
assert.Equal(t, *tt.expectedValue, *result)
}
})
}
}
func TestRefreshExcludedDevices(t *testing.T) {
tests := []struct {
name string

View File

@@ -7,13 +7,12 @@ import (
"log/slog"
"os"
"runtime"
"strconv"
"strings"
"time"
"github.com/henrygd/beszel"
"github.com/henrygd/beszel/agent/battery"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/agent/zfs"
"github.com/henrygd/beszel/internal/entities/container"
"github.com/henrygd/beszel/internal/entities/system"
@@ -108,7 +107,7 @@ func (a *Agent) refreshSystemDetails() {
}
// zfs
if _, err := zfs.ARCSize(); err != nil {
if _, err := getARCSize(); err != nil {
slog.Debug("Not monitoring ZFS ARC", "err", err)
} else {
a.zfs = true
@@ -128,13 +127,13 @@ func (a *Agent) getSystemStats(cacheTimeMs uint16) system.Stats {
// cpu metrics
cpuMetrics, err := getCpuMetrics(cacheTimeMs)
if err == nil {
systemStats.Cpu = utils.TwoDecimals(cpuMetrics.Total)
systemStats.Cpu = twoDecimals(cpuMetrics.Total)
systemStats.CpuBreakdown = []float64{
utils.TwoDecimals(cpuMetrics.User),
utils.TwoDecimals(cpuMetrics.System),
utils.TwoDecimals(cpuMetrics.Iowait),
utils.TwoDecimals(cpuMetrics.Steal),
utils.TwoDecimals(cpuMetrics.Idle),
twoDecimals(cpuMetrics.User),
twoDecimals(cpuMetrics.System),
twoDecimals(cpuMetrics.Iowait),
twoDecimals(cpuMetrics.Steal),
twoDecimals(cpuMetrics.Idle),
}
} else {
slog.Error("Error getting cpu metrics", "err", err)
@@ -158,8 +157,8 @@ func (a *Agent) getSystemStats(cacheTimeMs uint16) system.Stats {
// memory
if v, err := mem.VirtualMemory(); err == nil {
// swap
systemStats.Swap = utils.BytesToGigabytes(v.SwapTotal)
systemStats.SwapUsed = utils.BytesToGigabytes(v.SwapTotal - v.SwapFree - v.SwapCached)
systemStats.Swap = bytesToGigabytes(v.SwapTotal)
systemStats.SwapUsed = bytesToGigabytes(v.SwapTotal - v.SwapFree - v.SwapCached)
// cache + buffers value for default mem calculation
// note: gopsutil automatically adds SReclaimable to v.Cached
cacheBuff := v.Cached + v.Buffers - v.Shared
@@ -179,16 +178,16 @@ func (a *Agent) getSystemStats(cacheTimeMs uint16) system.Stats {
// }
// subtract ZFS ARC size from used memory and add as its own category
if a.zfs {
if arcSize, _ := zfs.ARCSize(); arcSize > 0 && arcSize < v.Used {
if arcSize, _ := getARCSize(); arcSize > 0 && arcSize < v.Used {
v.Used = v.Used - arcSize
v.UsedPercent = float64(v.Used) / float64(v.Total) * 100.0
systemStats.MemZfsArc = utils.BytesToGigabytes(arcSize)
systemStats.MemZfsArc = bytesToGigabytes(arcSize)
}
}
systemStats.Mem = utils.BytesToGigabytes(v.Total)
systemStats.MemBuffCache = utils.BytesToGigabytes(cacheBuff)
systemStats.MemUsed = utils.BytesToGigabytes(v.Used)
systemStats.MemPct = utils.TwoDecimals(v.UsedPercent)
systemStats.Mem = bytesToGigabytes(v.Total)
systemStats.MemBuffCache = bytesToGigabytes(cacheBuff)
systemStats.MemUsed = bytesToGigabytes(v.Used)
systemStats.MemPct = twoDecimals(v.UsedPercent)
}
// disk usage
@@ -251,6 +250,32 @@ func (a *Agent) getSystemStats(cacheTimeMs uint16) system.Stats {
return systemStats
}
// Returns the size of the ZFS ARC memory cache in bytes
func getARCSize() (uint64, error) {
file, err := os.Open("/proc/spl/kstat/zfs/arcstats")
if err != nil {
return 0, err
}
defer file.Close()
// Scan the lines
scanner := bufio.NewScanner(file)
for scanner.Scan() {
line := scanner.Text()
if strings.HasPrefix(line, "size") {
// Example line: size 4 15032385536
fields := strings.Fields(line)
if len(fields) < 3 {
return 0, err
}
// Return the size as uint64
return strconv.ParseUint(fields[2], 10, 64)
}
}
return 0, fmt.Errorf("failed to parse size field")
}
// getOsPrettyName attempts to get the pretty OS name from /etc/os-release on Linux systems
func getOsPrettyName() (string, error) {
file, err := os.Open("/etc/os-release")

View File

@@ -15,7 +15,6 @@ import (
"time"
"github.com/coreos/go-systemd/v22/dbus"
"github.com/henrygd/beszel/agent/utils"
"github.com/henrygd/beszel/internal/entities/systemd"
)
@@ -50,7 +49,7 @@ func isSystemdAvailable() bool {
// newSystemdManager creates a new systemdManager.
func newSystemdManager() (*systemdManager, error) {
if skipSystemd, _ := utils.GetEnv("SKIP_SYSTEMD"); skipSystemd == "true" {
if skipSystemd, _ := GetEnv("SKIP_SYSTEMD"); skipSystemd == "true" {
return nil, nil
}
@@ -295,7 +294,7 @@ func unescapeServiceName(name string) string {
// otherwise defaults to "*service".
func getServicePatterns() []string {
patterns := []string{}
if envPatterns, _ := utils.GetEnv("SERVICE_PATTERNS"); envPatterns != "" {
if envPatterns, _ := GetEnv("SERVICE_PATTERNS"); envPatterns != "" {
for pattern := range strings.SplitSeq(envPatterns, ",") {
pattern = strings.TrimSpace(pattern)
if pattern == "" {

15
agent/utils.go Normal file
View File

@@ -0,0 +1,15 @@
package agent
import "math"
func bytesToMegabytes(b float64) float64 {
return twoDecimals(b / 1048576)
}
func bytesToGigabytes(b uint64) float64 {
return twoDecimals(float64(b) / 1073741824)
}
func twoDecimals(value float64) float64 {
return math.Round(value*100) / 100
}

View File

@@ -1,88 +0,0 @@
package utils
import (
"io"
"math"
"os"
"strconv"
"strings"
)
// GetEnv retrieves an environment variable with a "BESZEL_AGENT_" prefix, or falls back to the unprefixed key.
func GetEnv(key string) (value string, exists bool) {
if value, exists = os.LookupEnv("BESZEL_AGENT_" + key); exists {
return value, exists
}
return os.LookupEnv(key)
}
// BytesToMegabytes converts bytes to megabytes and rounds to two decimal places.
func BytesToMegabytes(b float64) float64 {
return TwoDecimals(b / 1048576)
}
// BytesToGigabytes converts bytes to gigabytes and rounds to two decimal places.
func BytesToGigabytes(b uint64) float64 {
return TwoDecimals(float64(b) / 1073741824)
}
// TwoDecimals rounds a float64 value to two decimal places.
func TwoDecimals(value float64) float64 {
return math.Round(value*100) / 100
}
// func RoundFloat(val float64, precision uint) float64 {
// ratio := math.Pow(10, float64(precision))
// return math.Round(val*ratio) / ratio
// }
// ReadStringFile returns trimmed file contents or empty string on error.
func ReadStringFile(path string) string {
content, _ := ReadStringFileOK(path)
return content
}
// ReadStringFileOK returns trimmed file contents and read success.
func ReadStringFileOK(path string) (string, bool) {
b, err := os.ReadFile(path)
if err != nil {
return "", false
}
return strings.TrimSpace(string(b)), true
}
// ReadStringFileLimited reads a file into a string with a maximum size (in bytes) to avoid
// allocating large buffers and potential panics with pseudo-files when the size is misreported.
func ReadStringFileLimited(path string, maxSize int) (string, error) {
f, err := os.Open(path)
if err != nil {
return "", err
}
defer f.Close()
buf := make([]byte, maxSize)
n, err := f.Read(buf)
if err != nil && err != io.EOF {
return "", err
}
return strings.TrimSpace(string(buf[:n])), nil
}
// FileExists reports whether the given path exists.
func FileExists(path string) bool {
_, err := os.Stat(path)
return err == nil
}
// ReadUintFile parses a decimal uint64 value from a file.
func ReadUintFile(path string) (uint64, bool) {
raw, ok := ReadStringFileOK(path)
if !ok {
return 0, false
}
parsed, err := strconv.ParseUint(raw, 10, 64)
if err != nil {
return 0, false
}
return parsed, true
}

View File

@@ -1,165 +0,0 @@
package utils
import (
"os"
"path/filepath"
"testing"
"github.com/stretchr/testify/assert"
)
func TestTwoDecimals(t *testing.T) {
tests := []struct {
name string
input float64
expected float64
}{
{"round down", 1.234, 1.23},
{"round half up", 1.235, 1.24}, // math.Round rounds half up
{"no rounding needed", 1.23, 1.23},
{"negative number", -1.235, -1.24}, // math.Round rounds half up (more negative)
{"zero", 0.0, 0.0},
{"large number", 123.456, 123.46}, // rounds 5 up
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result := TwoDecimals(tt.input)
assert.Equal(t, tt.expected, result)
})
}
}
func TestBytesToMegabytes(t *testing.T) {
tests := []struct {
name string
input float64
expected float64
}{
{"1 MB", 1048576, 1.0},
{"512 KB", 524288, 0.5},
{"zero", 0, 0},
{"large value", 1073741824, 1024}, // 1 GB = 1024 MB
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result := BytesToMegabytes(tt.input)
assert.Equal(t, tt.expected, result)
})
}
}
func TestBytesToGigabytes(t *testing.T) {
tests := []struct {
name string
input uint64
expected float64
}{
{"1 GB", 1073741824, 1.0},
{"512 MB", 536870912, 0.5},
{"0 GB", 0, 0},
{"2 GB", 2147483648, 2.0},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result := BytesToGigabytes(tt.input)
assert.Equal(t, tt.expected, result)
})
}
}
func TestFileFunctions(t *testing.T) {
tmpDir := t.TempDir()
testFilePath := filepath.Join(tmpDir, "test.txt")
testContent := "hello world"
// Test FileExists (false)
assert.False(t, FileExists(testFilePath))
// Test ReadStringFileOK (false)
content, ok := ReadStringFileOK(testFilePath)
assert.False(t, ok)
assert.Empty(t, content)
// Test ReadStringFile (empty)
assert.Empty(t, ReadStringFile(testFilePath))
// Write file
err := os.WriteFile(testFilePath, []byte(testContent+"\n "), 0644)
assert.NoError(t, err)
// Test FileExists (true)
assert.True(t, FileExists(testFilePath))
// Test ReadStringFileOK (true)
content, ok = ReadStringFileOK(testFilePath)
assert.True(t, ok)
assert.Equal(t, testContent, content)
// Test ReadStringFile (content)
assert.Equal(t, testContent, ReadStringFile(testFilePath))
}
func TestReadUintFile(t *testing.T) {
tmpDir := t.TempDir()
t.Run("valid uint", func(t *testing.T) {
path := filepath.Join(tmpDir, "uint.txt")
os.WriteFile(path, []byte(" 12345\n"), 0644)
val, ok := ReadUintFile(path)
assert.True(t, ok)
assert.Equal(t, uint64(12345), val)
})
t.Run("invalid uint", func(t *testing.T) {
path := filepath.Join(tmpDir, "invalid.txt")
os.WriteFile(path, []byte("abc"), 0644)
val, ok := ReadUintFile(path)
assert.False(t, ok)
assert.Equal(t, uint64(0), val)
})
t.Run("missing file", func(t *testing.T) {
path := filepath.Join(tmpDir, "missing.txt")
val, ok := ReadUintFile(path)
assert.False(t, ok)
assert.Equal(t, uint64(0), val)
})
}
func TestGetEnv(t *testing.T) {
key := "TEST_VAR"
prefixedKey := "BESZEL_AGENT_" + key
t.Run("prefixed variable exists", func(t *testing.T) {
os.Setenv(prefixedKey, "prefixed_val")
os.Setenv(key, "unprefixed_val")
defer os.Unsetenv(prefixedKey)
defer os.Unsetenv(key)
val, exists := GetEnv(key)
assert.True(t, exists)
assert.Equal(t, "prefixed_val", val)
})
t.Run("only unprefixed variable exists", func(t *testing.T) {
os.Unsetenv(prefixedKey)
os.Setenv(key, "unprefixed_val")
defer os.Unsetenv(key)
val, exists := GetEnv(key)
assert.True(t, exists)
assert.Equal(t, "unprefixed_val", val)
})
t.Run("neither variable exists", func(t *testing.T) {
os.Unsetenv(prefixedKey)
os.Unsetenv(key)
val, exists := GetEnv(key)
assert.False(t, exists)
assert.Empty(t, val)
})
}

View File

@@ -1,11 +0,0 @@
//go:build freebsd
package zfs
import (
"golang.org/x/sys/unix"
)
func ARCSize() (uint64, error) {
return unix.SysctlUint64("kstat.zfs.misc.arcstats.size")
}

View File

@@ -1,34 +0,0 @@
//go:build linux
// Package zfs provides functions to read ZFS statistics.
package zfs
import (
"bufio"
"fmt"
"os"
"strconv"
"strings"
)
func ARCSize() (uint64, error) {
file, err := os.Open("/proc/spl/kstat/zfs/arcstats")
if err != nil {
return 0, err
}
defer file.Close()
scanner := bufio.NewScanner(file)
for scanner.Scan() {
line := scanner.Text()
if strings.HasPrefix(line, "size") {
fields := strings.Fields(line)
if len(fields) < 3 {
return 0, fmt.Errorf("unexpected arcstats size format: %s", line)
}
return strconv.ParseUint(fields[2], 10, 64)
}
}
return 0, fmt.Errorf("size field not found in arcstats")
}

View File

@@ -1,9 +0,0 @@
//go:build !linux && !freebsd
package zfs
import "errors"
func ARCSize() (uint64, error) {
return 0, errors.ErrUnsupported
}

View File

@@ -6,7 +6,7 @@ import "github.com/blang/semver"
const (
// Version is the current version of the application.
Version = "0.18.4"
Version = "0.18.3"
// AppName is the name of the application.
AppName = "beszel"
)

2
go.mod
View File

@@ -1,6 +1,6 @@
module github.com/henrygd/beszel
go 1.26.1
go 1.26.0
require (
github.com/blang/semver v3.5.1+incompatible

View File

@@ -21,7 +21,8 @@ type hubLike interface {
type AlertManager struct {
hub hubLike
stopOnce sync.Once
alertQueue chan alertTask
stopChan chan struct{}
pendingAlerts sync.Map
}
@@ -39,22 +40,16 @@ type UserNotificationSettings struct {
Webhooks []string `json:"webhooks"`
}
type SystemAlertFsStats struct {
DiskTotal float64 `json:"d"`
DiskUsed float64 `json:"du"`
}
// Values pulled from system_stats.stats that are relevant to alerts.
type SystemAlertStats struct {
Cpu float64 `json:"cpu"`
Mem float64 `json:"mp"`
Disk float64 `json:"dp"`
Bandwidth [2]uint64 `json:"b"`
NetSent float64 `json:"ns"`
NetRecv float64 `json:"nr"`
GPU map[string]SystemAlertGPUData `json:"g"`
Temperatures map[string]float32 `json:"t"`
LoadAvg [3]float64 `json:"la"`
Battery [2]uint8 `json:"bat"`
ExtraFs map[string]SystemAlertFsStats `json:"efs"`
}
type SystemAlertGPUData struct {
@@ -97,9 +92,12 @@ var supportsTitle = map[string]struct{}{
// NewAlertManager creates a new AlertManager instance.
func NewAlertManager(app hubLike) *AlertManager {
am := &AlertManager{
hub: app,
hub: app,
alertQueue: make(chan alertTask, 5),
stopChan: make(chan struct{}),
}
am.bindEvents()
go am.startWorker()
return am
}
@@ -108,16 +106,6 @@ func (am *AlertManager) bindEvents() {
am.hub.OnRecordAfterUpdateSuccess("alerts").BindFunc(updateHistoryOnAlertUpdate)
am.hub.OnRecordAfterDeleteSuccess("alerts").BindFunc(resolveHistoryOnAlertDelete)
am.hub.OnRecordAfterUpdateSuccess("smart_devices").BindFunc(am.handleSmartDeviceAlert)
am.hub.OnServe().BindFunc(func(e *core.ServeEvent) error {
if err := resolveStatusAlerts(e.App); err != nil {
e.App.Logger().Error("Failed to resolve stale status alerts", "err", err)
}
if err := am.restorePendingStatusAlerts(); err != nil {
e.App.Logger().Error("Failed to restore pending status alerts", "err", err)
}
return e.Next()
})
}
// IsNotificationSilenced checks if a notification should be silenced based on configured quiet hours
@@ -271,14 +259,13 @@ func (am *AlertManager) SendShoutrrrAlert(notificationUrl, title, message, link,
}
// Add link
switch scheme {
case "ntfy":
if scheme == "ntfy" {
queryParams.Add("Actions", fmt.Sprintf("view, %s, %s", linkText, link))
case "lark":
} else if scheme == "lark" {
queryParams.Add("link", link)
case "bark":
} else if scheme == "bark" {
queryParams.Add("url", link)
default:
} else {
message += "\n\n" + link
}

View File

@@ -1,155 +0,0 @@
//go:build testing
package alerts_test
import (
"encoding/json"
"testing"
"time"
"github.com/henrygd/beszel/internal/entities/system"
beszelTests "github.com/henrygd/beszel/internal/tests"
"github.com/pocketbase/dbx"
"github.com/pocketbase/pocketbase/tools/types"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
// TestDiskAlertExtraFsMultiMinute tests that multi-minute disk alerts correctly use
// historical per-minute values for extra (non-root) filesystems, not the current live snapshot.
func TestDiskAlertExtraFsMultiMinute(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
systems, err := beszelTests.CreateSystems(hub, 1, user.Id, "up")
require.NoError(t, err)
systemRecord := systems[0]
// Disk alert: threshold 80%, min=2 (requires historical averaging)
diskAlert, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Disk",
"system": systemRecord.Id,
"user": user.Id,
"value": 80, // threshold: 80%
"min": 2, // 2 minutes - requires historical averaging
})
require.NoError(t, err)
assert.False(t, diskAlert.GetBool("triggered"), "Alert should not be triggered initially")
am := hub.GetAlertManager()
now := time.Now().UTC()
extraFsHigh := map[string]*system.FsStats{
"/mnt/data": {DiskTotal: 1000, DiskUsed: 920}, // 92% - above threshold
}
// Insert 4 historical records spread over 3 minutes (same pattern as battery tests).
// The oldest record must predate (now - 2min) so the alert time window is valid.
recordTimes := []time.Duration{
-180 * time.Second, // 3 min ago - anchors oldest record before alert.time
-90 * time.Second,
-60 * time.Second,
-30 * time.Second,
}
for _, offset := range recordTimes {
stats := system.Stats{
DiskPct: 30, // root disk at 30% - below threshold
ExtraFs: extraFsHigh,
}
statsJSON, _ := json.Marshal(stats)
recordTime := now.Add(offset)
record, err := beszelTests.CreateRecord(hub, "system_stats", map[string]any{
"system": systemRecord.Id,
"type": "1m",
"stats": string(statsJSON),
})
require.NoError(t, err)
record.SetRaw("created", recordTime.Format(types.DefaultDateLayout))
err = hub.SaveNoValidate(record)
require.NoError(t, err)
}
combinedDataHigh := &system.CombinedData{
Stats: system.Stats{
DiskPct: 30,
ExtraFs: extraFsHigh,
},
Info: system.Info{
DiskPct: 30,
},
}
systemRecord.Set("updated", now)
err = hub.SaveNoValidate(systemRecord)
require.NoError(t, err)
err = am.HandleSystemAlerts(systemRecord, combinedDataHigh)
require.NoError(t, err)
time.Sleep(20 * time.Millisecond)
diskAlert, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": diskAlert.Id})
require.NoError(t, err)
assert.True(t, diskAlert.GetBool("triggered"),
"Alert SHOULD be triggered when extra disk average (92%%) exceeds threshold (80%%)")
// --- Resolution: extra disk drops to 50%, alert should resolve ---
extraFsLow := map[string]*system.FsStats{
"/mnt/data": {DiskTotal: 1000, DiskUsed: 500}, // 50% - below threshold
}
newNow := now.Add(2 * time.Minute)
recordTimesLow := []time.Duration{
-180 * time.Second,
-90 * time.Second,
-60 * time.Second,
-30 * time.Second,
}
for _, offset := range recordTimesLow {
stats := system.Stats{
DiskPct: 30,
ExtraFs: extraFsLow,
}
statsJSON, _ := json.Marshal(stats)
recordTime := newNow.Add(offset)
record, err := beszelTests.CreateRecord(hub, "system_stats", map[string]any{
"system": systemRecord.Id,
"type": "1m",
"stats": string(statsJSON),
})
require.NoError(t, err)
record.SetRaw("created", recordTime.Format(types.DefaultDateLayout))
err = hub.SaveNoValidate(record)
require.NoError(t, err)
}
combinedDataLow := &system.CombinedData{
Stats: system.Stats{
DiskPct: 30,
ExtraFs: extraFsLow,
},
Info: system.Info{
DiskPct: 30,
},
}
systemRecord.Set("updated", newNow)
err = hub.SaveNoValidate(systemRecord)
require.NoError(t, err)
err = am.HandleSystemAlerts(systemRecord, combinedDataLow)
require.NoError(t, err)
time.Sleep(20 * time.Millisecond)
diskAlert, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": diskAlert.Id})
require.NoError(t, err)
assert.False(t, diskAlert.GetBool("triggered"),
"Alert should be resolved when extra disk average (50%%) drops below threshold (80%%)")
}

View File

@@ -49,7 +49,7 @@ func TestAlertSilencedOneTime(t *testing.T) {
// Get alert manager
am := alerts.NewAlertManager(hub)
defer am.Stop()
defer am.StopWorker()
// Test that alert is silenced
silenced := am.IsNotificationSilenced(user.Id, system.Id)
@@ -106,7 +106,7 @@ func TestAlertSilencedDaily(t *testing.T) {
// Get alert manager
am := alerts.NewAlertManager(hub)
defer am.Stop()
defer am.StopWorker()
// Get current hour and create a window that includes current time
now := time.Now().UTC()
@@ -170,7 +170,7 @@ func TestAlertSilencedDailyMidnightCrossing(t *testing.T) {
// Get alert manager
am := alerts.NewAlertManager(hub)
defer am.Stop()
defer am.StopWorker()
// Create a window that crosses midnight: 22:00 - 02:00
startTime := time.Date(2000, 1, 1, 22, 0, 0, 0, time.UTC)
@@ -211,7 +211,7 @@ func TestAlertSilencedGlobal(t *testing.T) {
// Get alert manager
am := alerts.NewAlertManager(hub)
defer am.Stop()
defer am.StopWorker()
// Create a global quiet hours window (no system specified)
now := time.Now().UTC()
@@ -250,7 +250,7 @@ func TestAlertSilencedSystemSpecific(t *testing.T) {
// Get alert manager
am := alerts.NewAlertManager(hub)
defer am.Stop()
defer am.StopWorker()
// Create a system-specific quiet hours window for system1 only
now := time.Now().UTC()
@@ -296,7 +296,7 @@ func TestAlertSilencedMultiUser(t *testing.T) {
// Get alert manager
am := alerts.NewAlertManager(hub)
defer am.Stop()
defer am.StopWorker()
// Create a quiet hours window for user1 only
now := time.Now().UTC()
@@ -417,7 +417,7 @@ func TestAlertSilencedNoWindows(t *testing.T) {
// Get alert manager
am := alerts.NewAlertManager(hub)
defer am.Stop()
defer am.StopWorker()
// Without any quiet hours windows, alert should NOT be silenced
silenced := am.IsNotificationSilenced(user.Id, system.Id)

View File

@@ -9,25 +9,63 @@ import (
"github.com/pocketbase/pocketbase/core"
)
type alertTask struct {
action string // "schedule" or "cancel"
systemName string
alertRecord *core.Record
delay time.Duration
}
type alertInfo struct {
systemName string
alertRecord *core.Record
expireTime time.Time
timer *time.Timer
}
// Stop cancels all pending status alert timers.
func (am *AlertManager) Stop() {
am.stopOnce.Do(func() {
am.pendingAlerts.Range(func(key, value any) bool {
info := value.(*alertInfo)
if info.timer != nil {
info.timer.Stop()
// startWorker is a long-running goroutine that processes alert tasks
// every x seconds. It must be running to process status alerts.
func (am *AlertManager) startWorker() {
processPendingAlerts := time.Tick(15 * time.Second)
// check for status alerts that are not resolved when system comes up
// (can be removed if we figure out core bug in #1052)
checkStatusAlerts := time.Tick(561 * time.Second)
for {
select {
case <-am.stopChan:
return
case task := <-am.alertQueue:
switch task.action {
case "schedule":
am.pendingAlerts.Store(task.alertRecord.Id, &alertInfo{
systemName: task.systemName,
alertRecord: task.alertRecord,
expireTime: time.Now().Add(task.delay),
})
case "cancel":
am.pendingAlerts.Delete(task.alertRecord.Id)
}
am.pendingAlerts.Delete(key)
return true
})
})
case <-checkStatusAlerts:
resolveStatusAlerts(am.hub)
case <-processPendingAlerts:
// Check for expired alerts every tick
now := time.Now()
for key, value := range am.pendingAlerts.Range {
info := value.(*alertInfo)
if now.After(info.expireTime) {
// Downtime delay has passed, process alert
am.sendStatusAlert("down", info.systemName, info.alertRecord)
am.pendingAlerts.Delete(key)
}
}
}
}
}
// StopWorker shuts down the AlertManager.worker goroutine
func (am *AlertManager) StopWorker() {
close(am.stopChan)
}
// HandleStatusAlerts manages the logic when system status changes.
@@ -65,82 +103,44 @@ func (am *AlertManager) getSystemStatusAlerts(systemID string) ([]*core.Record,
return alertRecords, nil
}
// handleSystemDown manages the logic when a system status changes to "down". It schedules pending alerts for each alert record.
// Schedules delayed "down" alerts for each alert record.
func (am *AlertManager) handleSystemDown(systemName string, alertRecords []*core.Record) {
for _, alertRecord := range alertRecords {
// Continue if alert is already scheduled
if _, exists := am.pendingAlerts.Load(alertRecord.Id); exists {
continue
}
// Schedule by adding to queue
min := max(1, alertRecord.GetInt("min"))
am.schedulePendingStatusAlert(systemName, alertRecord, time.Duration(min)*time.Minute)
am.alertQueue <- alertTask{
action: "schedule",
systemName: systemName,
alertRecord: alertRecord,
delay: time.Duration(min) * time.Minute,
}
}
}
// schedulePendingStatusAlert sets up a timer to send a "down" alert after the specified delay if the system is still down.
// It returns true if the alert was scheduled, or false if an alert was already pending for the given alert record.
func (am *AlertManager) schedulePendingStatusAlert(systemName string, alertRecord *core.Record, delay time.Duration) bool {
alert := &alertInfo{
systemName: systemName,
alertRecord: alertRecord,
expireTime: time.Now().Add(delay),
}
storedAlert, loaded := am.pendingAlerts.LoadOrStore(alertRecord.Id, alert)
if loaded {
return false
}
stored := storedAlert.(*alertInfo)
stored.timer = time.AfterFunc(time.Until(stored.expireTime), func() {
am.processPendingAlert(alertRecord.Id)
})
return true
}
// handleSystemUp manages the logic when a system status changes to "up".
// It cancels any pending alerts and sends "up" alerts.
func (am *AlertManager) handleSystemUp(systemName string, alertRecords []*core.Record) {
for _, alertRecord := range alertRecords {
alertRecordID := alertRecord.Id
// If alert exists for record, delete and continue (down alert not sent)
if am.cancelPendingAlert(alertRecord.Id) {
continue
}
if !alertRecord.GetBool("triggered") {
if _, exists := am.pendingAlerts.Load(alertRecordID); exists {
am.alertQueue <- alertTask{
action: "cancel",
alertRecord: alertRecord,
}
continue
}
// No alert scheduled for this record, send "up" alert
if err := am.sendStatusAlert("up", systemName, alertRecord); err != nil {
am.hub.Logger().Error("Failed to send alert", "err", err)
}
}
}
// cancelPendingAlert stops the timer and removes the pending alert for the given alert ID. Returns true if a pending alert was found and cancelled.
func (am *AlertManager) cancelPendingAlert(alertID string) bool {
value, loaded := am.pendingAlerts.LoadAndDelete(alertID)
if !loaded {
return false
}
info := value.(*alertInfo)
if info.timer != nil {
info.timer.Stop()
}
return true
}
// processPendingAlert sends a "down" alert if the pending alert has expired and the system is still down.
func (am *AlertManager) processPendingAlert(alertID string) {
value, loaded := am.pendingAlerts.LoadAndDelete(alertID)
if !loaded {
return
}
info := value.(*alertInfo)
if info.alertRecord.GetBool("triggered") {
return
}
if err := am.sendStatusAlert("down", info.systemName, info.alertRecord); err != nil {
am.hub.Logger().Error("Failed to send alert", "err", err)
}
}
// sendStatusAlert sends a status alert ("up" or "down") to the users associated with the alert records.
func (am *AlertManager) sendStatusAlert(alertStatus string, systemName string, alertRecord *core.Record) error {
switch alertStatus {
@@ -174,8 +174,8 @@ func (am *AlertManager) sendStatusAlert(alertStatus string, systemName string, a
})
}
// resolveStatusAlerts resolves any triggered status alerts that weren't resolved
// when system came up (https://github.com/henrygd/beszel/issues/1052).
// resolveStatusAlerts resolves any status alerts that weren't resolved
// when system came up (https://github.com/henrygd/beszel/issues/1052)
func resolveStatusAlerts(app core.App) error {
db := app.DB()
// Find all active status alerts where the system is actually up
@@ -205,36 +205,3 @@ func resolveStatusAlerts(app core.App) error {
}
return nil
}
// restorePendingStatusAlerts re-queues untriggered status alerts for systems that
// are still down after a hub restart. This rebuilds the lost in-memory timer state.
func (am *AlertManager) restorePendingStatusAlerts() error {
type pendingStatusAlert struct {
AlertID string `db:"alert_id"`
SystemName string `db:"system_name"`
}
var pending []pendingStatusAlert
err := am.hub.DB().NewQuery(`
SELECT a.id AS alert_id, s.name AS system_name
FROM alerts a
JOIN systems s ON a.system = s.id
WHERE a.name = 'Status'
AND a.triggered = false
AND s.status = 'down'
`).All(&pending)
if err != nil {
return err
}
for _, item := range pending {
alertRecord, err := am.hub.FindRecordById("alerts", item.AlertID)
if err != nil {
return err
}
min := max(1, alertRecord.GetInt("min"))
am.schedulePendingStatusAlert(item.SystemName, alertRecord, time.Duration(min)*time.Minute)
}
return nil
}

View File

@@ -1,628 +0,0 @@
//go:build testing
package alerts_test
import (
"testing"
"testing/synctest"
"time"
"github.com/henrygd/beszel/internal/alerts"
beszelTests "github.com/henrygd/beszel/internal/tests"
"github.com/pocketbase/dbx"
"github.com/pocketbase/pocketbase/core"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
func TestStatusAlerts(t *testing.T) {
synctest.Test(t, func(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
systems, err := beszelTests.CreateSystems(hub, 4, user.Id, "paused")
assert.NoError(t, err)
var alerts []*core.Record
for i, system := range systems {
alert, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": system.Id,
"user": user.Id,
"min": i + 1,
})
assert.NoError(t, err)
alerts = append(alerts, alert)
}
time.Sleep(10 * time.Millisecond)
for _, alert := range alerts {
assert.False(t, alert.GetBool("triggered"), "Alert should not be triggered immediately")
}
if hub.TestMailer.TotalSend() != 0 {
assert.Zero(t, hub.TestMailer.TotalSend(), "Expected 0 messages, got %d", hub.TestMailer.TotalSend())
}
for _, system := range systems {
assert.EqualValues(t, "paused", system.GetString("status"), "System should be paused")
}
for _, system := range systems {
system.Set("status", "up")
err = hub.SaveNoValidate(system)
assert.NoError(t, err)
}
time.Sleep(time.Second)
assert.EqualValues(t, 0, hub.GetPendingAlertsCount(), "should have 0 alerts in the pendingAlerts map")
for _, system := range systems {
system.Set("status", "down")
err = hub.SaveNoValidate(system)
assert.NoError(t, err)
}
// after 30 seconds, should have 4 alerts in the pendingAlerts map, no triggered alerts
time.Sleep(time.Second * 30)
assert.EqualValues(t, 4, hub.GetPendingAlertsCount(), "should have 4 alerts in the pendingAlerts map")
triggeredCount, err := hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.EqualValues(t, 0, triggeredCount, "should have 0 alert triggered")
assert.EqualValues(t, 0, hub.TestMailer.TotalSend(), "should have 0 messages sent")
// after 1:30 seconds, should have 1 triggered alert and 3 pending alerts
time.Sleep(time.Second * 60)
assert.EqualValues(t, 3, hub.GetPendingAlertsCount(), "should have 3 alerts in the pendingAlerts map")
triggeredCount, err = hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.EqualValues(t, 1, triggeredCount, "should have 1 alert triggered")
assert.EqualValues(t, 1, hub.TestMailer.TotalSend(), "should have 1 messages sent")
// after 2:30 seconds, should have 2 triggered alerts and 2 pending alerts
time.Sleep(time.Second * 60)
assert.EqualValues(t, 2, hub.GetPendingAlertsCount(), "should have 2 alerts in the pendingAlerts map")
triggeredCount, err = hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.EqualValues(t, 2, triggeredCount, "should have 2 alert triggered")
assert.EqualValues(t, 2, hub.TestMailer.TotalSend(), "should have 2 messages sent")
// now we will bring the remaning systems back up
for _, system := range systems {
system.Set("status", "up")
err = hub.SaveNoValidate(system)
assert.NoError(t, err)
}
time.Sleep(time.Second)
// should have 0 alerts in the pendingAlerts map and 0 alerts triggered
assert.EqualValues(t, 0, hub.GetPendingAlertsCount(), "should have 0 alerts in the pendingAlerts map")
triggeredCount, err = hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.Zero(t, triggeredCount, "should have 0 alert triggered")
// 4 messages sent, 2 down alerts and 2 up alerts for first 2 systems
assert.EqualValues(t, 4, hub.TestMailer.TotalSend(), "should have 4 messages sent")
})
}
func TestStatusAlertRecoveryBeforeDeadline(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
// Ensure user settings have an email
userSettings, _ := hub.FindFirstRecordByFilter("user_settings", "user={:user}", map[string]any{"user": user.Id})
userSettings.Set("settings", `{"emails":["test@example.com"],"webhooks":[]}`)
hub.Save(userSettings)
// Initial email count
initialEmailCount := hub.TestMailer.TotalSend()
systemCollection, _ := hub.FindCollectionByNameOrId("systems")
system := core.NewRecord(systemCollection)
system.Set("name", "test-system")
system.Set("status", "up")
system.Set("host", "127.0.0.1")
system.Set("users", []string{user.Id})
hub.Save(system)
alertCollection, _ := hub.FindCollectionByNameOrId("alerts")
alert := core.NewRecord(alertCollection)
alert.Set("user", user.Id)
alert.Set("system", system.Id)
alert.Set("name", "Status")
alert.Set("triggered", false)
alert.Set("min", 1)
hub.Save(alert)
am := hub.AlertManager
// 1. System goes down
am.HandleStatusAlerts("down", system)
assert.Equal(t, 1, am.GetPendingAlertsCount(), "Alert should be scheduled")
// 2. System goes up BEFORE delay expires
// Triggering HandleStatusAlerts("up") SHOULD NOT send an alert.
am.HandleStatusAlerts("up", system)
assert.Equal(t, 0, am.GetPendingAlertsCount(), "Alert should be canceled if system recovers before delay expires")
// Verify that NO email was sent.
assert.Equal(t, initialEmailCount, hub.TestMailer.TotalSend(), "Recovery notification should not be sent if system never went down")
}
func TestStatusAlertNormalRecovery(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
// Ensure user settings have an email
userSettings, _ := hub.FindFirstRecordByFilter("user_settings", "user={:user}", map[string]any{"user": user.Id})
userSettings.Set("settings", `{"emails":["test@example.com"],"webhooks":[]}`)
hub.Save(userSettings)
systemCollection, _ := hub.FindCollectionByNameOrId("systems")
system := core.NewRecord(systemCollection)
system.Set("name", "test-system")
system.Set("status", "up")
system.Set("host", "127.0.0.1")
system.Set("users", []string{user.Id})
hub.Save(system)
alertCollection, _ := hub.FindCollectionByNameOrId("alerts")
alert := core.NewRecord(alertCollection)
alert.Set("user", user.Id)
alert.Set("system", system.Id)
alert.Set("name", "Status")
alert.Set("triggered", true) // System was confirmed DOWN
hub.Save(alert)
am := hub.AlertManager
initialEmailCount := hub.TestMailer.TotalSend()
// System goes up
am.HandleStatusAlerts("up", system)
// Verify that an email WAS sent (normal recovery).
assert.Equal(t, initialEmailCount+1, hub.TestMailer.TotalSend(), "Recovery notification should be sent if system was triggered as down")
}
func TestHandleStatusAlertsDoesNotSendRecoveryWhileDownIsOnlyPending(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
userSettings, err := hub.FindFirstRecordByFilter("user_settings", "user={:user}", map[string]any{"user": user.Id})
require.NoError(t, err)
userSettings.Set("settings", `{"emails":["test@example.com"],"webhooks":[]}`)
require.NoError(t, hub.Save(userSettings))
systemCollection, err := hub.FindCollectionByNameOrId("systems")
require.NoError(t, err)
system := core.NewRecord(systemCollection)
system.Set("name", "test-system")
system.Set("status", "up")
system.Set("host", "127.0.0.1")
system.Set("users", []string{user.Id})
require.NoError(t, hub.Save(system))
alertCollection, err := hub.FindCollectionByNameOrId("alerts")
require.NoError(t, err)
alert := core.NewRecord(alertCollection)
alert.Set("user", user.Id)
alert.Set("system", system.Id)
alert.Set("name", "Status")
alert.Set("triggered", false)
alert.Set("min", 1)
require.NoError(t, hub.Save(alert))
initialEmailCount := hub.TestMailer.TotalSend()
am := alerts.NewTestAlertManagerWithoutWorker(hub)
require.NoError(t, am.HandleStatusAlerts("down", system))
assert.Equal(t, 1, am.GetPendingAlertsCount(), "down transition should register a pending alert immediately")
require.NoError(t, am.HandleStatusAlerts("up", system))
assert.Zero(t, am.GetPendingAlertsCount(), "recovery should cancel the pending down alert")
assert.Equal(t, initialEmailCount, hub.TestMailer.TotalSend(), "recovery notification should not be sent before a down alert triggers")
alertRecord, err := hub.FindRecordById("alerts", alert.Id)
require.NoError(t, err)
assert.False(t, alertRecord.GetBool("triggered"), "alert should remain untriggered when downtime never matured")
}
func TestStatusAlertTimerCancellationPreventsBoundaryDelivery(t *testing.T) {
synctest.Test(t, func(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
userSettings, err := hub.FindFirstRecordByFilter("user_settings", "user={:user}", map[string]any{"user": user.Id})
require.NoError(t, err)
userSettings.Set("settings", `{"emails":["test@example.com"],"webhooks":[]}`)
require.NoError(t, hub.Save(userSettings))
systemCollection, err := hub.FindCollectionByNameOrId("systems")
require.NoError(t, err)
system := core.NewRecord(systemCollection)
system.Set("name", "test-system")
system.Set("status", "up")
system.Set("host", "127.0.0.1")
system.Set("users", []string{user.Id})
require.NoError(t, hub.Save(system))
alertCollection, err := hub.FindCollectionByNameOrId("alerts")
require.NoError(t, err)
alert := core.NewRecord(alertCollection)
alert.Set("user", user.Id)
alert.Set("system", system.Id)
alert.Set("name", "Status")
alert.Set("triggered", false)
alert.Set("min", 1)
require.NoError(t, hub.Save(alert))
initialEmailCount := hub.TestMailer.TotalSend()
am := alerts.NewTestAlertManagerWithoutWorker(hub)
require.NoError(t, am.HandleStatusAlerts("down", system))
assert.Equal(t, 1, am.GetPendingAlertsCount(), "down transition should register a pending alert immediately")
require.True(t, am.ResetPendingAlertTimer(alert.Id, 25*time.Millisecond), "test should shorten the pending alert timer")
time.Sleep(10 * time.Millisecond)
require.NoError(t, am.HandleStatusAlerts("up", system))
assert.Zero(t, am.GetPendingAlertsCount(), "recovery should remove the pending alert before the timer callback runs")
time.Sleep(40 * time.Millisecond)
assert.Equal(t, initialEmailCount, hub.TestMailer.TotalSend(), "timer callback should not deliver after recovery cancels the pending alert")
alertRecord, err := hub.FindRecordById("alerts", alert.Id)
require.NoError(t, err)
assert.False(t, alertRecord.GetBool("triggered"), "alert should remain untriggered when cancellation wins the timer race")
time.Sleep(time.Minute)
synctest.Wait()
})
}
func TestStatusAlertDownFiresAfterDelayExpires(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
userSettings, err := hub.FindFirstRecordByFilter("user_settings", "user={:user}", map[string]any{"user": user.Id})
require.NoError(t, err)
userSettings.Set("settings", `{"emails":["test@example.com"],"webhooks":[]}`)
require.NoError(t, hub.Save(userSettings))
systemCollection, err := hub.FindCollectionByNameOrId("systems")
require.NoError(t, err)
system := core.NewRecord(systemCollection)
system.Set("name", "test-system")
system.Set("status", "up")
system.Set("host", "127.0.0.1")
system.Set("users", []string{user.Id})
require.NoError(t, hub.Save(system))
alertCollection, err := hub.FindCollectionByNameOrId("alerts")
require.NoError(t, err)
alert := core.NewRecord(alertCollection)
alert.Set("user", user.Id)
alert.Set("system", system.Id)
alert.Set("name", "Status")
alert.Set("triggered", false)
alert.Set("min", 1)
require.NoError(t, hub.Save(alert))
initialEmailCount := hub.TestMailer.TotalSend()
am := alerts.NewTestAlertManagerWithoutWorker(hub)
require.NoError(t, am.HandleStatusAlerts("down", system))
assert.Equal(t, 1, am.GetPendingAlertsCount(), "alert should be pending after system goes down")
// Expire the pending alert and process it
am.ForceExpirePendingAlerts()
processed, err := am.ProcessPendingAlerts()
require.NoError(t, err)
assert.Len(t, processed, 1, "one alert should have been processed")
assert.Equal(t, 0, am.GetPendingAlertsCount(), "pending alert should be consumed after processing")
// Verify down email was sent
assert.Equal(t, initialEmailCount+1, hub.TestMailer.TotalSend(), "down notification should be sent after delay expires")
// Verify triggered flag is set in the DB
alertRecord, err := hub.FindRecordById("alerts", alert.Id)
require.NoError(t, err)
assert.True(t, alertRecord.GetBool("triggered"), "alert should be marked triggered after downtime matures")
}
func TestStatusAlertDuplicateDownCallIsIdempotent(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
userSettings, err := hub.FindFirstRecordByFilter("user_settings", "user={:user}", map[string]any{"user": user.Id})
require.NoError(t, err)
userSettings.Set("settings", `{"emails":["test@example.com"],"webhooks":[]}`)
require.NoError(t, hub.Save(userSettings))
systemCollection, err := hub.FindCollectionByNameOrId("systems")
require.NoError(t, err)
system := core.NewRecord(systemCollection)
system.Set("name", "test-system")
system.Set("status", "up")
system.Set("host", "127.0.0.1")
system.Set("users", []string{user.Id})
require.NoError(t, hub.Save(system))
alertCollection, err := hub.FindCollectionByNameOrId("alerts")
require.NoError(t, err)
alert := core.NewRecord(alertCollection)
alert.Set("user", user.Id)
alert.Set("system", system.Id)
alert.Set("name", "Status")
alert.Set("triggered", false)
alert.Set("min", 5)
require.NoError(t, hub.Save(alert))
am := alerts.NewTestAlertManagerWithoutWorker(hub)
require.NoError(t, am.HandleStatusAlerts("down", system))
require.NoError(t, am.HandleStatusAlerts("down", system))
require.NoError(t, am.HandleStatusAlerts("down", system))
assert.Equal(t, 1, am.GetPendingAlertsCount(), "repeated down calls should not schedule duplicate pending alerts")
}
func TestStatusAlertNoAlertRecord(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
systemCollection, err := hub.FindCollectionByNameOrId("systems")
require.NoError(t, err)
system := core.NewRecord(systemCollection)
system.Set("name", "test-system")
system.Set("status", "up")
system.Set("host", "127.0.0.1")
system.Set("users", []string{user.Id})
require.NoError(t, hub.Save(system))
// No Status alert record created for this system
initialEmailCount := hub.TestMailer.TotalSend()
am := alerts.NewTestAlertManagerWithoutWorker(hub)
require.NoError(t, am.HandleStatusAlerts("down", system))
assert.Equal(t, 0, am.GetPendingAlertsCount(), "no pending alert when no alert record exists")
require.NoError(t, am.HandleStatusAlerts("up", system))
assert.Equal(t, initialEmailCount, hub.TestMailer.TotalSend(), "no email when no alert record exists")
}
func TestRestorePendingStatusAlertsRequeuesDownSystemsAfterRestart(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
userSettings, err := hub.FindFirstRecordByFilter("user_settings", "user={:user}", map[string]any{"user": user.Id})
require.NoError(t, err)
userSettings.Set("settings", `{"emails":["test@example.com"],"webhooks":[]}`)
require.NoError(t, hub.Save(userSettings))
systems, err := beszelTests.CreateSystems(hub, 1, user.Id, "down")
require.NoError(t, err)
system := systems[0]
alertCollection, err := hub.FindCollectionByNameOrId("alerts")
require.NoError(t, err)
alert := core.NewRecord(alertCollection)
alert.Set("user", user.Id)
alert.Set("system", system.Id)
alert.Set("name", "Status")
alert.Set("triggered", false)
alert.Set("min", 1)
require.NoError(t, hub.Save(alert))
initialEmailCount := hub.TestMailer.TotalSend()
am := alerts.NewTestAlertManagerWithoutWorker(hub)
require.NoError(t, am.RestorePendingStatusAlerts())
assert.Equal(t, 1, am.GetPendingAlertsCount(), "startup restore should requeue a pending down alert for a system still marked down")
am.ForceExpirePendingAlerts()
processed, err := am.ProcessPendingAlerts()
require.NoError(t, err)
assert.Len(t, processed, 1, "restored pending alert should be processable after the delay expires")
assert.Equal(t, initialEmailCount+1, hub.TestMailer.TotalSend(), "restored pending alert should send the down notification")
alertRecord, err := hub.FindRecordById("alerts", alert.Id)
require.NoError(t, err)
assert.True(t, alertRecord.GetBool("triggered"), "restored pending alert should mark the alert as triggered once delivered")
}
func TestRestorePendingStatusAlertsSkipsNonDownOrAlreadyTriggeredAlerts(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
systemsDown, err := beszelTests.CreateSystems(hub, 2, user.Id, "down")
require.NoError(t, err)
systemDownPending := systemsDown[0]
systemDownTriggered := systemsDown[1]
systemUp, err := beszelTests.CreateRecord(hub, "systems", map[string]any{
"name": "up-system",
"users": []string{user.Id},
"host": "127.0.0.2",
"status": "up",
})
require.NoError(t, err)
_, err = beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": systemDownPending.Id,
"user": user.Id,
"min": 1,
"triggered": false,
})
require.NoError(t, err)
_, err = beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": systemUp.Id,
"user": user.Id,
"min": 1,
"triggered": false,
})
require.NoError(t, err)
_, err = beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": systemDownTriggered.Id,
"user": user.Id,
"min": 1,
"triggered": true,
})
require.NoError(t, err)
am := alerts.NewTestAlertManagerWithoutWorker(hub)
require.NoError(t, am.RestorePendingStatusAlerts())
assert.Equal(t, 1, am.GetPendingAlertsCount(), "only untriggered alerts for currently down systems should be restored")
}
func TestRestorePendingStatusAlertsIsIdempotent(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
systems, err := beszelTests.CreateSystems(hub, 1, user.Id, "down")
require.NoError(t, err)
system := systems[0]
_, err = beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": system.Id,
"user": user.Id,
"min": 1,
"triggered": false,
})
require.NoError(t, err)
am := alerts.NewTestAlertManagerWithoutWorker(hub)
require.NoError(t, am.RestorePendingStatusAlerts())
require.NoError(t, am.RestorePendingStatusAlerts())
assert.Equal(t, 1, am.GetPendingAlertsCount(), "restoring twice should not create duplicate pending alerts")
am.ForceExpirePendingAlerts()
processed, err := am.ProcessPendingAlerts()
require.NoError(t, err)
assert.Len(t, processed, 1, "restored alert should still be processable exactly once")
assert.Zero(t, am.GetPendingAlertsCount(), "processing the restored alert should empty the pending map")
}
func TestResolveStatusAlertsFixesStaleTriggered(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
// CreateSystems uses SaveNoValidate after initial save to bypass the
// onRecordCreate hook that forces status = "pending".
systems, err := beszelTests.CreateSystems(hub, 1, user.Id, "up")
require.NoError(t, err)
system := systems[0]
alertCollection, err := hub.FindCollectionByNameOrId("alerts")
require.NoError(t, err)
alert := core.NewRecord(alertCollection)
alert.Set("user", user.Id)
alert.Set("system", system.Id)
alert.Set("name", "Status")
alert.Set("triggered", true) // Stale: system is up but alert still says triggered
require.NoError(t, hub.Save(alert))
// resolveStatusAlerts should clear the stale triggered flag
require.NoError(t, alerts.ResolveStatusAlerts(hub))
alertRecord, err := hub.FindRecordById("alerts", alert.Id)
require.NoError(t, err)
assert.False(t, alertRecord.GetBool("triggered"), "stale triggered flag should be cleared when system is up")
}
func TestResolveStatusAlerts(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
// Create a systemUp
systemUp, err := beszelTests.CreateRecord(hub, "systems", map[string]any{
"name": "test-system",
"users": []string{user.Id},
"host": "127.0.0.1",
"status": "up",
})
assert.NoError(t, err)
systemDown, err := beszelTests.CreateRecord(hub, "systems", map[string]any{
"name": "test-system-2",
"users": []string{user.Id},
"host": "127.0.0.2",
"status": "up",
})
assert.NoError(t, err)
// Create a status alertUp for the system
alertUp, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": systemUp.Id,
"user": user.Id,
"min": 1,
})
assert.NoError(t, err)
alertDown, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": systemDown.Id,
"user": user.Id,
"min": 1,
})
assert.NoError(t, err)
// Verify alert is not triggered initially
assert.False(t, alertUp.GetBool("triggered"), "Alert should not be triggered initially")
// Set the system to 'up' (this should not trigger the alert)
systemUp.Set("status", "up")
err = hub.SaveNoValidate(systemUp)
assert.NoError(t, err)
systemDown.Set("status", "down")
err = hub.SaveNoValidate(systemDown)
assert.NoError(t, err)
// Wait a moment for any processing
time.Sleep(10 * time.Millisecond)
// Verify alertUp is still not triggered after setting system to up
alertUp, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": alertUp.Id})
assert.NoError(t, err)
assert.False(t, alertUp.GetBool("triggered"), "Alert should not be triggered when system is up")
// Manually set both alerts triggered to true
alertUp.Set("triggered", true)
err = hub.SaveNoValidate(alertUp)
assert.NoError(t, err)
alertDown.Set("triggered", true)
err = hub.SaveNoValidate(alertDown)
assert.NoError(t, err)
// Verify we have exactly one alert with triggered true
triggeredCount, err := hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.EqualValues(t, 2, triggeredCount, "Should have exactly two alerts with triggered true")
// Verify the specific alertUp is triggered
alertUp, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": alertUp.Id})
assert.NoError(t, err)
assert.True(t, alertUp.GetBool("triggered"), "Alert should be triggered")
// Verify we have two unresolved alert history records
alertHistoryCount, err := hub.CountRecords("alerts_history", dbx.HashExp{"resolved": ""})
assert.NoError(t, err)
assert.EqualValues(t, 2, alertHistoryCount, "Should have exactly two unresolved alert history records")
err = alerts.ResolveStatusAlerts(hub)
assert.NoError(t, err)
// Verify alertUp is not triggered after resolving
alertUp, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": alertUp.Id})
assert.NoError(t, err)
assert.False(t, alertUp.GetBool("triggered"), "Alert should not be triggered after resolving")
// Verify alertDown is still triggered
alertDown, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": alertDown.Id})
assert.NoError(t, err)
assert.True(t, alertDown.GetBool("triggered"), "Alert should still be triggered after resolving")
// Verify we have one unresolved alert history record
alertHistoryCount, err = hub.CountRecords("alerts_history", dbx.HashExp{"resolved": ""})
assert.NoError(t, err)
assert.EqualValues(t, 1, alertHistoryCount, "Should have exactly one unresolved alert history record")
}

View File

@@ -11,6 +11,7 @@ import (
"github.com/pocketbase/dbx"
"github.com/pocketbase/pocketbase/core"
"github.com/pocketbase/pocketbase/tools/types"
"github.com/spf13/cast"
)
func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *system.CombinedData) error {
@@ -37,7 +38,7 @@ func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *syst
case "Memory":
val = data.Info.MemPct
case "Bandwidth":
val = float64(data.Info.BandwidthBytes) / (1024 * 1024)
val = data.Info.Bandwidth
unit = " MB/s"
case "Disk":
maxUsedPct := data.Info.DiskPct
@@ -91,7 +92,7 @@ func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *syst
}
}
min := max(1, uint8(alertRecord.GetInt("min")))
min := max(1, cast.ToUint8(alertRecord.Get("min")))
alert := SystemAlertData{
systemRecord: systemRecord,
@@ -191,24 +192,22 @@ func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *syst
case "Memory":
alert.val += stats.Mem
case "Bandwidth":
alert.val += float64(stats.Bandwidth[0]+stats.Bandwidth[1]) / (1024 * 1024)
alert.val += stats.NetSent + stats.NetRecv
case "Disk":
if alert.mapSums == nil {
alert.mapSums = make(map[string]float32, len(stats.ExtraFs)+1)
alert.mapSums = make(map[string]float32, len(data.Stats.ExtraFs)+1)
}
// add root disk
if _, ok := alert.mapSums["root"]; !ok {
alert.mapSums["root"] = 0.0
}
alert.mapSums["root"] += float32(stats.Disk)
// add extra disks from historical record
for key, fs := range stats.ExtraFs {
if fs.DiskTotal > 0 {
if _, ok := alert.mapSums[key]; !ok {
alert.mapSums[key] = 0.0
}
alert.mapSums[key] += float32(fs.DiskUsed / fs.DiskTotal * 100)
// add extra disks
for key, fs := range data.Stats.ExtraFs {
if _, ok := alert.mapSums[key]; !ok {
alert.mapSums[key] = 0.0
}
alert.mapSums[key] += float32(fs.DiskUsed / fs.DiskTotal * 100)
}
case "Temperature":
if alert.mapSums == nil {

View File

@@ -12,6 +12,7 @@ import (
"testing/synctest"
"time"
"github.com/henrygd/beszel/internal/alerts"
beszelTests "github.com/henrygd/beszel/internal/tests"
"github.com/pocketbase/dbx"
@@ -368,6 +369,87 @@ func TestUserAlertsApi(t *testing.T) {
}
}
func TestStatusAlerts(t *testing.T) {
synctest.Test(t, func(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
systems, err := beszelTests.CreateSystems(hub, 4, user.Id, "paused")
assert.NoError(t, err)
var alerts []*core.Record
for i, system := range systems {
alert, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": system.Id,
"user": user.Id,
"min": i + 1,
})
assert.NoError(t, err)
alerts = append(alerts, alert)
}
time.Sleep(10 * time.Millisecond)
for _, alert := range alerts {
assert.False(t, alert.GetBool("triggered"), "Alert should not be triggered immediately")
}
if hub.TestMailer.TotalSend() != 0 {
assert.Zero(t, hub.TestMailer.TotalSend(), "Expected 0 messages, got %d", hub.TestMailer.TotalSend())
}
for _, system := range systems {
assert.EqualValues(t, "paused", system.GetString("status"), "System should be paused")
}
for _, system := range systems {
system.Set("status", "up")
err = hub.SaveNoValidate(system)
assert.NoError(t, err)
}
time.Sleep(time.Second)
assert.EqualValues(t, 0, hub.GetPendingAlertsCount(), "should have 0 alerts in the pendingAlerts map")
for _, system := range systems {
system.Set("status", "down")
err = hub.SaveNoValidate(system)
assert.NoError(t, err)
}
// after 30 seconds, should have 4 alerts in the pendingAlerts map, no triggered alerts
time.Sleep(time.Second * 30)
assert.EqualValues(t, 4, hub.GetPendingAlertsCount(), "should have 4 alerts in the pendingAlerts map")
triggeredCount, err := hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.EqualValues(t, 0, triggeredCount, "should have 0 alert triggered")
assert.EqualValues(t, 0, hub.TestMailer.TotalSend(), "should have 0 messages sent")
// after 1:30 seconds, should have 1 triggered alert and 3 pending alerts
time.Sleep(time.Second * 60)
assert.EqualValues(t, 3, hub.GetPendingAlertsCount(), "should have 3 alerts in the pendingAlerts map")
triggeredCount, err = hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.EqualValues(t, 1, triggeredCount, "should have 1 alert triggered")
assert.EqualValues(t, 1, hub.TestMailer.TotalSend(), "should have 1 messages sent")
// after 2:30 seconds, should have 2 triggered alerts and 2 pending alerts
time.Sleep(time.Second * 60)
assert.EqualValues(t, 2, hub.GetPendingAlertsCount(), "should have 2 alerts in the pendingAlerts map")
triggeredCount, err = hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.EqualValues(t, 2, triggeredCount, "should have 2 alert triggered")
assert.EqualValues(t, 2, hub.TestMailer.TotalSend(), "should have 2 messages sent")
// now we will bring the remaning systems back up
for _, system := range systems {
system.Set("status", "up")
err = hub.SaveNoValidate(system)
assert.NoError(t, err)
}
time.Sleep(time.Second)
// should have 0 alerts in the pendingAlerts map and 0 alerts triggered
assert.EqualValues(t, 0, hub.GetPendingAlertsCount(), "should have 0 alerts in the pendingAlerts map")
triggeredCount, err = hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.Zero(t, triggeredCount, "should have 0 alert triggered")
// 4 messages sent, 2 down alerts and 2 up alerts for first 2 systems
assert.EqualValues(t, 4, hub.TestMailer.TotalSend(), "should have 4 messages sent")
})
}
func TestAlertsHistory(t *testing.T) {
synctest.Test(t, func(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
@@ -496,3 +578,102 @@ func TestAlertsHistory(t *testing.T) {
assert.EqualValues(t, 2, totalHistoryCount, "Should have 2 total alert history records")
})
}
func TestResolveStatusAlerts(t *testing.T) {
hub, user := beszelTests.GetHubWithUser(t)
defer hub.Cleanup()
// Create a systemUp
systemUp, err := beszelTests.CreateRecord(hub, "systems", map[string]any{
"name": "test-system",
"users": []string{user.Id},
"host": "127.0.0.1",
"status": "up",
})
assert.NoError(t, err)
systemDown, err := beszelTests.CreateRecord(hub, "systems", map[string]any{
"name": "test-system-2",
"users": []string{user.Id},
"host": "127.0.0.2",
"status": "up",
})
assert.NoError(t, err)
// Create a status alertUp for the system
alertUp, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": systemUp.Id,
"user": user.Id,
"min": 1,
})
assert.NoError(t, err)
alertDown, err := beszelTests.CreateRecord(hub, "alerts", map[string]any{
"name": "Status",
"system": systemDown.Id,
"user": user.Id,
"min": 1,
})
assert.NoError(t, err)
// Verify alert is not triggered initially
assert.False(t, alertUp.GetBool("triggered"), "Alert should not be triggered initially")
// Set the system to 'up' (this should not trigger the alert)
systemUp.Set("status", "up")
err = hub.SaveNoValidate(systemUp)
assert.NoError(t, err)
systemDown.Set("status", "down")
err = hub.SaveNoValidate(systemDown)
assert.NoError(t, err)
// Wait a moment for any processing
time.Sleep(10 * time.Millisecond)
// Verify alertUp is still not triggered after setting system to up
alertUp, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": alertUp.Id})
assert.NoError(t, err)
assert.False(t, alertUp.GetBool("triggered"), "Alert should not be triggered when system is up")
// Manually set both alerts triggered to true
alertUp.Set("triggered", true)
err = hub.SaveNoValidate(alertUp)
assert.NoError(t, err)
alertDown.Set("triggered", true)
err = hub.SaveNoValidate(alertDown)
assert.NoError(t, err)
// Verify we have exactly one alert with triggered true
triggeredCount, err := hub.CountRecords("alerts", dbx.HashExp{"triggered": true})
assert.NoError(t, err)
assert.EqualValues(t, 2, triggeredCount, "Should have exactly two alerts with triggered true")
// Verify the specific alertUp is triggered
alertUp, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": alertUp.Id})
assert.NoError(t, err)
assert.True(t, alertUp.GetBool("triggered"), "Alert should be triggered")
// Verify we have two unresolved alert history records
alertHistoryCount, err := hub.CountRecords("alerts_history", dbx.HashExp{"resolved": ""})
assert.NoError(t, err)
assert.EqualValues(t, 2, alertHistoryCount, "Should have exactly two unresolved alert history records")
err = alerts.ResolveStatusAlerts(hub)
assert.NoError(t, err)
// Verify alertUp is not triggered after resolving
alertUp, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": alertUp.Id})
assert.NoError(t, err)
assert.False(t, alertUp.GetBool("triggered"), "Alert should not be triggered after resolving")
// Verify alertDown is still triggered
alertDown, err = hub.FindFirstRecordByFilter("alerts", "id={:id}", dbx.Params{"id": alertDown.Id})
assert.NoError(t, err)
assert.True(t, alertDown.GetBool("triggered"), "Alert should still be triggered after resolving")
// Verify we have one unresolved alert history record
alertHistoryCount, err = hub.CountRecords("alerts_history", dbx.HashExp{"resolved": ""})
assert.NoError(t, err)
assert.EqualValues(t, 1, alertHistoryCount, "Should have exactly one unresolved alert history record")
}

View File

@@ -9,12 +9,6 @@ import (
"github.com/pocketbase/pocketbase/core"
)
func NewTestAlertManagerWithoutWorker(app hubLike) *AlertManager {
return &AlertManager{
hub: app,
}
}
func (am *AlertManager) GetAlertManager() *AlertManager {
return am
}
@@ -40,11 +34,12 @@ func (am *AlertManager) ProcessPendingAlerts() ([]*core.Record, error) {
am.pendingAlerts.Range(func(key, value any) bool {
info := value.(*alertInfo)
if now.After(info.expireTime) {
if info.timer != nil {
info.timer.Stop()
// Downtime delay has passed, process alert
if err := am.sendStatusAlert("down", info.systemName, info.alertRecord); err != nil {
lastErr = err
}
am.processPendingAlert(key.(string))
processedAlerts = append(processedAlerts, info.alertRecord)
am.pendingAlerts.Delete(key)
}
return true
})
@@ -61,27 +56,6 @@ func (am *AlertManager) ForceExpirePendingAlerts() {
})
}
func (am *AlertManager) ResetPendingAlertTimer(alertID string, delay time.Duration) bool {
value, loaded := am.pendingAlerts.Load(alertID)
if !loaded {
return false
}
info := value.(*alertInfo)
if info.timer != nil {
info.timer.Stop()
}
info.expireTime = time.Now().Add(delay)
info.timer = time.AfterFunc(delay, func() {
am.processPendingAlert(alertID)
})
return true
}
func ResolveStatusAlerts(app core.App) error {
return resolveStatusAlerts(app)
}
func (am *AlertManager) RestorePendingStatusAlerts() error {
return am.restorePendingStatusAlerts()
}

View File

@@ -9,7 +9,6 @@ import (
"github.com/henrygd/beszel"
"github.com/henrygd/beszel/agent"
"github.com/henrygd/beszel/agent/health"
"github.com/henrygd/beszel/agent/utils"
"github.com/spf13/pflag"
"golang.org/x/crypto/ssh"
)
@@ -117,12 +116,12 @@ func (opts *cmdOptions) loadPublicKeys() ([]ssh.PublicKey, error) {
}
// Try environment variable
if key, ok := utils.GetEnv("KEY"); ok && key != "" {
if key, ok := agent.GetEnv("KEY"); ok && key != "" {
return agent.ParseKeys(key)
}
// Try key file
keyFile, ok := utils.GetEnv("KEY_FILE")
keyFile, ok := agent.GetEnv("KEY_FILE")
if !ok {
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")
}

View File

@@ -10,19 +10,10 @@ type ApiInfo struct {
Status string
State string
Image string
Health struct {
Status string
// FailingStreak int
}
Ports []struct {
// PrivatePort uint16
PublicPort uint16
// IP string
// Type string
}
// ImageID string
// Command string
// Created int64
// Ports []Port
// SizeRw int64 `json:",omitempty"`
// SizeRootFs int64 `json:",omitempty"`
// Labels map[string]string
@@ -149,7 +140,6 @@ type Stats struct {
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:"-"`

View File

@@ -143,8 +143,8 @@ type AtaDeviceStatisticsPage struct {
}
type AtaDeviceStatisticsEntry struct {
Name string `json:"name"`
Value *int64 `json:"value,omitempty"`
Name string `json:"name"`
Value *uint64 `json:"value,omitempty"`
}
type AtaSmartAttribute struct {
@@ -356,8 +356,8 @@ type SmartInfoForSata struct {
SmartStatus SmartStatusInfo `json:"smart_status"`
// AtaSmartData AtaSmartData `json:"ata_smart_data"`
// AtaSctCapabilities AtaSctCapabilities `json:"ata_sct_capabilities"`
AtaSmartAttributes AtaSmartAttributes `json:"ata_smart_attributes"`
AtaDeviceStatistics json.RawMessage `json:"ata_device_statistics"`
AtaSmartAttributes AtaSmartAttributes `json:"ata_smart_attributes"`
AtaDeviceStatistics AtaDeviceStatistics `json:"ata_device_statistics"`
// PowerOnTime PowerOnTimeInfo `json:"power_on_time"`
// PowerCycleCount uint16 `json:"power_cycle_count"`
Temperature TemperatureInfo `json:"temperature"`

View File

@@ -12,9 +12,8 @@ import (
type Stats struct {
Cpu float64 `json:"cpu" cbor:"0,keyasint"`
MaxCpu float64 `json:"cpum,omitempty" cbor:"-"`
MaxCpu float64 `json:"cpum,omitempty" cbor:"1,keyasint,omitempty"`
Mem float64 `json:"m" cbor:"2,keyasint"`
MaxMem float64 `json:"mm,omitempty" cbor:"-"`
MemUsed float64 `json:"mu" cbor:"3,keyasint"`
MemPct float64 `json:"mp" cbor:"4,keyasint"`
MemBuffCache float64 `json:"mb" cbor:"5,keyasint"`
@@ -24,25 +23,26 @@ type Stats struct {
DiskTotal float64 `json:"d" cbor:"9,keyasint"`
DiskUsed float64 `json:"du" cbor:"10,keyasint"`
DiskPct float64 `json:"dp" cbor:"11,keyasint"`
DiskReadPs float64 `json:"dr,omitzero" cbor:"12,keyasint,omitzero"`
DiskWritePs float64 `json:"dw,omitzero" cbor:"13,keyasint,omitzero"`
MaxDiskReadPs float64 `json:"drm,omitempty" cbor:"-"`
MaxDiskWritePs float64 `json:"dwm,omitempty" cbor:"-"`
DiskReadPs float64 `json:"dr" cbor:"12,keyasint"`
DiskWritePs float64 `json:"dw" cbor:"13,keyasint"`
MaxDiskReadPs float64 `json:"drm,omitempty" cbor:"14,keyasint,omitempty"`
MaxDiskWritePs float64 `json:"dwm,omitempty" cbor:"15,keyasint,omitempty"`
NetworkSent float64 `json:"ns,omitzero" cbor:"16,keyasint,omitzero"`
NetworkRecv float64 `json:"nr,omitzero" cbor:"17,keyasint,omitzero"`
MaxNetworkSent float64 `json:"nsm,omitempty" cbor:"-"`
MaxNetworkRecv float64 `json:"nrm,omitempty" cbor:"-"`
MaxNetworkSent float64 `json:"nsm,omitempty" cbor:"18,keyasint,omitempty"`
MaxNetworkRecv float64 `json:"nrm,omitempty" cbor:"19,keyasint,omitempty"`
Temperatures map[string]float64 `json:"t,omitempty" cbor:"20,keyasint,omitempty"`
ExtraFs map[string]*FsStats `json:"efs,omitempty" cbor:"21,keyasint,omitempty"`
GPUData map[string]GPUData `json:"g,omitempty" cbor:"22,keyasint,omitempty"`
// LoadAvg1 float64 `json:"l1,omitempty" cbor:"23,keyasint,omitempty"`
// LoadAvg5 float64 `json:"l5,omitempty" cbor:"24,keyasint,omitempty"`
// LoadAvg15 float64 `json:"l15,omitempty" cbor:"25,keyasint,omitempty"`
Bandwidth [2]uint64 `json:"b,omitzero" cbor:"26,keyasint,omitzero"` // [sent bytes, recv bytes]
MaxBandwidth [2]uint64 `json:"bm,omitzero" cbor:"-"` // [sent bytes, recv bytes]
LoadAvg1 float64 `json:"l1,omitempty" cbor:"23,keyasint,omitempty"`
LoadAvg5 float64 `json:"l5,omitempty" cbor:"24,keyasint,omitempty"`
LoadAvg15 float64 `json:"l15,omitempty" cbor:"25,keyasint,omitempty"`
Bandwidth [2]uint64 `json:"b,omitzero" cbor:"26,keyasint,omitzero"` // [sent bytes, recv bytes]
MaxBandwidth [2]uint64 `json:"bm,omitzero" cbor:"27,keyasint,omitzero"` // [sent bytes, recv bytes]
// TODO: remove other load fields in future release in favor of load avg array
LoadAvg [3]float64 `json:"la,omitempty" cbor:"28,keyasint"`
Battery [2]uint8 `json:"bat,omitzero" cbor:"29,keyasint,omitzero"` // [percent, charge state, current]
Battery [2]uint8 `json:"bat,omitzero" cbor:"29,keyasint,omitzero"` // [percent, charge state, current]
MaxMem float64 `json:"mm,omitempty" cbor:"30,keyasint,omitempty"`
NetworkInterfaces map[string][4]uint64 `json:"ni,omitempty" cbor:"31,keyasint,omitempty"` // [upload bytes, download bytes, total upload, total download]
DiskIO [2]uint64 `json:"dio,omitzero" cbor:"32,keyasint,omitzero"` // [read bytes, write bytes]
MaxDiskIO [2]uint64 `json:"diom,omitzero" cbor:"-"` // [max read bytes, max write bytes]
@@ -90,8 +90,8 @@ type FsStats struct {
TotalWrite uint64 `json:"-"`
DiskReadPs float64 `json:"r" cbor:"2,keyasint"`
DiskWritePs float64 `json:"w" cbor:"3,keyasint"`
MaxDiskReadPS float64 `json:"rm,omitempty" cbor:"-"`
MaxDiskWritePS float64 `json:"wm,omitempty" cbor:"-"`
MaxDiskReadPS float64 `json:"rm,omitempty" cbor:"4,keyasint,omitempty"`
MaxDiskWritePS float64 `json:"wm,omitempty" cbor:"5,keyasint,omitempty"`
// TODO: remove DiskReadPs and DiskWritePs in future release in favor of DiskReadBytes and DiskWriteBytes
DiskReadBytes uint64 `json:"rb" cbor:"6,keyasint,omitempty"`
DiskWriteBytes uint64 `json:"wb" cbor:"7,keyasint,omitempty"`
@@ -129,23 +129,23 @@ type Info struct {
KernelVersion string `json:"k,omitempty" cbor:"1,keyasint,omitempty"` // deprecated - moved to Details struct
Cores int `json:"c,omitzero" cbor:"2,keyasint,omitzero"` // deprecated - moved to Details struct
// Threads is needed in Info struct to calculate load average thresholds
Threads int `json:"t,omitempty" cbor:"3,keyasint,omitempty"`
CpuModel string `json:"m,omitempty" cbor:"4,keyasint,omitempty"` // deprecated - moved to Details struct
Uptime uint64 `json:"u" cbor:"5,keyasint"`
Cpu float64 `json:"cpu" cbor:"6,keyasint"`
MemPct float64 `json:"mp" cbor:"7,keyasint"`
DiskPct float64 `json:"dp" cbor:"8,keyasint"`
Bandwidth float64 `json:"b,omitzero" cbor:"9,keyasint"` // deprecated in favor of BandwidthBytes
AgentVersion string `json:"v" cbor:"10,keyasint"`
Podman bool `json:"p,omitempty" cbor:"11,keyasint,omitempty"` // deprecated - moved to Details struct
GpuPct float64 `json:"g,omitempty" cbor:"12,keyasint,omitempty"`
DashboardTemp float64 `json:"dt,omitempty" cbor:"13,keyasint,omitempty"`
Os Os `json:"os,omitempty" cbor:"14,keyasint,omitempty"` // deprecated - moved to Details struct
// LoadAvg1 float64 `json:"l1,omitempty" cbor:"15,keyasint,omitempty"` // deprecated - use `la` array instead
// LoadAvg5 float64 `json:"l5,omitempty" cbor:"16,keyasint,omitempty"` // deprecated - use `la` array instead
// LoadAvg15 float64 `json:"l15,omitempty" cbor:"17,keyasint,omitempty"` // deprecated - use `la` array instead
Threads int `json:"t,omitempty" cbor:"3,keyasint,omitempty"`
CpuModel string `json:"m,omitempty" cbor:"4,keyasint,omitempty"` // deprecated - moved to Details struct
Uptime uint64 `json:"u" cbor:"5,keyasint"`
Cpu float64 `json:"cpu" cbor:"6,keyasint"`
MemPct float64 `json:"mp" cbor:"7,keyasint"`
DiskPct float64 `json:"dp" cbor:"8,keyasint"`
Bandwidth float64 `json:"b" cbor:"9,keyasint"`
AgentVersion string `json:"v" cbor:"10,keyasint"`
Podman bool `json:"p,omitempty" cbor:"11,keyasint,omitempty"` // deprecated - moved to Details struct
GpuPct float64 `json:"g,omitempty" cbor:"12,keyasint,omitempty"`
DashboardTemp float64 `json:"dt,omitempty" cbor:"13,keyasint,omitempty"`
Os Os `json:"os,omitempty" cbor:"14,keyasint,omitempty"` // deprecated - moved to Details struct
LoadAvg1 float64 `json:"l1,omitempty" cbor:"15,keyasint,omitempty"` // deprecated - use `la` array instead
LoadAvg5 float64 `json:"l5,omitempty" cbor:"16,keyasint,omitempty"` // deprecated - use `la` array instead
LoadAvg15 float64 `json:"l15,omitempty" cbor:"17,keyasint,omitempty"` // deprecated - use `la` array instead
BandwidthBytes uint64 `json:"bb" cbor:"18,keyasint"`
BandwidthBytes uint64 `json:"bb" cbor:"18,keyasint"`
LoadAvg [3]float64 `json:"la,omitempty" cbor:"19,keyasint"`
ConnectionType ConnectionType `json:"ct,omitempty" cbor:"20,keyasint,omitempty,omitzero"`
ExtraFsPct map[string]float64 `json:"efs,omitempty" cbor:"21,keyasint,omitempty"`

View File

@@ -917,7 +917,7 @@ func TestAgentWebSocketIntegration(t *testing.T) {
// Wait for connection result
maxWait := 2 * time.Second
time.Sleep(40 * time.Millisecond)
time.Sleep(20 * time.Millisecond)
checkInterval := 20 * time.Millisecond
timeout := time.After(maxWait)
ticker := time.Tick(checkInterval)

View File

@@ -1,39 +1,35 @@
// Package expirymap provides a thread-safe map with expiring entries.
// It supports TTL-based expiration with both lazy cleanup on access
// and periodic background cleanup.
package expirymap
import (
"sync"
"reflect"
"time"
"github.com/pocketbase/pocketbase/tools/store"
)
type val[T comparable] struct {
type val[T any] struct {
value T
expires time.Time
}
type ExpiryMap[T comparable] struct {
store store.Store[string, val[T]]
stopChan chan struct{}
stopOnce sync.Once
type ExpiryMap[T any] struct {
store *store.Store[string, *val[T]]
cleanupInterval time.Duration
}
// New creates a new expiry map with custom cleanup interval
func New[T comparable](cleanupInterval time.Duration) *ExpiryMap[T] {
func New[T any](cleanupInterval time.Duration) *ExpiryMap[T] {
m := &ExpiryMap[T]{
store: *store.New(map[string]val[T]{}),
stopChan: make(chan struct{}),
store: store.New(map[string]*val[T]{}),
cleanupInterval: cleanupInterval,
}
go m.startCleaner(cleanupInterval)
m.startCleaner()
return m
}
// Set stores a value with the given TTL
func (m *ExpiryMap[T]) Set(key string, value T, ttl time.Duration) {
m.store.Set(key, val[T]{
m.store.Set(key, &val[T]{
value: value,
expires: time.Now().Add(ttl),
})
@@ -59,7 +55,7 @@ func (m *ExpiryMap[T]) GetOk(key string) (T, bool) {
// GetByValue retrieves a value by value
func (m *ExpiryMap[T]) GetByValue(val T) (key string, value T, ok bool) {
for key, v := range m.store.GetAll() {
if v.value == val {
if reflect.DeepEqual(v.value, val) {
// check if expired
if v.expires.Before(time.Now()) {
m.store.Remove(key)
@@ -79,7 +75,7 @@ func (m *ExpiryMap[T]) Remove(key string) {
// RemovebyValue removes a value by value
func (m *ExpiryMap[T]) RemovebyValue(value T) (T, bool) {
for key, val := range m.store.GetAll() {
if val.value == value {
if reflect.DeepEqual(val.value, value) {
m.store.Remove(key)
return val.value, true
}
@@ -88,23 +84,13 @@ func (m *ExpiryMap[T]) RemovebyValue(value T) (T, bool) {
}
// startCleaner runs the background cleanup process
func (m *ExpiryMap[T]) startCleaner(interval time.Duration) {
tick := time.Tick(interval)
for {
select {
case <-tick:
func (m *ExpiryMap[T]) startCleaner() {
go func() {
tick := time.Tick(m.cleanupInterval)
for range tick {
m.cleanup()
case <-m.stopChan:
return
}
}
}
// StopCleaner stops the background cleanup process
func (m *ExpiryMap[T]) StopCleaner() {
m.stopOnce.Do(func() {
close(m.stopChan)
})
}()
}
// cleanup removes all expired entries
@@ -116,12 +102,3 @@ func (m *ExpiryMap[T]) cleanup() {
}
}
}
// UpdateExpiration updates the expiration time of a key
func (m *ExpiryMap[T]) UpdateExpiration(key string, ttl time.Duration) {
value, ok := m.store.GetOk(key)
if ok {
value.expires = time.Now().Add(ttl)
m.store.Set(key, value)
}
}

View File

@@ -4,7 +4,6 @@ package expirymap
import (
"testing"
"testing/synctest"
"time"
"github.com/stretchr/testify/assert"
@@ -178,33 +177,6 @@ func TestExpiryMap_GenericTypes(t *testing.T) {
})
}
func TestExpiryMap_UpdateExpiration(t *testing.T) {
em := New[string](time.Hour)
// Set a value with short TTL
em.Set("key1", "value1", time.Millisecond*50)
// Verify it exists
assert.True(t, em.Has("key1"))
// Update expiration to a longer TTL
em.UpdateExpiration("key1", time.Hour)
// Wait for the original TTL to pass
time.Sleep(time.Millisecond * 100)
// Should still exist because expiration was updated
assert.True(t, em.Has("key1"))
value, ok := em.GetOk("key1")
assert.True(t, ok)
assert.Equal(t, "value1", value)
// Try updating non-existent key (should not panic)
assert.NotPanics(t, func() {
em.UpdateExpiration("nonexistent", time.Hour)
})
}
func TestExpiryMap_ZeroValues(t *testing.T) {
em := New[string](time.Hour)
@@ -501,52 +473,3 @@ func TestExpiryMap_ValueOperations_Integration(t *testing.T) {
assert.Equal(t, "unique", value)
assert.Equal(t, "key2", key)
}
func TestExpiryMap_Cleaner(t *testing.T) {
synctest.Test(t, func(t *testing.T) {
em := New[string](time.Second)
defer em.StopCleaner()
em.Set("test", "value", 500*time.Millisecond)
// Wait 600ms, value is expired but cleaner hasn't run yet (interval is 1s)
time.Sleep(600 * time.Millisecond)
synctest.Wait()
// Map should still hold the value in its internal store before lazy access or cleaner
assert.Equal(t, 1, len(em.store.GetAll()), "store should still have 1 item before cleaner runs")
// Wait another 500ms so cleaner (1s interval) runs
time.Sleep(500 * time.Millisecond)
synctest.Wait() // Wait for background goroutine to process the tick
assert.Equal(t, 0, len(em.store.GetAll()), "store should be empty after cleaner runs")
})
}
func TestExpiryMap_StopCleaner(t *testing.T) {
em := New[string](time.Hour)
// Initially, stopChan is open, reading would block
select {
case <-em.stopChan:
t.Fatal("stopChan should be open initially")
default:
// success
}
em.StopCleaner()
// After StopCleaner, stopChan is closed, reading returns immediately
select {
case <-em.stopChan:
// success
default:
t.Fatal("stopChan was not closed by StopCleaner")
}
// Calling StopCleaner again should NOT panic thanks to sync.Once
assert.NotPanics(t, func() {
em.StopCleaner()
})
}

View File

@@ -48,6 +48,7 @@ type System struct {
detailsFetched atomic.Bool // True if static system details have been fetched and saved
smartFetching atomic.Bool // True if SMART devices are currently being fetched
smartInterval time.Duration // Interval for periodic SMART data updates
lastSmartFetch atomic.Int64 // Unix milliseconds of last SMART data fetch
}
func (sm *SystemManager) NewSystem(systemId string) *System {
@@ -133,34 +134,19 @@ func (sys *System) update() error {
return err
}
// ensure deprecated fields from older agents are migrated to current fields
migrateDeprecatedFields(data, !sys.detailsFetched.Load())
// create system records
_, err = sys.createRecords(data)
// if details were included and fetched successfully, mark details as fetched and update smart interval if set by agent
if err == nil && data.Details != nil {
sys.detailsFetched.Store(true)
// update smart interval if it's set on the agent side
if data.Details.SmartInterval > 0 {
sys.smartInterval = data.Details.SmartInterval
// make sure we reset expiration of lastFetch to remain as long as the new smart interval
// to prevent premature expiration leading to new fetch if interval is different.
sys.manager.smartFetchMap.UpdateExpiration(sys.Id, sys.smartInterval+time.Minute)
}
}
// Fetch and save SMART devices when system first comes online or at intervals
if backgroundSmartFetchEnabled() && sys.detailsFetched.Load() {
if backgroundSmartFetchEnabled() {
if sys.smartInterval <= 0 {
sys.smartInterval = time.Hour
}
lastFetch, _ := sys.manager.smartFetchMap.GetOk(sys.Id)
if time.Since(time.UnixMilli(lastFetch-1e4)) >= sys.smartInterval && sys.smartFetching.CompareAndSwap(false, true) {
lastFetch := sys.lastSmartFetch.Load()
if time.Since(time.UnixMilli(lastFetch)) >= sys.smartInterval && sys.smartFetching.CompareAndSwap(false, true) {
go func() {
defer sys.smartFetching.Store(false)
sys.manager.smartFetchMap.Set(sys.Id, time.Now().UnixMilli(), sys.smartInterval+time.Minute)
sys.lastSmartFetch.Store(time.Now().UnixMilli())
_ = sys.FetchAndSaveSmartDevices()
}()
}
@@ -235,6 +221,11 @@ func (sys *System) createRecords(data *system.CombinedData) (*core.Record, error
if err := createSystemDetailsRecord(txApp, data.Details, sys.Id); err != nil {
return err
}
sys.detailsFetched.Store(true)
// update smart interval if it's set on the agent side
if data.Details.SmartInterval > 0 {
sys.smartInterval = data.Details.SmartInterval
}
}
// update system record (do this last because it triggers alerts and we need above records to be inserted first)
@@ -318,11 +309,10 @@ func createContainerRecords(app core.App, data []*container.Stats, systemId stri
valueStrings := make([]string, 0, len(data))
for i, container := range data {
suffix := fmt.Sprintf("%d", i)
valueStrings = append(valueStrings, fmt.Sprintf("({:id%[1]s}, {:system}, {:name%[1]s}, {:image%[1]s}, {:ports%[1]s}, {:status%[1]s}, {:health%[1]s}, {:cpu%[1]s}, {:memory%[1]s}, {:net%[1]s}, {:updated})", suffix))
valueStrings = append(valueStrings, fmt.Sprintf("({:id%[1]s}, {:system}, {:name%[1]s}, {:image%[1]s}, {:status%[1]s}, {:health%[1]s}, {:cpu%[1]s}, {:memory%[1]s}, {:net%[1]s}, {:updated})", suffix))
params["id"+suffix] = container.Id
params["name"+suffix] = container.Name
params["image"+suffix] = container.Image
params["ports"+suffix] = container.Ports
params["status"+suffix] = container.Status
params["health"+suffix] = container.Health
params["cpu"+suffix] = container.Cpu
@@ -334,7 +324,7 @@ func createContainerRecords(app core.App, data []*container.Stats, systemId stri
params["net"+suffix] = netBytes
}
queryString := fmt.Sprintf(
"INSERT INTO containers (id, system, name, image, ports, status, health, cpu, memory, net, updated) VALUES %s ON CONFLICT(id) DO UPDATE SET system = excluded.system, name = excluded.name, image = excluded.image, ports = excluded.ports, status = excluded.status, health = excluded.health, cpu = excluded.cpu, memory = excluded.memory, net = excluded.net, updated = excluded.updated",
"INSERT INTO containers (id, system, name, image, status, health, cpu, memory, net, updated) VALUES %s ON CONFLICT(id) DO UPDATE SET system = excluded.system, name = excluded.name, image = excluded.image, status = excluded.status, health = excluded.health, cpu = excluded.cpu, memory = excluded.memory, net = excluded.net, updated = excluded.updated",
strings.Join(valueStrings, ","),
)
_, err := app.DB().NewQuery(queryString).Bind(params).Execute()
@@ -713,50 +703,3 @@ func getJitter() <-chan time.Time {
msDelay := (interval * minPercent / 100) + rand.Intn(interval*jitterRange/100)
return time.After(time.Duration(msDelay) * time.Millisecond)
}
// migrateDeprecatedFields moves values from deprecated fields to their new locations if the new
// fields are not already populated. Deprecated fields and refs may be removed at least 30 days
// and one minor version release after the release that includes the migration.
//
// This is run when processing incoming system data from agents, which may be on older versions.
func migrateDeprecatedFields(cd *system.CombinedData, createDetails bool) {
// migration added 0.19.0
if cd.Stats.Bandwidth[0] == 0 && cd.Stats.Bandwidth[1] == 0 {
cd.Stats.Bandwidth[0] = uint64(cd.Stats.NetworkSent * 1024 * 1024)
cd.Stats.Bandwidth[1] = uint64(cd.Stats.NetworkRecv * 1024 * 1024)
cd.Stats.NetworkSent, cd.Stats.NetworkRecv = 0, 0
}
// migration added 0.19.0
if cd.Info.BandwidthBytes == 0 {
cd.Info.BandwidthBytes = uint64(cd.Info.Bandwidth * 1024 * 1024)
cd.Info.Bandwidth = 0
}
// migration added 0.19.0
if cd.Stats.DiskIO[0] == 0 && cd.Stats.DiskIO[1] == 0 {
cd.Stats.DiskIO[0] = uint64(cd.Stats.DiskReadPs * 1024 * 1024)
cd.Stats.DiskIO[1] = uint64(cd.Stats.DiskWritePs * 1024 * 1024)
cd.Stats.DiskReadPs, cd.Stats.DiskWritePs = 0, 0
}
// migration added 0.19.0 - Move deprecated Info fields to Details struct
if cd.Details == nil && cd.Info.Hostname != "" {
if createDetails {
cd.Details = &system.Details{
Hostname: cd.Info.Hostname,
Kernel: cd.Info.KernelVersion,
Cores: cd.Info.Cores,
Threads: cd.Info.Threads,
CpuModel: cd.Info.CpuModel,
Podman: cd.Info.Podman,
Os: cd.Info.Os,
MemoryTotal: uint64(cd.Stats.Mem * 1024 * 1024 * 1024),
}
}
// zero the deprecated fields to prevent saving them in systems.info DB json payload
cd.Info.Hostname = ""
cd.Info.KernelVersion = ""
cd.Info.Cores = 0
cd.Info.CpuModel = ""
cd.Info.Podman = false
cd.Info.Os = 0
}
}

View File

@@ -8,7 +8,6 @@ import (
"github.com/henrygd/beszel/internal/hub/ws"
"github.com/henrygd/beszel/internal/entities/system"
"github.com/henrygd/beszel/internal/hub/expirymap"
"github.com/henrygd/beszel/internal/common"
@@ -41,10 +40,9 @@ var errSystemExists = errors.New("system exists")
// SystemManager manages a collection of monitored systems and their connections.
// It handles system lifecycle, status updates, and maintains both SSH and WebSocket connections.
type SystemManager struct {
hub hubLike // Hub interface for database and alert operations
systems *store.Store[string, *System] // Thread-safe store of active systems
sshConfig *ssh.ClientConfig // SSH client configuration for system connections
smartFetchMap *expirymap.ExpiryMap[int64] // Stores last SMART fetch time per system ID
hub hubLike // Hub interface for database and alert operations
systems *store.Store[string, *System] // Thread-safe store of active systems
sshConfig *ssh.ClientConfig // SSH client configuration for system connections
}
// hubLike defines the interface requirements for the hub dependency.
@@ -60,9 +58,8 @@ type hubLike interface {
// The hub must implement the hubLike interface to provide database and alert functionality.
func NewSystemManager(hub hubLike) *SystemManager {
return &SystemManager{
systems: store.New(map[string]*System{}),
hub: hub,
smartFetchMap: expirymap.New[int64](time.Hour),
systems: store.New(map[string]*System{}),
hub: hub,
}
}

View File

@@ -1,159 +0,0 @@
//go:build testing
package systems
import (
"testing"
"github.com/henrygd/beszel/internal/entities/system"
)
func TestCombinedData_MigrateDeprecatedFields(t *testing.T) {
t.Run("Migrate NetworkSent and NetworkRecv to Bandwidth", func(t *testing.T) {
cd := &system.CombinedData{
Stats: system.Stats{
NetworkSent: 1.5, // 1.5 MB
NetworkRecv: 2.5, // 2.5 MB
},
}
migrateDeprecatedFields(cd, true)
expectedSent := uint64(1.5 * 1024 * 1024)
expectedRecv := uint64(2.5 * 1024 * 1024)
if cd.Stats.Bandwidth[0] != expectedSent {
t.Errorf("expected Bandwidth[0] %d, got %d", expectedSent, cd.Stats.Bandwidth[0])
}
if cd.Stats.Bandwidth[1] != expectedRecv {
t.Errorf("expected Bandwidth[1] %d, got %d", expectedRecv, cd.Stats.Bandwidth[1])
}
if cd.Stats.NetworkSent != 0 || cd.Stats.NetworkRecv != 0 {
t.Errorf("expected NetworkSent and NetworkRecv to be reset, got %f, %f", cd.Stats.NetworkSent, cd.Stats.NetworkRecv)
}
})
t.Run("Migrate Info.Bandwidth to Info.BandwidthBytes", func(t *testing.T) {
cd := &system.CombinedData{
Info: system.Info{
Bandwidth: 10.0, // 10 MB
},
}
migrateDeprecatedFields(cd, true)
expected := uint64(10 * 1024 * 1024)
if cd.Info.BandwidthBytes != expected {
t.Errorf("expected BandwidthBytes %d, got %d", expected, cd.Info.BandwidthBytes)
}
if cd.Info.Bandwidth != 0 {
t.Errorf("expected Info.Bandwidth to be reset, got %f", cd.Info.Bandwidth)
}
})
t.Run("Migrate DiskReadPs and DiskWritePs to DiskIO", func(t *testing.T) {
cd := &system.CombinedData{
Stats: system.Stats{
DiskReadPs: 3.0, // 3 MB
DiskWritePs: 4.0, // 4 MB
},
}
migrateDeprecatedFields(cd, true)
expectedRead := uint64(3 * 1024 * 1024)
expectedWrite := uint64(4 * 1024 * 1024)
if cd.Stats.DiskIO[0] != expectedRead {
t.Errorf("expected DiskIO[0] %d, got %d", expectedRead, cd.Stats.DiskIO[0])
}
if cd.Stats.DiskIO[1] != expectedWrite {
t.Errorf("expected DiskIO[1] %d, got %d", expectedWrite, cd.Stats.DiskIO[1])
}
if cd.Stats.DiskReadPs != 0 || cd.Stats.DiskWritePs != 0 {
t.Errorf("expected DiskReadPs and DiskWritePs to be reset, got %f, %f", cd.Stats.DiskReadPs, cd.Stats.DiskWritePs)
}
})
t.Run("Migrate Info fields to Details struct", func(t *testing.T) {
cd := &system.CombinedData{
Stats: system.Stats{
Mem: 16.0, // 16 GB
},
Info: system.Info{
Hostname: "test-host",
KernelVersion: "6.8.0",
Cores: 8,
Threads: 16,
CpuModel: "Intel i7",
Podman: true,
Os: system.Linux,
},
}
migrateDeprecatedFields(cd, true)
if cd.Details == nil {
t.Fatal("expected Details struct to be created")
}
if cd.Details.Hostname != "test-host" {
t.Errorf("expected Hostname 'test-host', got '%s'", cd.Details.Hostname)
}
if cd.Details.Kernel != "6.8.0" {
t.Errorf("expected Kernel '6.8.0', got '%s'", cd.Details.Kernel)
}
if cd.Details.Cores != 8 {
t.Errorf("expected Cores 8, got %d", cd.Details.Cores)
}
if cd.Details.Threads != 16 {
t.Errorf("expected Threads 16, got %d", cd.Details.Threads)
}
if cd.Details.CpuModel != "Intel i7" {
t.Errorf("expected CpuModel 'Intel i7', got '%s'", cd.Details.CpuModel)
}
if cd.Details.Podman != true {
t.Errorf("expected Podman true, got %v", cd.Details.Podman)
}
if cd.Details.Os != system.Linux {
t.Errorf("expected Os Linux, got %d", cd.Details.Os)
}
expectedMem := uint64(16 * 1024 * 1024 * 1024)
if cd.Details.MemoryTotal != expectedMem {
t.Errorf("expected MemoryTotal %d, got %d", expectedMem, cd.Details.MemoryTotal)
}
if cd.Info.Hostname != "" || cd.Info.KernelVersion != "" || cd.Info.Cores != 0 || cd.Info.CpuModel != "" || cd.Info.Podman != false || cd.Info.Os != 0 {
t.Errorf("expected Info fields to be reset, got %+v", cd.Info)
}
})
t.Run("Do not migrate if Details already exists", func(t *testing.T) {
cd := &system.CombinedData{
Details: &system.Details{Hostname: "existing-host"},
Info: system.Info{
Hostname: "deprecated-host",
},
}
migrateDeprecatedFields(cd, true)
if cd.Details.Hostname != "existing-host" {
t.Errorf("expected Hostname 'existing-host', got '%s'", cd.Details.Hostname)
}
if cd.Info.Hostname != "deprecated-host" {
t.Errorf("expected Info.Hostname to remain 'deprecated-host', got '%s'", cd.Info.Hostname)
}
})
t.Run("Do not create details if migrateDetails is false", func(t *testing.T) {
cd := &system.CombinedData{
Info: system.Info{
Hostname: "deprecated-host",
},
}
migrateDeprecatedFields(cd, false)
if cd.Details != nil {
t.Fatal("expected Details struct to not be created")
}
if cd.Info.Hostname != "" {
t.Errorf("expected Info.Hostname to be reset, got '%s'", cd.Info.Hostname)
}
})
}

View File

@@ -113,5 +113,4 @@ func (sm *SystemManager) RemoveAllSystems() {
for _, system := range sm.systems.GetAll() {
sm.RemoveSystem(system.Id)
}
sm.smartFetchMap.StopCleaner()
}

View File

@@ -977,6 +977,18 @@ func init() {
"system": false,
"type": "number"
},
{
"hidden": false,
"id": "number3332085495",
"max": null,
"min": null,
"name": "updated",
"onlyInt": true,
"presentable": false,
"required": true,
"system": false,
"type": "number"
},
{
"autogeneratePattern": "",
"hidden": false,
@@ -990,32 +1002,6 @@ func init() {
"required": false,
"system": false,
"type": "text"
},
{
"autogeneratePattern": "",
"hidden": false,
"id": "text2308952269",
"max": 0,
"min": 0,
"name": "ports",
"pattern": "",
"presentable": false,
"primaryKey": false,
"required": false,
"system": false,
"type": "text"
},
{
"hidden": false,
"id": "number3332085495",
"max": null,
"min": null,
"name": "updated",
"onlyInt": true,
"presentable": false,
"required": true,
"system": false,
"type": "number"
}
],
"indexes": [

View File

@@ -1,5 +1,5 @@
{
"$schema": "https://biomejs.dev/schemas/2.2.4/schema.json",
"$schema": "https://biomejs.dev/schemas/2.2.3/schema.json",
"vcs": {
"enabled": true,
"clientKind": "git",
@@ -12,7 +12,7 @@
"lineWidth": 120,
"formatWithErrors": true
},
"assist": { "actions": { "source": { "organizeImports": "off" } } },
"assist": { "actions": { "source": { "organizeImports": "on" } } },
"linter": {
"enabled": true,
"rules": {

View File

@@ -1,7 +1,7 @@
{
"name": "beszel",
"private": true,
"version": "0.18.4",
"version": "0.18.3",
"type": "module",
"scripts": {
"dev": "vite --host",

View File

@@ -16,16 +16,19 @@ import { useYAxisWidth } from "./hooks"
export default memo(function LoadAverageChart({ chartData }: { chartData: ChartData }) {
const { yAxisWidth, updateYAxisWidth } = useYAxisWidth()
const keys: { color: string; label: string }[] = [
const keys: { legacy: keyof SystemStats; color: string; label: string }[] = [
{
legacy: "l1",
color: "hsl(271, 81%, 60%)", // Purple
label: t({ message: `1 min`, comment: "Load average" }),
},
{
legacy: "l5",
color: "hsl(217, 91%, 60%)", // Blue
label: t({ message: `5 min`, comment: "Load average" }),
},
{
legacy: "l15",
color: "hsl(25, 95%, 53%)", // Orange
label: t({ message: `15 min`, comment: "Load average" }),
},
@@ -63,18 +66,27 @@ export default memo(function LoadAverageChart({ chartData }: { chartData: ChartD
/>
}
/>
{keys.map(({ color, label }, i) => (
<Line
key={label}
dataKey={(value: { stats: SystemStats }) => value.stats?.la?.[i]}
name={label}
type="monotoneX"
dot={false}
strokeWidth={1.5}
stroke={color}
isAnimationActive={false}
/>
))}
{keys.map(({ legacy, color, label }, i) => {
const dataKey = (value: { stats: SystemStats }) => {
const { minor, patch } = chartData.agentVersion
if (minor <= 12 && patch < 1) {
return value.stats?.[legacy]
}
return value.stats?.la?.[i] ?? value.stats?.[legacy]
}
return (
<Line
key={label}
dataKey={dataKey}
name={label}
type="monotoneX"
dot={false}
strokeWidth={1.5}
stroke={color}
isAnimationActive={false}
/>
)
})}
<ChartLegend content={<ChartLegendContent />} />
</LineChart>
</ChartContainer>

View File

@@ -4,6 +4,7 @@ import { cn, decimalString, formatBytes, hourWithSeconds } from "@/lib/utils"
import type { ContainerRecord } from "@/types"
import { ContainerHealth, ContainerHealthLabels } from "@/lib/enums"
import {
ArrowUpDownIcon,
ClockIcon,
ContainerIcon,
CpuIcon,
@@ -12,12 +13,11 @@ import {
ServerIcon,
ShieldCheckIcon,
} from "lucide-react"
import { EthernetIcon, HourglassIcon, SquareArrowRightEnterIcon } from "../ui/icons"
import { EthernetIcon, HourglassIcon } from "../ui/icons"
import { Badge } from "../ui/badge"
import { t } from "@lingui/core/macro"
import { $allSystemsById } from "@/lib/stores"
import { useStore } from "@nanostores/react"
import { Tooltip, TooltipContent, TooltipTrigger } from "../ui/tooltip"
// Unit names and their corresponding number of seconds for converting docker status strings
const unitSeconds = [
@@ -134,29 +134,6 @@ export const containerChartCols: ColumnDef<ContainerRecord>[] = [
)
},
},
{
id: "ports",
accessorFn: (record) => record.ports || undefined,
header: ({ column }) => (
<HeaderButton column={column} name={t({ message: "Ports", context: "Container ports" })} Icon={SquareArrowRightEnterIcon} />
),
cell: ({ getValue }) => {
const val = getValue() as string
if (!val) {
return <span className="ms-2">-</span>
}
const className = "ms-1.5 w-20 block truncate tabular-nums"
if (val.length > 9) {
return (
<Tooltip>
<TooltipTrigger className={className}>{val}</TooltipTrigger>
<TooltipContent>{val}</TooltipContent>
</Tooltip>
)
}
return <span className={className}>{val}</span>
},
},
{
id: "image",
sortingFn: (a, b) => a.original.image.localeCompare(b.original.image),
@@ -211,7 +188,7 @@ function HeaderButton({
>
{Icon && <Icon className="size-4" />}
{name}
{/* <ArrowUpDownIcon className="size-4" /> */}
<ArrowUpDownIcon className="size-4" />
</Button>
)
}

View File

@@ -51,7 +51,7 @@ export default function ContainersTable({ systemId }: { systemId?: string }) {
function fetchData(systemId?: string) {
pb.collection<ContainerRecord>("containers")
.getList(0, 2000, {
fields: "id,name,image,ports,cpu,memory,net,health,status,system,updated",
fields: "id,name,image,cpu,memory,net,health,status,system,updated",
filter: systemId ? pb.filter("system={:system}", { system: systemId }) : undefined,
})
.then(({ items }) => {
@@ -134,8 +134,7 @@ export default function ContainersTable({ systemId }: { systemId?: string }) {
const status = container.status ?? ""
const healthLabel = ContainerHealthLabels[container.health as ContainerHealth] ?? ""
const image = container.image ?? ""
const ports = container.ports ?? ""
const searchString = `${systemName} ${id} ${name} ${healthLabel} ${status} ${image} ${ports}`.toLowerCase()
const searchString = `${systemName} ${id} ${name} ${healthLabel} ${status} ${image}`.toLowerCase()
return (filterValue as string)
.toLowerCase()
@@ -379,14 +378,8 @@ function ContainerSheet({
{container.image}
<Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
{container.id}
{/* {container.ports && (
<>
<Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
{container.ports}
</>
)} */}
{/* <Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
{ContainerHealthLabels[container.health as ContainerHealth]} */}
<Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
{ContainerHealthLabels[container.health as ContainerHealth]}
</SheetDescription>
</SheetHeader>
<div className="px-3 pb-3 -mt-4 flex flex-col gap-3 h-full items-start">

View File

@@ -43,7 +43,7 @@ export function copyDockerCompose(port = "45876", publicKey: string, token: stri
export function copyDockerRun(port = "45876", publicKey: string, token: string) {
copyToClipboard(
`docker run -d --name beszel-agent --network host --restart unless-stopped -v /var/run/docker.sock:/var/run/docker.sock:ro -v beszel_agent_data:/var/lib/beszel-agent -e KEY="${publicKey}" -e LISTEN=${port} -e TOKEN="${token}" -e HUB_URL="${getHubURL()}" henrygd/beszel-agent`
`docker run -d --name beszel-agent --network host --restart unless-stopped -v /var/run/docker.sock:/var/run/docker.sock:ro -v ./beszel_agent_data:/var/lib/beszel-agent -e KEY="${publicKey}" -e LISTEN=${port} -e TOKEN="${token}" -e HUB_URL="${getHubURL()}" henrygd/beszel-agent`
)
}

View File

@@ -32,10 +32,7 @@ export function LangToggle() {
className={cn("px-2.5 flex gap-2.5 cursor-pointer", lang === i18n.locale && "bg-accent/70 font-medium")}
onClick={() => dynamicActivate(lang)}
>
<span>
{e || <code className="font-mono bg-muted text-[.65em] w-5 h-4 grid place-items-center">{lang}</code>}
</span>{" "}
{label}
<span>{e}</span> {label}
</DropdownMenuItem>
))}
</DropdownMenuContent>

View File

@@ -70,16 +70,7 @@ export default function SettingsProfilePage({ userSettings }: { userSettings: Us
<SelectContent>
{languages.map(([lang, label, e]) => (
<SelectItem key={lang} value={lang}>
<span className="me-2.5">
{e || (
<code
aria-hidden="true"
className="font-mono bg-muted text-[.65em] w-5 h-4 inline-grid place-items-center"
>
{lang}
</code>
)}
</span>
<span className="me-2.5">{e}</span>
{label}
</SelectItem>
))}

View File

@@ -1,6 +1,7 @@
import { t } from "@lingui/core/macro"
import { Trans } from "@lingui/react/macro"
import { redirectPage } from "@nanostores/router"
import clsx from "clsx"
import { LoaderCircleIcon, SendIcon } from "lucide-react"
import { useEffect, useState } from "react"
import { $router } from "@/components/router"
@@ -9,7 +10,6 @@ import { Button } from "@/components/ui/button"
import { Separator } from "@/components/ui/separator"
import { toast } from "@/components/ui/use-toast"
import { isAdmin, pb } from "@/lib/api"
import { cn } from "@/lib/utils"
interface HeartbeatStatus {
enabled: boolean
@@ -37,10 +37,10 @@ export default function HeartbeatSettings() {
setIsLoading(true)
const res = await pb.send<HeartbeatStatus>("/api/beszel/heartbeat-status", {})
setStatus(res)
} catch (error: unknown) {
} catch (error: any) {
toast({
title: t`Error`,
description: (error as Error).message,
description: error.message,
variant: "destructive",
})
} finally {
@@ -66,10 +66,10 @@ export default function HeartbeatSettings() {
variant: "destructive",
})
}
} catch (error: unknown) {
} catch (error: any) {
toast({
title: t`Error`,
description: (error as Error).message,
description: error.message,
variant: "destructive",
})
} finally {
@@ -77,6 +77,8 @@ export default function HeartbeatSettings() {
}
}
const TestIcon = isTesting ? LoaderCircleIcon : SendIcon
return (
<div>
<div>
@@ -92,107 +94,91 @@ export default function HeartbeatSettings() {
</div>
<Separator className="my-4" />
{status?.enabled ? (
<EnabledState status={status} isTesting={isTesting} sendTestHeartbeat={sendTestHeartbeat} />
) : (
<NotEnabledState isLoading={isLoading} />
)}
</div>
)
}
function EnabledState({
status,
isTesting,
sendTestHeartbeat,
}: {
status: HeartbeatStatus
isTesting: boolean
sendTestHeartbeat: () => void
}) {
const TestIcon = isTesting ? LoaderCircleIcon : SendIcon
return (
<div className="space-y-5">
<div className="flex items-center gap-2">
<Badge variant="success">
<Trans>Active</Trans>
</Badge>
</div>
<div className="grid gap-4 sm:grid-cols-2">
<ConfigItem label={t`Endpoint URL`} value={status.url ?? ""} mono />
<ConfigItem label={t`Interval`} value={`${status.interval}s`} />
<ConfigItem label={t`HTTP Method`} value={status.method ?? "POST"} />
</div>
<Separator />
<div>
<h4 className="text-base font-medium mb-1">
<Trans>Test heartbeat</Trans>
</h4>
<p className="text-sm text-muted-foreground leading-relaxed mb-3">
<Trans>Send a single heartbeat ping to verify your endpoint is working.</Trans>
</p>
<Button
type="button"
variant="outline"
className="flex items-center gap-1.5"
onClick={sendTestHeartbeat}
disabled={isTesting}
>
<TestIcon className={cn("size-4", isTesting && "animate-spin")} />
<Trans>Send test heartbeat</Trans>
</Button>
</div>
<Separator />
<div>
<h4 className="text-base font-medium mb-2">
<Trans>Payload format</Trans>
</h4>
<p className="text-sm text-muted-foreground leading-relaxed mb-2">
<Trans>
When using POST, each heartbeat includes a JSON payload with system status summary, list of down systems,
and triggered alerts.
</Trans>
</p>
<p className="text-sm text-muted-foreground leading-relaxed">
<Trans>
The overall status is <code className="bg-muted rounded-sm px-1 text-primary">ok</code> when all systems are
up, <code className="bg-muted rounded-sm px-1 text-primary">warn</code> when alerts are triggered, and{" "}
<code className="bg-muted rounded-sm px-1 text-primary">error</code> when any system is down.
</Trans>
</p>
</div>
</div>
)
}
function NotEnabledState({ isLoading }: { isLoading?: boolean }) {
return (
<div className={cn("grid gap-4", isLoading && "animate-pulse")}>
<div>
<p className="text-sm text-muted-foreground leading-relaxed mb-3">
<Trans>Set the following environment variables on your Beszel hub to enable heartbeat monitoring:</Trans>
</p>
<div className="grid gap-2.5">
<EnvVarItem
name="HEARTBEAT_URL"
description={t`Endpoint URL to ping (required)`}
example="https://uptime.betterstack.com/api/v1/heartbeat/xxxx"
/>
<EnvVarItem name="HEARTBEAT_INTERVAL" description={t`Seconds between pings (default: 60)`} example="60" />
<EnvVarItem
name="HEARTBEAT_METHOD"
description={t`HTTP method: POST, GET, or HEAD (default: POST)`}
example="POST"
/>
{isLoading ? (
<div className="flex items-center gap-2 text-muted-foreground py-4">
<LoaderCircleIcon className="h-4 w-4 animate-spin" />
<Trans>Loading...</Trans>
</div>
</div>
<p className="text-sm text-muted-foreground leading-relaxed">
<Trans>After setting the environment variables, restart your Beszel hub for changes to take effect.</Trans>
</p>
) : status?.enabled ? (
<div className="space-y-5">
<div className="flex items-center gap-2">
<Badge variant="success">
<Trans>Active</Trans>
</Badge>
</div>
<div className="grid gap-4 sm:grid-cols-2">
<ConfigItem label={t`Endpoint URL`} value={status.url ?? ""} mono />
<ConfigItem label={t`Interval`} value={`${status.interval}s`} />
<ConfigItem label={t`HTTP Method`} value={status.method ?? "POST"} />
</div>
<Separator />
<div>
<h4 className="text-base font-medium mb-1">
<Trans>Test heartbeat</Trans>
</h4>
<p className="text-sm text-muted-foreground leading-relaxed mb-3">
<Trans>Send a single heartbeat ping to verify your endpoint is working.</Trans>
</p>
<Button
type="button"
variant="outline"
className="flex items-center gap-1.5"
onClick={sendTestHeartbeat}
disabled={isTesting}
>
<TestIcon className={clsx("h-4 w-4", isTesting && "animate-spin")} />
<Trans>Send test heartbeat</Trans>
</Button>
</div>
<Separator />
<div>
<h4 className="text-base font-medium mb-2">
<Trans>Payload format</Trans>
</h4>
<p className="text-sm text-muted-foreground leading-relaxed mb-2">
<Trans>
When using POST, each heartbeat includes a JSON payload with system status summary, list of down
systems, and triggered alerts.
</Trans>
</p>
<p className="text-sm text-muted-foreground leading-relaxed">
<Trans>
The overall status is <code className="bg-muted rounded-sm px-1 text-primary">ok</code> when all systems
are up, <code className="bg-muted rounded-sm px-1 text-primary">warn</code> when alerts are triggered,
and <code className="bg-muted rounded-sm px-1 text-primary">error</code> when any system is down.
</Trans>
</p>
</div>
</div>
) : (
<div className="grid gap-4">
<div>
<p className="text-sm text-muted-foreground leading-relaxed mb-3">
<Trans>Set the following environment variables on your Beszel hub to enable heartbeat monitoring:</Trans>
</p>
<div className="grid gap-2.5">
<EnvVarItem
name="HEARTBEAT_URL"
description={t`Endpoint URL to ping (required)`}
example="https://uptime.betterstack.com/api/v1/heartbeat/xxxx"
/>
<EnvVarItem name="HEARTBEAT_INTERVAL" description={t`Seconds between pings (default: 60)`} example="60" />
<EnvVarItem
name="HEARTBEAT_METHOD"
description={t`HTTP method: POST, GET, or HEAD (default: POST)`}
example="POST"
/>
</div>
</div>
<p className="text-sm text-muted-foreground leading-relaxed">
<Trans>After setting the environment variables, restart your Beszel hub for changes to take effect.</Trans>
</p>
</div>
)}
</div>
)
}
@@ -201,14 +187,14 @@ function ConfigItem({ label, value, mono }: { label: string; value: string; mono
return (
<div>
<p className="text-sm font-medium mb-0.5">{label}</p>
<p className={cn("text-sm text-muted-foreground break-all", mono && "font-mono")}>{value}</p>
<p className={clsx("text-sm text-muted-foreground break-all", mono && "font-mono")}>{value}</p>
</div>
)
}
function EnvVarItem({ name, description, example }: { name: string; description: string; example: string }) {
return (
<div className="bg-muted/50 rounded-md px-3 py-2.5 grid gap-1.5">
<div className="bg-muted/50 rounded-md px-3 py-2 grid gap-1.5">
<code className="text-sm font-mono text-primary font-medium leading-tight">{name}</code>
<p className="text-sm text-muted-foreground">{description}</p>
<p className="text-xs text-muted-foreground">

View File

@@ -654,7 +654,7 @@ export default memo(function SystemDetail({ id }: { id: string }) {
)}
{/* Load Average chart */}
{chartData.agentVersion?.minor > 12 && (
{chartData.agentVersion?.minor >= 12 && (
<ChartCard
empty={dataEmpty}
grid={grid}

View File

@@ -621,8 +621,8 @@ function DiskSheet({
const deviceName = disk?.name || unknown
const model = disk?.model || unknown
const capacity = disk?.capacity ? formatCapacity(disk.capacity) : unknown
const serialNumber = disk?.serial
const firmwareVersion = disk?.firmware
const serialNumber = disk?.serial || unknown
const firmwareVersion = disk?.firmware || unknown
const status = disk?.state || unknown
return (
@@ -636,32 +636,24 @@ function DiskSheet({
{model}
<Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
{capacity}
{serialNumber && (
<>
<Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
<Tooltip>
<TooltipTrigger asChild>
<span>{serialNumber}</span>
</TooltipTrigger>
<TooltipContent>
<Trans>Serial Number</Trans>
</TooltipContent>
</Tooltip>
</>
)}
{firmwareVersion && (
<>
<Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
<Tooltip>
<TooltipTrigger asChild>
<span>{firmwareVersion}</span>
</TooltipTrigger>
<TooltipContent>
<Trans>Firmware</Trans>
</TooltipContent>
</Tooltip>
</>
)}
<Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
<Tooltip>
<TooltipTrigger asChild>
<span>{serialNumber}</span>
</TooltipTrigger>
<TooltipContent>
<Trans>Serial Number</Trans>
</TooltipContent>
</Tooltip>
<Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
<Tooltip>
<TooltipTrigger asChild>
<span>{firmwareVersion}</span>
</TooltipTrigger>
<TooltipContent>
<Trans>Firmware</Trans>
</TooltipContent>
</Tooltip>
</SheetDescription>
</SheetHeader>
<div className="flex-1 overflow-hidden p-4 flex flex-col gap-4">

View File

@@ -198,19 +198,32 @@ export function SystemsTableColumns(viewMode: "table" | "grid"): ColumnDef<Syste
},
{
id: "loadAverage",
accessorFn: ({ info }) => info.la?.reduce((acc, curr) => acc + curr, 0),
accessorFn: ({ info }) => {
const sum = info.la?.reduce((acc, curr) => acc + curr, 0)
// TODO: remove this in future release in favor of la array
if (!sum) {
return (info.l1 ?? 0) + (info.l5 ?? 0) + (info.l15 ?? 0) || undefined
}
return sum || undefined
},
name: () => t({ message: "Load Avg", comment: "Short label for load average" }),
size: 0,
Icon: HourglassIcon,
header: sortableHeader,
cell(info: CellContext<SystemRecord, unknown>) {
const { info: sysInfo, status } = info.row.original
const { major, minor } = parseSemVer(sysInfo.v)
const { colorWarn = 65, colorCrit = 90 } = useStore($userSettings, { keys: ["colorWarn", "colorCrit"] })
const loadAverages = sysInfo.la || []
// agent version
const { minor, patch } = parseSemVer(sysInfo.v)
let loadAverages = sysInfo.la
// use legacy load averages if agent version is less than 12.1.0
if (!loadAverages || (minor === 12 && patch < 1)) {
loadAverages = [sysInfo.l1 ?? 0, sysInfo.l5 ?? 0, sysInfo.l15 ?? 0]
}
const max = Math.max(...loadAverages)
if (max === 0 && (status === SystemStatus.Paused || (major < 1 && minor < 13))) {
if (max === 0 && (status === SystemStatus.Paused || minor < 12)) {
return null
}
@@ -235,20 +248,19 @@ export function SystemsTableColumns(viewMode: "table" | "grid"): ColumnDef<Syste
},
},
{
accessorFn: ({ info, status }) => (status !== SystemStatus.Up ? undefined : info.bb),
accessorFn: ({ info }) => info.bb || (info.b || 0) * 1024 * 1024 || undefined,
id: "net",
name: () => t`Net`,
size: 0,
Icon: EthernetIcon,
header: sortableHeader,
sortUndefined: "last",
cell(info) {
const val = info.getValue() as number | undefined
if (val === undefined) {
const sys = info.row.original
const userSettings = useStore($userSettings, { keys: ["unitNet"] })
if (sys.status === SystemStatus.Paused) {
return null
}
const userSettings = useStore($userSettings, { keys: ["unitNet"] })
const { value, unit } = formatBytes(val, true, userSettings.unitNet, false)
const { value, unit } = formatBytes((info.getValue() || 0) as number, true, userSettings.unitNet, false)
return (
<span className="tabular-nums whitespace-nowrap">
{decimalString(value, value >= 100 ? 1 : 2)} {unit}

View File

@@ -185,14 +185,3 @@ export function PlugChargingIcon(props: SVGProps<SVGSVGElement>) {
</svg>
)
}
// Lucide Icons (ISC) - used for ports
export function SquareArrowRightEnterIcon(props: SVGProps<SVGSVGElement>) {
return (
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" strokeWidth="2" strokeLinecap="round" {...props}>
<path d="m10 16 4-4-4-4" />
<path d="M3 12h11" />
<path d="M3 8V5a2 2 0 0 1 2-2h14a2 2 0 0 1 2 2v14a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2v-3" />
</svg>
)
}

View File

@@ -8,7 +8,7 @@ export default [
["es", "Español", "🇪🇸"],
["fa", "فارسی", "🇮🇷"],
["fr", "Français", "🇫🇷"],
["he", "עברית", ""],
["he", "עברית", "🕎"],
["hr", "Hrvatski", "🇭🇷"],
["hu", "Magyar", "🇭🇺"],
["id", "Indonesia", "🇮🇩"],

View File

@@ -937,6 +937,7 @@ msgstr "متوسط التحميل"
msgid "Load state"
msgstr "حالة التحميل"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "جاري التحميل..."

View File

@@ -8,7 +8,7 @@ msgstr ""
"Language: bg\n"
"Project-Id-Version: beszel\n"
"Report-Msgid-Bugs-To: \n"
"PO-Revision-Date: 2026-01-31 21:16\n"
"PO-Revision-Date: 2026-02-25 11:49\n"
"Last-Translator: \n"
"Language-Team: Bulgarian\n"
"Plural-Forms: nplurals=2; plural=(n != 1);\n"
@@ -357,7 +357,7 @@ msgstr "Провери log-овете за повече информация."
#: src/components/routes/settings/heartbeat.tsx
msgid "Check your monitoring service"
msgstr "Проверете вашата услуга за мониторинг"
msgstr "Проверете мониторинг услугата си"
#: src/components/routes/settings/notifications.tsx
msgid "Check your notification service"
@@ -1377,7 +1377,7 @@ msgstr "Избери {foo}"
#: src/components/routes/settings/heartbeat.tsx
msgid "Send a single heartbeat ping to verify your endpoint is working."
msgstr "Изпратете единичен heartbeat пинг, за да проверите дали вашата крайна точка работи."
msgstr "Изпратете единичен heartbeat пинг, за да се уверите, че крайната Ви точка работи."
#: src/components/routes/settings/heartbeat.tsx
msgid "Send periodic outbound pings to an external monitoring service so you can monitor Beszel without exposing it to the internet."
@@ -1409,7 +1409,7 @@ msgstr "Задайте процентни прагове за цветовете
#: src/components/routes/settings/heartbeat.tsx
msgid "Set the following environment variables on your Beszel hub to enable heartbeat monitoring:"
msgstr "Задайте следните променливи на средата на вашия Beszel hub, за да активирате мониторинга на heartbeat:"
msgstr "Задайте следните променливи на средата на вашия Beszel hub, за да активирате heartbeat мониторинг:"
#: src/components/command-palette.tsx
#: src/components/command-palette.tsx
@@ -1728,7 +1728,7 @@ msgstr "Качване"
#: src/components/routes/system/info-bar.tsx
#: src/components/systems-table/systems-table-columns.tsx
msgid "Uptime"
msgstr "Uptime"
msgstr "Време на работа"
#: src/components/routes/system.tsx
#: src/components/routes/system.tsx
@@ -1799,7 +1799,7 @@ msgstr "Webhook / Пуш нотификации"
#: src/components/routes/settings/tokens-fingerprints.tsx
msgid "When enabled, this token allows agents to self-register without prior system creation."
msgstr "Когато е активиран, този символ позволява на агентите да се регистрират сами без предварително създаване на система."
msgstr "Когато е активиран, този токен позволява на агентите да се регистрират сами без предварително създаване на система."
#: src/components/routes/settings/heartbeat.tsx
msgid "When using POST, each heartbeat includes a JSON payload with system status summary, list of down systems, and triggered alerts."
@@ -1834,3 +1834,4 @@ msgstr "Да"
#: src/components/routes/settings/layout.tsx
msgid "Your user settings have been updated."
msgstr "Настройките за потребителя ти са обновени."

View File

@@ -937,6 +937,7 @@ msgstr "Prům. zatížení"
msgid "Load state"
msgstr "Stav načtení"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Načítání..."

View File

@@ -937,6 +937,7 @@ msgstr "Belastning gns."
msgid "Load state"
msgstr "Indlæsningstilstand"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Indlæser..."

View File

@@ -8,7 +8,7 @@ msgstr ""
"Language: de\n"
"Project-Id-Version: beszel\n"
"Report-Msgid-Bugs-To: \n"
"PO-Revision-Date: 2026-01-31 21:16\n"
"PO-Revision-Date: 2026-03-14 20:37\n"
"Last-Translator: \n"
"Language-Team: German\n"
"Plural-Forms: nplurals=2; plural=(n != 1);\n"
@@ -236,7 +236,7 @@ msgstr "Bandbreite"
#. Battery label in systems table header
#: src/components/systems-table/systems-table-columns.tsx
msgid "Bat"
msgstr "Bat"
msgstr "Batt"
#: src/components/routes/system.tsx
#: src/lib/alerts.ts
@@ -1177,7 +1177,7 @@ msgstr "Prozentsatz der Zeit in jedem Zustand"
#: src/components/routes/settings/tokens-fingerprints.tsx
msgid "Permanent"
msgstr "Permanent"
msgstr "Dauerhaft"
#: src/components/routes/settings/tokens-fingerprints.tsx
msgid "Persistence"
@@ -1834,3 +1834,4 @@ msgstr "Ja"
#: src/components/routes/settings/layout.tsx
msgid "Your user settings have been updated."
msgstr "Deine Benutzereinstellungen wurden aktualisiert."

View File

@@ -932,6 +932,7 @@ msgstr "Load Avg"
msgid "Load state"
msgstr "Load state"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Loading..."

View File

@@ -937,6 +937,7 @@ msgstr "Carga media"
msgid "Load state"
msgstr "Estado de carga"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Cargando..."

View File

@@ -937,6 +937,7 @@ msgstr "میانگین بار"
msgid "Load state"
msgstr "وضعیت بارگذاری"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "در حال بارگذاری..."

View File

@@ -937,6 +937,7 @@ msgstr "Charge moy."
msgid "Load state"
msgstr "État de charge"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Chargement..."

View File

@@ -937,6 +937,7 @@ msgstr "ממוצע עומס"
msgid "Load state"
msgstr "מצב עומס"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "טוען..."

View File

@@ -937,6 +937,7 @@ msgstr "Prosječno Opterećenje"
msgid "Load state"
msgstr "Stanje učitavanja"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Učitavanje..."

View File

@@ -937,6 +937,7 @@ msgstr "Terhelési átlag"
msgid "Load state"
msgstr "Betöltési állapot"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Betöltés..."

View File

@@ -937,6 +937,7 @@ msgstr "Rata-rata Beban"
msgid "Load state"
msgstr "Beban saat ini"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Memuat..."

View File

@@ -937,6 +937,7 @@ msgstr "Carico Medio"
msgid "Load state"
msgstr "Stato di caricamento"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Caricamento..."

View File

@@ -937,6 +937,7 @@ msgstr "負荷平均"
msgid "Load state"
msgstr "ロード状態"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "読み込み中..."

View File

@@ -937,6 +937,7 @@ msgstr "부하 평균"
msgid "Load state"
msgstr "로드 상태"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "로딩 중..."

View File

@@ -8,7 +8,7 @@ msgstr ""
"Language: nl\n"
"Project-Id-Version: beszel\n"
"Report-Msgid-Bugs-To: \n"
"PO-Revision-Date: 2026-02-19 19:40\n"
"PO-Revision-Date: 2026-02-24 11:37\n"
"Last-Translator: \n"
"Language-Team: Dutch\n"
"Plural-Forms: nplurals=2; plural=(n != 1);\n"
@@ -836,7 +836,7 @@ msgstr "Gezondheid"
#: src/components/routes/settings/layout.tsx
msgid "Heartbeat"
msgstr "Heartbeat"
msgstr ""
#: src/components/routes/settings/heartbeat.tsx
msgid "Heartbeat Monitoring"
@@ -884,7 +884,7 @@ msgstr "Inactief"
#: src/components/routes/settings/heartbeat.tsx
msgid "Interval"
msgstr "Interval"
msgstr ""
#: src/components/login/auth-form.tsx
msgid "Invalid email address."
@@ -1834,3 +1834,4 @@ msgstr "Ja"
#: src/components/routes/settings/layout.tsx
msgid "Your user settings have been updated."
msgstr "Je gebruikersinstellingen zijn bijgewerkt."

View File

@@ -937,6 +937,7 @@ msgstr "Snittbelastning"
msgid "Load state"
msgstr "Lastetilstand"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Laster..."

View File

@@ -937,6 +937,7 @@ msgstr "Śr. obciążenie"
msgid "Load state"
msgstr "Stan obciążenia"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Ładowanie..."

View File

@@ -937,6 +937,7 @@ msgstr "Carga Média"
msgid "Load state"
msgstr "Estado de carga"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Carregando..."

View File

@@ -8,7 +8,7 @@ msgstr ""
"Language: ru\n"
"Project-Id-Version: beszel\n"
"Report-Msgid-Bugs-To: \n"
"PO-Revision-Date: 2026-01-31 21:16\n"
"PO-Revision-Date: 2026-02-21 09:46\n"
"Last-Translator: \n"
"Language-Team: Russian\n"
"Plural-Forms: nplurals=4; plural=((n%10==1 && n%100!=11) ? 0 : ((n%10 >= 2 && n%10 <=4 && (n%100 < 12 || n%100 > 14)) ? 1 : ((n%10 == 0 || (n%10 >= 5 && n%10 <=9)) || (n%100 >= 11 && n%100 <= 14)) ? 2 : 3));\n"
@@ -1728,7 +1728,7 @@ msgstr "Отдача"
#: src/components/routes/system/info-bar.tsx
#: src/components/systems-table/systems-table-columns.tsx
msgid "Uptime"
msgstr "Uptime"
msgstr "Время работы"
#: src/components/routes/system.tsx
#: src/components/routes/system.tsx
@@ -1834,3 +1834,4 @@ msgstr "Да"
#: src/components/routes/settings/layout.tsx
msgid "Your user settings have been updated."
msgstr "Ваши настройки пользователя были обновлены."

View File

@@ -937,6 +937,7 @@ msgstr "Povpr. obrem."
msgid "Load state"
msgstr "Stanje nalaganja"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Nalaganje..."

View File

@@ -8,7 +8,7 @@ msgstr ""
"Language: sr\n"
"Project-Id-Version: beszel\n"
"Report-Msgid-Bugs-To: \n"
"PO-Revision-Date: 2026-02-03 15:27\n"
"PO-Revision-Date: 2026-02-23 15:11\n"
"Last-Translator: \n"
"Language-Team: Serbian (Cyrillic)\n"
"Plural-Forms: nplurals=3; plural=(n%10==1 && n%100!=11 ? 0 : n%10>=2 && n%10<=4 && (n%100<10 || n%100>=20) ? 1 : 2);\n"
@@ -836,7 +836,7 @@ msgstr "Здравље"
#: src/components/routes/settings/layout.tsx
msgid "Heartbeat"
msgstr "Heartbeat"
msgstr "Оркуцај"
#: src/components/routes/settings/heartbeat.tsx
msgid "Heartbeat Monitoring"
@@ -884,7 +884,7 @@ msgstr "Неактивно"
#: src/components/routes/settings/heartbeat.tsx
msgid "Interval"
msgstr "Interval"
msgstr "Интервал"
#: src/components/login/auth-form.tsx
msgid "Invalid email address."
@@ -1728,7 +1728,7 @@ msgstr "Отпреми"
#: src/components/routes/system/info-bar.tsx
#: src/components/systems-table/systems-table-columns.tsx
msgid "Uptime"
msgstr "Uptime"
msgstr "Време рада"
#: src/components/routes/system.tsx
#: src/components/routes/system.tsx
@@ -1834,3 +1834,4 @@ msgstr "Да"
#: src/components/routes/settings/layout.tsx
msgid "Your user settings have been updated."
msgstr "Ваша корисничка подешавања су ажурирана."

View File

@@ -937,6 +937,7 @@ msgstr "Belastning"
msgid "Load state"
msgstr "Laddningstillstånd"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Laddar..."

View File

@@ -8,7 +8,7 @@ msgstr ""
"Language: tr\n"
"Project-Id-Version: beszel\n"
"Report-Msgid-Bugs-To: \n"
"PO-Revision-Date: 2026-01-31 21:16\n"
"PO-Revision-Date: 2026-03-02 13:16\n"
"Last-Translator: \n"
"Language-Team: Turkish\n"
"Plural-Forms: nplurals=2; plural=(n != 1);\n"
@@ -143,7 +143,7 @@ msgstr "Sonra"
#: src/components/routes/settings/heartbeat.tsx
msgid "After setting the environment variables, restart your Beszel hub for changes to take effect."
msgstr "Ortam değişkenlerini ayarladıktan sonra, değişikliklerin etkili olması için Beszel hub'ınızı yeniden başlatın."
msgstr "Ortam değişkenlerini ayarladıktan sonra, değişikliklerin etkili olması için Beszel Hub'ınızı yeniden başlatın."
#: src/components/systems-table/systems-table-columns.tsx
msgid "Agent"
@@ -466,7 +466,7 @@ msgstr "YAML'ı kopyala"
#: src/components/systemd-table/systemd-table-columns.tsx
#: src/components/systems-table/systems-table-columns.tsx
msgid "CPU"
msgstr "CPU"
msgstr "İşlemci"
#: src/components/routes/system/cpu-sheet.tsx
msgid "CPU Cores"
@@ -565,7 +565,7 @@ msgstr "Boşalıyor"
#: src/components/systems-table/systems-table-columns.tsx
msgid "Disk"
msgstr "Disk"
msgstr "Depolama alanı"
#: src/components/routes/system.tsx
msgid "Disk I/O"
@@ -836,7 +836,7 @@ msgstr "Sağlık"
#: src/components/routes/settings/layout.tsx
msgid "Heartbeat"
msgstr "Heartbeat"
msgstr ""
#: src/components/routes/settings/heartbeat.tsx
msgid "Heartbeat Monitoring"
@@ -1010,7 +1010,7 @@ msgstr "Docker konteynerlerinin bellek kullanımı"
#: src/components/routes/system/smart-table.tsx
msgid "Model"
msgstr "Model"
msgstr ""
#: src/components/add-system.tsx
#: src/components/alerts-history-columns.tsx
@@ -1530,7 +1530,7 @@ msgstr "<0>URL</0>'yi Test Et"
#: src/components/routes/settings/heartbeat.tsx
msgid "Test heartbeat"
msgstr "Test heartbeat"
msgstr ""
#: src/components/routes/settings/notifications.tsx
msgid "Test notification sent"
@@ -1581,7 +1581,7 @@ msgstr "Temayı değiştir"
#: src/components/add-system.tsx
#: src/components/routes/settings/tokens-fingerprints.tsx
msgid "Token"
msgstr "Token"
msgstr "Erişim Anahtarı"
#: src/components/command-palette.tsx
#: src/components/routes/settings/layout.tsx
@@ -1728,7 +1728,7 @@ msgstr "Yükle"
#: src/components/routes/system/info-bar.tsx
#: src/components/systems-table/systems-table-columns.tsx
msgid "Uptime"
msgstr "Uptime"
msgstr "Çalışma Süresi"
#: src/components/routes/system.tsx
#: src/components/routes/system.tsx
@@ -1834,3 +1834,4 @@ msgstr "Evet"
#: src/components/routes/settings/layout.tsx
msgid "Your user settings have been updated."
msgstr "Kullanıcı ayarlarınız güncellendi."

View File

@@ -937,6 +937,7 @@ msgstr "Сер. навантаження"
msgid "Load state"
msgstr "Стан завантаження"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Завантаження..."

View File

@@ -937,6 +937,7 @@ msgstr "Tải TB"
msgid "Load state"
msgstr "Trạng thái tải"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "Đang tải..."

View File

@@ -8,7 +8,7 @@ msgstr ""
"Language: zh\n"
"Project-Id-Version: beszel\n"
"Report-Msgid-Bugs-To: \n"
"PO-Revision-Date: 2026-01-31 21:16\n"
"PO-Revision-Date: 2026-03-01 17:14\n"
"Last-Translator: \n"
"Language-Team: Chinese Simplified\n"
"Plural-Forms: nplurals=1; plural=0;\n"
@@ -836,7 +836,7 @@ msgstr "健康"
#: src/components/routes/settings/layout.tsx
msgid "Heartbeat"
msgstr "Heartbeat"
msgstr "心跳"
#: src/components/routes/settings/heartbeat.tsx
msgid "Heartbeat Monitoring"
@@ -1728,7 +1728,7 @@ msgstr "上传"
#: src/components/routes/system/info-bar.tsx
#: src/components/systems-table/systems-table-columns.tsx
msgid "Uptime"
msgstr "运行时间"
msgstr "正常运行时间"
#: src/components/routes/system.tsx
#: src/components/routes/system.tsx
@@ -1834,3 +1834,4 @@ msgstr "是"
#: src/components/routes/settings/layout.tsx
msgid "Your user settings have been updated."
msgstr "您的用户设置已更新。"

View File

@@ -8,7 +8,7 @@ msgstr ""
"Language: zh\n"
"Project-Id-Version: beszel\n"
"Report-Msgid-Bugs-To: \n"
"PO-Revision-Date: 2026-01-31 21:16\n"
"PO-Revision-Date: 2026-02-20 02:25\n"
"Last-Translator: \n"
"Language-Team: Chinese Traditional, Hong Kong\n"
"Plural-Forms: nplurals=1; plural=0;\n"
@@ -836,7 +836,7 @@ msgstr "健康狀態"
#: src/components/routes/settings/layout.tsx
msgid "Heartbeat"
msgstr "Heartbeat"
msgstr ""
#: src/components/routes/settings/heartbeat.tsx
msgid "Heartbeat Monitoring"
@@ -937,6 +937,7 @@ msgstr "平均負載"
msgid "Load state"
msgstr "載入狀態"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "載入中..."
@@ -1728,7 +1729,7 @@ msgstr "上傳"
#: src/components/routes/system/info-bar.tsx
#: src/components/systems-table/systems-table-columns.tsx
msgid "Uptime"
msgstr "運行時間"
msgstr "正常運行時間"
#: src/components/routes/system.tsx
#: src/components/routes/system.tsx
@@ -1834,3 +1835,4 @@ msgstr "是"
#: src/components/routes/settings/layout.tsx
msgid "Your user settings have been updated."
msgstr "您的用戶設置已更新。"

View File

@@ -937,6 +937,7 @@ msgstr "平均負載"
msgid "Load state"
msgstr "載入狀態"
#: src/components/routes/settings/heartbeat.tsx
#: src/components/systemd-table/systemd-table.tsx
msgid "Loading..."
msgstr "載入中..."

View File

@@ -45,6 +45,12 @@ export interface SystemInfo {
c: number
/** cpu model */
m: string
/** load average 1 minute */
l1?: number
/** load average 5 minutes */
l5?: number
/** load average 15 minutes */
l15?: number
/** load average */
la?: [number, number, number]
/** operating system */
@@ -88,6 +94,13 @@ export interface SystemStats {
cpub?: number[]
/** per-core cpu usage [CPU0..] (0-100 integers) */
cpus?: number[]
// TODO: remove these in future release in favor of la
/** load average 1 minute */
l1?: number
/** load average 5 minutes */
l5?: number
/** load average 15 minutes */
l15?: number
/** load average */
la?: [number, number, number]
/** total memory (gb) */
@@ -254,7 +267,6 @@ export interface ContainerRecord extends RecordModel {
system: string
name: string
image: string
ports: string
cpu: number
memory: number
net: number

View File

@@ -98,7 +98,7 @@ func ClearCollection(t testing.TB, app core.App, collectionName string) error {
}
func (h *TestHub) Cleanup() {
h.GetAlertManager().Stop()
h.GetAlertManager().StopWorker()
h.GetSystemManager().RemoveAllSystems()
h.TestApp.Cleanup()
}

Some files were not shown because too many files have changed in this diff Show More