updates

# Conflicts:
#	agent/connection_manager.go

agent/agent.go

@@ -48,6 +48,7 @@ type Agent struct {
 	keys                      []gossh.PublicKey                                     // SSH public keys
 	smartManager              *SmartManager                                         // Manages SMART data
 	systemdManager            *systemdManager                                       // Manages systemd services
+	probeManager              *ProbeManager                                         // Manages network probes
 }
 
 // NewAgent creates a new agent with the given data directory for persisting data.
@@ -121,6 +122,9 @@ func NewAgent(dataDir ...string) (agent *Agent, err error) {
 	// initialize handler registry
 	agent.handlerRegistry = NewHandlerRegistry()
 
+	// initialize probe manager
+	agent.probeManager = newProbeManager()
+
 	// initialize disk info
 	agent.initializeDiskInfo()
 
@@ -178,6 +182,11 @@ func (a *Agent) gatherStats(options common.DataRequestOptions) *system.CombinedD
 		}
 	}
 
+	if a.probeManager != nil {
+		data.Probes = a.probeManager.GetResults(cacheTimeMs)
+		slog.Debug("Probes", "data", data.Probes)
+	}
+
 	// skip updating systemd services if cache time is not the default 60sec interval
 	if a.systemdManager != nil && cacheTimeMs == defaultDataCacheTimeMs {
 		totalCount := uint16(a.systemdManager.getServiceStatsCount())

agent/connection_manager.go

@@ -4,11 +4,15 @@ import (
 	"context"
 	"errors"
 	"log/slog"
+	"net"
+	"os"
 	"os/signal"
+	"strings"
 	"syscall"
 	"time"
 
 	"github.com/henrygd/beszel/agent/health"
+	"github.com/henrygd/beszel/agent/utils"
 	"github.com/henrygd/beszel/internal/entities/system"
 )
 
@@ -111,13 +115,37 @@ func (c *ConnectionManager) Start(serverOptions ServerOptions) error {
 			_ = health.Update()
 		case <-sigCtx.Done():
 			slog.Info("Shutting down", "cause", context.Cause(sigCtx))
-			_ = c.agent.StopServer()
-			c.closeWebSocket()
-			return health.CleanUp()
+			return c.stop()
 		}
 	}
 }
 
+// stop does not stop the connection manager itself, just any active connections. The manager will attempt to reconnect after stopping, so this should only be called immediately before shutting down the entire agent.
+//
+// If we need or want to expose a graceful Stop method in the future, do something like this to actually stop the manager:
+//
+//	func (c *ConnectionManager) Start(serverOptions ServerOptions) error {
+//		ctx, cancel := context.WithCancel(context.Background())
+//		c.cancel = cancel
+//
+//		for {
+//			select {
+//			case <-ctx.Done():
+//				return c.stop()
+//			}
+//		}
+//	}
+//
+//	func (c *ConnectionManager) Stop() {
+//		c.cancel()
+//	}
+func (c *ConnectionManager) stop() error {
+	_ = c.agent.StopServer()
+	c.agent.probeManager.Stop()
+	c.closeWebSocket()
+	return health.CleanUp()
+}
+
 // handleEvent processes connection events and updates the connection state accordingly.
 func (c *ConnectionManager) handleEvent(event ConnectionEvent) {
 	switch event {
@@ -185,9 +213,16 @@ func (c *ConnectionManager) connect() {
 
 	// Try WebSocket first, if it fails, start SSH server
 	err := c.startWebSocketConnection()
-	if err != nil && c.State == Disconnected {
-		c.startSSHServer()
-		c.startWsTicker()
+	if err != nil {
+		if shouldExitOnErr(err) {
+			time.Sleep(2 * time.Second) // prevent tight restart loop
+			_ = c.stop()
+			os.Exit(1)
+		}
+		if c.State == Disconnected {
+			c.startSSHServer()
+			c.startWsTicker()
+		}
 	}
 }
 
@@ -224,3 +259,14 @@ func (c *ConnectionManager) closeWebSocket() {
 		c.wsClient.Close()
 	}
 }
+
+// shouldExitOnErr checks if the error is a DNS resolution failure and if the
+// EXIT_ON_DNS_ERROR env var is set. https://github.com/henrygd/beszel/issues/1924.
+func shouldExitOnErr(err error) bool {
+	if val, _ := utils.GetEnv("EXIT_ON_DNS_ERROR"); val == "true" {
+		if opErr, ok := errors.AsType[*net.OpError](err); ok {
+			return strings.Contains(opErr.Err.Error(), "lookup")
+		}
+	}
+	return false
+}

agent/connection_manager_test.go

@@ -4,6 +4,7 @@ package agent
 
 import (
 	"crypto/ed25519"
+	"errors"
 	"fmt"
 	"net"
 	"net/url"
@@ -298,3 +299,65 @@ func TestConnectionManager_ConnectFlow(t *testing.T) {
 		cm.connect()
 	}, "Connect should not panic without WebSocket client")
 }
+
+func TestShouldExitOnErr(t *testing.T) {
+	createDialErr := func(msg string) error {
+		return &net.OpError{
+			Op:  "dial",
+			Net: "tcp",
+			Err: errors.New(msg),
+		}
+	}
+
+	tests := []struct {
+		name     string
+		err      error
+		envValue string
+		expected bool
+	}{
+		{
+			name:     "no env var",
+			err:      createDialErr("lookup lkahsdfasdf: no such host"),
+			envValue: "",
+			expected: false,
+		},
+		{
+			name:     "env var false",
+			err:      createDialErr("lookup lkahsdfasdf: no such host"),
+			envValue: "false",
+			expected: false,
+		},
+		{
+			name:     "env var true, matching error",
+			err:      createDialErr("lookup lkahsdfasdf: no such host"),
+			envValue: "true",
+			expected: true,
+		},
+		{
+			name:     "env var true, matching error with extra context",
+			err:      createDialErr("lookup beszel.server.lan on [::1]:53: read udp [::1]:44557->[::1]:53: read: connection refused"),
+			envValue: "true",
+			expected: true,
+		},
+		{
+			name:     "env var true, non-matching error",
+			err:      errors.New("connection refused"),
+			envValue: "true",
+			expected: false,
+		},
+		{
+			name:     "env var true, dial but not lookup",
+			err:      createDialErr("connection timeout"),
+			envValue: "true",
+			expected: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Setenv("EXIT_ON_DNS_ERROR", tt.envValue)
+			result := shouldExitOnErr(tt.err)
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}

agent/handlers.go

@@ -7,6 +7,7 @@ import (
 
 	"github.com/fxamacker/cbor/v2"
 	"github.com/henrygd/beszel/internal/common"
+	"github.com/henrygd/beszel/internal/entities/probe"
 	"github.com/henrygd/beszel/internal/entities/smart"
 
 	"log/slog"
@@ -51,6 +52,7 @@ func NewHandlerRegistry() *HandlerRegistry {
 	registry.Register(common.GetContainerInfo, &GetContainerInfoHandler{})
 	registry.Register(common.GetSmartData, &GetSmartDataHandler{})
 	registry.Register(common.GetSystemdInfo, &GetSystemdInfoHandler{})
+	registry.Register(common.SyncNetworkProbes, &SyncNetworkProbesHandler{})
 
 	return registry
 }
@@ -203,3 +205,19 @@ func (h *GetSystemdInfoHandler) Handle(hctx *HandlerContext) error {
 
 	return hctx.SendResponse(details, hctx.RequestID)
 }
+
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+
+// SyncNetworkProbesHandler handles probe configuration sync from hub
+type SyncNetworkProbesHandler struct{}
+
+func (h *SyncNetworkProbesHandler) Handle(hctx *HandlerContext) error {
+	var configs []probe.Config
+	if err := cbor.Unmarshal(hctx.Request.Data, &configs); err != nil {
+		return err
+	}
+	hctx.Agent.probeManager.SyncProbes(configs)
+	slog.Info("network probes synced", "count", len(configs))
+	return hctx.SendResponse("ok", hctx.RequestID)
+}

agent/probe.go

@@ -0,0 +1,387 @@
+package agent
+
+import (
+	"fmt"
+	"math"
+	"net"
+	"net/http"
+	"sync"
+	"time"
+
+	"log/slog"
+
+	"github.com/henrygd/beszel/internal/entities/probe"
+)
+
+// Probe functionality overview:
+// Probes run at user-defined intervals (e.g., every 10s).
+// To keep memory usage low and constant, data is stored in two layers:
+// 1. Raw samples: The most recent individual results (kept for probeRawRetention).
+// 2. Minute buckets: A fixed-size ring buffer of 61 buckets, each representing one
+//    wall-clock minute. Samples collected within the same minute are aggregated
+//    (sum, min, max, count) into a single bucket.
+//
+// Short-term requests (<= 2m) use raw samples for perfect accuracy.
+// Long-term requests (up to 1h) use the minute buckets to avoid storing thousands
+// of individual data points.
+
+const (
+	// probeRawRetention is the duration to keep individual samples for high-precision short-term requests
+	probeRawRetention = 80 * time.Second
+	// probeMinuteBucketLen is the number of 1-minute buckets to keep (1 hour + 1 for partials)
+	probeMinuteBucketLen int32 = 61
+)
+
+// ProbeManager manages network probe tasks.
+type ProbeManager struct {
+	mu         sync.RWMutex
+	probes     map[string]*probeTask // key = probe.Config.Key()
+	httpClient *http.Client
+}
+
+// probeTask owns retention buffers and cancellation for a single probe config.
+type probeTask struct {
+	config  probe.Config
+	cancel  chan struct{}
+	mu      sync.Mutex
+	samples []probeSample
+	buckets [probeMinuteBucketLen]probeBucket
+}
+
+// probeSample stores one probe attempt and its collection time.
+type probeSample struct {
+	responseMs float64 // -1 means loss
+	timestamp  time.Time
+}
+
+// probeBucket stores one minute of aggregated probe data.
+type probeBucket struct {
+	minute int32
+	filled bool
+	stats  probeAggregate
+}
+
+// probeAggregate accumulates successful response stats and total sample counts.
+type probeAggregate struct {
+	sumMs        float64
+	minMs        float64
+	maxMs        float64
+	totalCount   int
+	successCount int
+}
+
+func newProbeManager() *ProbeManager {
+	return &ProbeManager{
+		probes:     make(map[string]*probeTask),
+		httpClient: &http.Client{Timeout: 10 * time.Second},
+	}
+}
+
+// newProbeAggregate initializes an aggregate with an unset minimum value.
+func newProbeAggregate() probeAggregate {
+	return probeAggregate{minMs: math.MaxFloat64}
+}
+
+// addResponse folds a single probe sample into the aggregate.
+func (agg *probeAggregate) addResponse(responseMs float64) {
+	agg.totalCount++
+	if responseMs < 0 {
+		return
+	}
+	agg.successCount++
+	agg.sumMs += responseMs
+	if responseMs < agg.minMs {
+		agg.minMs = responseMs
+	}
+	if responseMs > agg.maxMs {
+		agg.maxMs = responseMs
+	}
+}
+
+// addAggregate merges another aggregate into this one.
+func (agg *probeAggregate) addAggregate(other probeAggregate) {
+	if other.totalCount == 0 {
+		return
+	}
+	agg.totalCount += other.totalCount
+	agg.successCount += other.successCount
+	agg.sumMs += other.sumMs
+	if other.successCount == 0 {
+		return
+	}
+	if agg.minMs == math.MaxFloat64 || other.minMs < agg.minMs {
+		agg.minMs = other.minMs
+	}
+	if other.maxMs > agg.maxMs {
+		agg.maxMs = other.maxMs
+	}
+}
+
+// hasData reports whether the aggregate contains any samples.
+func (agg probeAggregate) hasData() bool {
+	return agg.totalCount > 0
+}
+
+// result converts the aggregate into the probe result slice format.
+func (agg probeAggregate) result() probe.Result {
+	avg := agg.avgResponse()
+	minMs := 0.0
+	if agg.successCount > 0 {
+		minMs = math.Round(agg.minMs*100) / 100
+	}
+	return probe.Result{
+		avg,
+		minMs,
+		math.Round(agg.maxMs*100) / 100,
+		agg.lossPercentage(),
+	}
+}
+
+// avgResponse returns the rounded average of successful samples.
+func (agg probeAggregate) avgResponse() float64 {
+	if agg.successCount == 0 {
+		return 0
+	}
+	return math.Round(agg.sumMs/float64(agg.successCount)*100) / 100
+}
+
+// lossPercentage returns the rounded failure rate for the aggregate.
+func (agg probeAggregate) lossPercentage() float64 {
+	if agg.totalCount == 0 {
+		return 0
+	}
+	return math.Round(float64(agg.totalCount-agg.successCount)/float64(agg.totalCount)*10000) / 100
+}
+
+// SyncProbes replaces all probe tasks with the given configs.
+func (pm *ProbeManager) SyncProbes(configs []probe.Config) {
+	pm.mu.Lock()
+	defer pm.mu.Unlock()
+
+	// Build set of new keys
+	newKeys := make(map[string]probe.Config, len(configs))
+	for _, cfg := range configs {
+		if cfg.ID == "" {
+			continue
+		}
+		newKeys[cfg.ID] = cfg
+	}
+
+	// Stop removed probes
+	for key, task := range pm.probes {
+		if _, exists := newKeys[key]; !exists {
+			close(task.cancel)
+			delete(pm.probes, key)
+		}
+	}
+
+	// Start new probes (skip existing ones with same key)
+	for key, cfg := range newKeys {
+		if _, exists := pm.probes[key]; exists {
+			continue
+		}
+		task := &probeTask{
+			config:  cfg,
+			cancel:  make(chan struct{}),
+			samples: make([]probeSample, 0, 64),
+		}
+		pm.probes[key] = task
+		go pm.runProbe(task)
+	}
+}
+
+// GetResults returns aggregated results for all probes over the last supplied duration in ms.
+func (pm *ProbeManager) GetResults(durationMs uint16) map[string]probe.Result {
+	pm.mu.RLock()
+	defer pm.mu.RUnlock()
+
+	results := make(map[string]probe.Result, len(pm.probes))
+	now := time.Now()
+	duration := time.Duration(durationMs) * time.Millisecond
+
+	for _, task := range pm.probes {
+		task.mu.Lock()
+		agg := task.aggregateLocked(duration, now)
+		hourAgg := task.aggregateLocked(time.Hour, now)
+		task.mu.Unlock()
+
+		if !agg.hasData() {
+			continue
+		}
+
+		result := agg.result()
+		hourAvg := hourAgg.avgResponse()
+		hourLoss := hourAgg.lossPercentage()
+		if hourAgg.successCount > 0 {
+			result = probe.Result{
+				result[0],
+				hourAvg,
+				math.Round(hourAgg.minMs*100) / 100,
+				math.Round(hourAgg.maxMs*100) / 100,
+				hourLoss,
+			}
+		} else {
+			result = probe.Result{result[0], hourAvg, 0, 0, hourLoss}
+		}
+		results[task.config.ID] = result
+	}
+
+	return results
+}
+
+// Stop stops all probe tasks.
+func (pm *ProbeManager) Stop() {
+	pm.mu.Lock()
+	defer pm.mu.Unlock()
+	for key, task := range pm.probes {
+		close(task.cancel)
+		delete(pm.probes, key)
+	}
+}
+
+// runProbe executes a single probe task in a loop.
+func (pm *ProbeManager) runProbe(task *probeTask) {
+	interval := time.Duration(task.config.Interval) * time.Second
+	if interval < time.Second {
+		interval = 10 * time.Second
+	}
+	ticker := time.Tick(interval)
+
+	// Run immediately on start
+	pm.executeProbe(task)
+
+	for {
+		select {
+		case <-task.cancel:
+			return
+		case <-ticker:
+			pm.executeProbe(task)
+		}
+	}
+}
+
+// aggregateLocked collects probe data for the requested time window.
+func (task *probeTask) aggregateLocked(duration time.Duration, now time.Time) probeAggregate {
+	cutoff := now.Add(-duration)
+	// Keep short windows exact; longer windows read from minute buckets to avoid raw-sample retention.
+	if duration <= probeRawRetention {
+		return aggregateSamplesSince(task.samples, cutoff)
+	}
+	return aggregateBucketsSince(task.buckets[:], cutoff, now)
+}
+
+// aggregateSamplesSince aggregates raw samples newer than the cutoff.
+func aggregateSamplesSince(samples []probeSample, cutoff time.Time) probeAggregate {
+	agg := newProbeAggregate()
+	for _, sample := range samples {
+		if sample.timestamp.Before(cutoff) {
+			continue
+		}
+		agg.addResponse(sample.responseMs)
+	}
+	return agg
+}
+
+// aggregateBucketsSince aggregates minute buckets overlapping the requested window.
+func aggregateBucketsSince(buckets []probeBucket, cutoff, now time.Time) probeAggregate {
+	agg := newProbeAggregate()
+	startMinute := int32(cutoff.Unix() / 60)
+	endMinute := int32(now.Unix() / 60)
+	for _, bucket := range buckets {
+		if !bucket.filled || bucket.minute < startMinute || bucket.minute > endMinute {
+			continue
+		}
+		agg.addAggregate(bucket.stats)
+	}
+	return agg
+}
+
+// addSampleLocked stores a fresh sample in both raw and per-minute retention buffers.
+func (task *probeTask) addSampleLocked(sample probeSample) {
+	cutoff := sample.timestamp.Add(-probeRawRetention)
+	start := 0
+	for i := range task.samples {
+		if !task.samples[i].timestamp.Before(cutoff) {
+			start = i
+			break
+		}
+		if i == len(task.samples)-1 {
+			start = len(task.samples)
+		}
+	}
+	if start > 0 {
+		size := copy(task.samples, task.samples[start:])
+		task.samples = task.samples[:size]
+	}
+	task.samples = append(task.samples, sample)
+
+	minute := int32(sample.timestamp.Unix() / 60)
+	// Each slot stores one wall-clock minute, so the ring stays fixed-size at ~1h per probe.
+	bucket := &task.buckets[minute%probeMinuteBucketLen]
+	if !bucket.filled || bucket.minute != minute {
+		bucket.minute = minute
+		bucket.filled = true
+		bucket.stats = newProbeAggregate()
+	}
+	bucket.stats.addResponse(sample.responseMs)
+}
+
+// executeProbe runs the configured probe and records the sample.
+func (pm *ProbeManager) executeProbe(task *probeTask) {
+	var responseMs float64
+
+	switch task.config.Protocol {
+	case "icmp":
+		responseMs = probeICMP(task.config.Target)
+	case "tcp":
+		responseMs = probeTCP(task.config.Target, task.config.Port)
+	case "http":
+		responseMs = probeHTTP(pm.httpClient, task.config.Target)
+	default:
+		slog.Warn("unknown probe protocol", "protocol", task.config.Protocol)
+		return
+	}
+
+	sample := probeSample{
+		responseMs: responseMs,
+		timestamp:  time.Now(),
+	}
+
+	task.mu.Lock()
+	task.addSampleLocked(sample)
+	task.mu.Unlock()
+}
+
+// probeTCP measures pure TCP handshake response (excluding DNS resolution).
+// Returns -1 on failure.
+func probeTCP(target string, port uint16) float64 {
+	// Resolve DNS first, outside the timing window
+	ips, err := net.LookupHost(target)
+	if err != nil || len(ips) == 0 {
+		return -1
+	}
+	addr := net.JoinHostPort(ips[0], fmt.Sprintf("%d", port))
+
+	// Measure only the TCP handshake
+	start := time.Now()
+	conn, err := net.DialTimeout("tcp", addr, 3*time.Second)
+	if err != nil {
+		return -1
+	}
+	conn.Close()
+	return float64(time.Since(start).Microseconds()) / 1000.0
+}
+
+// probeHTTP measures HTTP GET request response. Returns -1 on failure.
+func probeHTTP(client *http.Client, url string) float64 {
+	start := time.Now()
+	resp, err := client.Get(url)
+	if err != nil {
+		return -1
+	}
+	resp.Body.Close()
+	if resp.StatusCode >= 400 {
+		return -1
+	}
+	return float64(time.Since(start).Microseconds()) / 1000.0
+}

agent/probe_ping.go

@@ -0,0 +1,242 @@
+package agent
+
+import (
+	"net"
+	"os"
+	"os/exec"
+	"regexp"
+	"runtime"
+	"strconv"
+	"sync"
+	"time"
+
+	"golang.org/x/net/icmp"
+	"golang.org/x/net/ipv4"
+	"golang.org/x/net/ipv6"
+
+	"log/slog"
+)
+
+var pingTimeRegex = regexp.MustCompile(`time[=<]([\d.]+)\s*ms`)
+
+type icmpPacketConn interface {
+	Close() error
+}
+
+// icmpMethod tracks which ICMP approach to use. Once a method succeeds or
+// all native methods fail, the choice is cached so subsequent probes skip
+// the trial-and-error overhead.
+type icmpMethod int
+
+const (
+	icmpUntried      icmpMethod = iota // haven't tried yet
+	icmpRaw                            // privileged raw socket
+	icmpDatagram                       // unprivileged datagram socket
+	icmpExecFallback                   // shell out to system ping command
+)
+
+// icmpFamily holds the network parameters and cached detection result for one address family.
+type icmpFamily struct {
+	rawNetwork   string    // e.g. "ip4:icmp" or "ip6:ipv6-icmp"
+	dgramNetwork string    // e.g. "udp4" or "udp6"
+	listenAddr   string    // "0.0.0.0" or "::"
+	echoType     icmp.Type // outgoing echo request type
+	replyType    icmp.Type // expected echo reply type
+	proto        int       // IANA protocol number for parsing replies
+	isIPv6       bool
+	mode         icmpMethod // cached detection result (guarded by icmpModeMu)
+}
+
+var (
+	icmpV4 = icmpFamily{
+		rawNetwork:   "ip4:icmp",
+		dgramNetwork: "udp4",
+		listenAddr:   "0.0.0.0",
+		echoType:     ipv4.ICMPTypeEcho,
+		replyType:    ipv4.ICMPTypeEchoReply,
+		proto:        1,
+	}
+	icmpV6 = icmpFamily{
+		rawNetwork:   "ip6:ipv6-icmp",
+		dgramNetwork: "udp6",
+		listenAddr:   "::",
+		echoType:     ipv6.ICMPTypeEchoRequest,
+		replyType:    ipv6.ICMPTypeEchoReply,
+		proto:        58,
+		isIPv6:       true,
+	}
+	icmpModeMu sync.Mutex
+	icmpListen = func(network, listenAddr string) (icmpPacketConn, error) {
+		return icmp.ListenPacket(network, listenAddr)
+	}
+)
+
+// probeICMP sends an ICMP echo request and measures round-trip response.
+// Supports both IPv4 and IPv6 targets. The ICMP method (raw socket,
+// unprivileged datagram, or exec fallback) is detected once per address
+// family and cached for subsequent probes.
+// Returns response in milliseconds, or -1 on failure.
+func probeICMP(target string) float64 {
+	family, ip := resolveICMPTarget(target)
+	if family == nil {
+		return -1
+	}
+
+	icmpModeMu.Lock()
+	if family.mode == icmpUntried {
+		family.mode = detectICMPMode(family, icmpListen)
+	}
+	mode := family.mode
+	icmpModeMu.Unlock()
+
+	switch mode {
+	case icmpRaw:
+		return probeICMPNative(family.rawNetwork, family, &net.IPAddr{IP: ip})
+	case icmpDatagram:
+		return probeICMPNative(family.dgramNetwork, family, &net.UDPAddr{IP: ip})
+	case icmpExecFallback:
+		return probeICMPExec(target, family.isIPv6)
+	default:
+		return -1
+	}
+}
+
+// resolveICMPTarget resolves a target hostname or IP to determine the address
+// family and concrete IP address. Prefers IPv4 for dual-stack hostnames.
+func resolveICMPTarget(target string) (*icmpFamily, net.IP) {
+	if ip := net.ParseIP(target); ip != nil {
+		if ip.To4() != nil {
+			return &icmpV4, ip.To4()
+		}
+		return &icmpV6, ip
+	}
+
+	ips, err := net.LookupIP(target)
+	if err != nil || len(ips) == 0 {
+		return nil, nil
+	}
+	for _, ip := range ips {
+		if v4 := ip.To4(); v4 != nil {
+			return &icmpV4, v4
+		}
+	}
+	return &icmpV6, ips[0]
+}
+
+func detectICMPMode(family *icmpFamily, listen func(network, listenAddr string) (icmpPacketConn, error)) icmpMethod {
+	label := "IPv4"
+	if family.isIPv6 {
+		label = "IPv6"
+	}
+
+	if conn, err := listen(family.rawNetwork, family.listenAddr); err == nil {
+		conn.Close()
+		slog.Info("ICMP probe using raw socket", "family", label)
+		return icmpRaw
+	} else {
+		slog.Debug("ICMP raw socket unavailable", "family", label, "err", err)
+	}
+
+	if conn, err := listen(family.dgramNetwork, family.listenAddr); err == nil {
+		conn.Close()
+		slog.Info("ICMP probe using unprivileged datagram socket", "family", label)
+		return icmpDatagram
+	} else {
+		slog.Debug("ICMP datagram socket unavailable", "family", label, "err", err)
+	}
+
+	slog.Info("ICMP probe falling back to system ping command", "family", label)
+	return icmpExecFallback
+}
+
+// probeICMPNative sends an ICMP echo request using Go's x/net/icmp package.
+func probeICMPNative(network string, family *icmpFamily, dst net.Addr) float64 {
+	conn, err := icmp.ListenPacket(network, family.listenAddr)
+	if err != nil {
+		return -1
+	}
+	defer conn.Close()
+
+	// Build ICMP echo request
+	msg := &icmp.Message{
+		Type: family.echoType,
+		Code: 0,
+		Body: &icmp.Echo{
+			ID:   os.Getpid() & 0xffff,
+			Seq:  1,
+			Data: []byte("beszel-probe"),
+		},
+	}
+	msgBytes, err := msg.Marshal(nil)
+	if err != nil {
+		return -1
+	}
+
+	// Set deadline before sending
+	conn.SetDeadline(time.Now().Add(3 * time.Second))
+
+	start := time.Now()
+	if _, err := conn.WriteTo(msgBytes, dst); err != nil {
+		return -1
+	}
+
+	// Read reply
+	buf := make([]byte, 1500)
+	for {
+		n, _, err := conn.ReadFrom(buf)
+		if err != nil {
+			return -1
+		}
+
+		reply, err := icmp.ParseMessage(family.proto, buf[:n])
+		if err != nil {
+			return -1
+		}
+
+		if reply.Type == family.replyType {
+			return float64(time.Since(start).Microseconds()) / 1000.0
+		}
+		// Ignore non-echo-reply messages (e.g. destination unreachable) and keep reading
+	}
+}
+
+// probeICMPExec falls back to the system ping command. Returns -1 on failure.
+func probeICMPExec(target string, isIPv6 bool) float64 {
+	var cmd *exec.Cmd
+	switch runtime.GOOS {
+	case "windows":
+		if isIPv6 {
+			cmd = exec.Command("ping", "-6", "-n", "1", "-w", "3000", target)
+		} else {
+			cmd = exec.Command("ping", "-n", "1", "-w", "3000", target)
+		}
+	default: // linux, darwin, freebsd
+		if isIPv6 {
+			cmd = exec.Command("ping", "-6", "-c", "1", "-W", "3", target)
+		} else {
+			cmd = exec.Command("ping", "-c", "1", "-W", "3", target)
+		}
+	}
+
+	start := time.Now()
+	output, err := cmd.Output()
+	if err != nil {
+		// If ping fails but we got output, still try to parse
+		if len(output) == 0 {
+			return -1
+		}
+	}
+
+	matches := pingTimeRegex.FindSubmatch(output)
+	if len(matches) >= 2 {
+		if ms, err := strconv.ParseFloat(string(matches[1]), 64); err == nil {
+			return ms
+		}
+	}
+
+	// Fallback: use wall clock time if ping succeeded but parsing failed
+	if err == nil {
+		return float64(time.Since(start).Microseconds()) / 1000.0
+	}
+	return -1
+}

agent/probe_ping_test.go

@@ -0,0 +1,118 @@
+//go:build testing
+
+package agent
+
+import (
+	"errors"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+type testICMPPacketConn struct{}
+
+func (testICMPPacketConn) Close() error { return nil }
+
+func TestDetectICMPMode(t *testing.T) {
+	tests := []struct {
+		name         string
+		family       *icmpFamily
+		rawErr       error
+		udpErr       error
+		want         icmpMethod
+		wantNetworks []string
+	}{
+		{
+			name:         "IPv4 prefers raw socket when available",
+			family:       &icmpV4,
+			want:         icmpRaw,
+			wantNetworks: []string{"ip4:icmp"},
+		},
+		{
+			name:         "IPv4 uses datagram when raw unavailable",
+			family:       &icmpV4,
+			rawErr:       errors.New("operation not permitted"),
+			want:         icmpDatagram,
+			wantNetworks: []string{"ip4:icmp", "udp4"},
+		},
+		{
+			name:         "IPv4 falls back to exec when both unavailable",
+			family:       &icmpV4,
+			rawErr:       errors.New("operation not permitted"),
+			udpErr:       errors.New("protocol not supported"),
+			want:         icmpExecFallback,
+			wantNetworks: []string{"ip4:icmp", "udp4"},
+		},
+		{
+			name:         "IPv6 prefers raw socket when available",
+			family:       &icmpV6,
+			want:         icmpRaw,
+			wantNetworks: []string{"ip6:ipv6-icmp"},
+		},
+		{
+			name:         "IPv6 uses datagram when raw unavailable",
+			family:       &icmpV6,
+			rawErr:       errors.New("operation not permitted"),
+			want:         icmpDatagram,
+			wantNetworks: []string{"ip6:ipv6-icmp", "udp6"},
+		},
+		{
+			name:         "IPv6 falls back to exec when both unavailable",
+			family:       &icmpV6,
+			rawErr:       errors.New("operation not permitted"),
+			udpErr:       errors.New("protocol not supported"),
+			want:         icmpExecFallback,
+			wantNetworks: []string{"ip6:ipv6-icmp", "udp6"},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			calls := make([]string, 0, 2)
+			listen := func(network, listenAddr string) (icmpPacketConn, error) {
+				require.Equal(t, tt.family.listenAddr, listenAddr)
+				calls = append(calls, network)
+				switch network {
+				case tt.family.rawNetwork:
+					if tt.rawErr != nil {
+						return nil, tt.rawErr
+					}
+				case tt.family.dgramNetwork:
+					if tt.udpErr != nil {
+						return nil, tt.udpErr
+					}
+				default:
+					t.Fatalf("unexpected network %q", network)
+				}
+				return testICMPPacketConn{}, nil
+			}
+
+			assert.Equal(t, tt.want, detectICMPMode(tt.family, listen))
+			assert.Equal(t, tt.wantNetworks, calls)
+		})
+	}
+}
+
+func TestResolveICMPTarget(t *testing.T) {
+	t.Run("IPv4 literal", func(t *testing.T) {
+		family, ip := resolveICMPTarget("127.0.0.1")
+		require.NotNil(t, family)
+		assert.False(t, family.isIPv6)
+		assert.Equal(t, "127.0.0.1", ip.String())
+	})
+
+	t.Run("IPv6 literal", func(t *testing.T) {
+		family, ip := resolveICMPTarget("::1")
+		require.NotNil(t, family)
+		assert.True(t, family.isIPv6)
+		assert.Equal(t, "::1", ip.String())
+	})
+
+	t.Run("IPv4-mapped IPv6 resolves as IPv4", func(t *testing.T) {
+		family, ip := resolveICMPTarget("::ffff:127.0.0.1")
+		require.NotNil(t, family)
+		assert.False(t, family.isIPv6)
+		assert.Equal(t, "127.0.0.1", ip.String())
+	})
+}

agent/probe_test.go

@@ -0,0 +1,216 @@
+package agent
+
+import (
+	"net"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+	"time"
+
+	"github.com/henrygd/beszel/internal/entities/probe"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestProbeTaskAggregateLockedUsesRawSamplesForShortWindows(t *testing.T) {
+	now := time.Date(2026, time.April, 21, 12, 0, 0, 0, time.UTC)
+	task := &probeTask{}
+
+	task.addSampleLocked(probeSample{responseMs: 10, timestamp: now.Add(-90 * time.Second)})
+	task.addSampleLocked(probeSample{responseMs: 20, timestamp: now.Add(-30 * time.Second)})
+	task.addSampleLocked(probeSample{responseMs: -1, timestamp: now.Add(-10 * time.Second)})
+
+	agg := task.aggregateLocked(time.Minute, now)
+	require.True(t, agg.hasData())
+	assert.Equal(t, 2, agg.totalCount)
+	assert.Equal(t, 1, agg.successCount)
+	assert.Equal(t, 20.0, agg.result()[0])
+	assert.Equal(t, 20.0, agg.result()[1])
+	assert.Equal(t, 20.0, agg.result()[2])
+	assert.Equal(t, 50.0, agg.result()[3])
+}
+
+func TestProbeTaskAggregateLockedUsesMinuteBucketsForLongWindows(t *testing.T) {
+	now := time.Date(2026, time.April, 21, 12, 0, 30, 0, time.UTC)
+	task := &probeTask{}
+
+	task.addSampleLocked(probeSample{responseMs: 10, timestamp: now.Add(-11 * time.Minute)})
+	task.addSampleLocked(probeSample{responseMs: 20, timestamp: now.Add(-9 * time.Minute)})
+	task.addSampleLocked(probeSample{responseMs: 40, timestamp: now.Add(-5 * time.Minute)})
+	task.addSampleLocked(probeSample{responseMs: -1, timestamp: now.Add(-90 * time.Second)})
+	task.addSampleLocked(probeSample{responseMs: 30, timestamp: now.Add(-30 * time.Second)})
+
+	agg := task.aggregateLocked(10*time.Minute, now)
+	require.True(t, agg.hasData())
+	assert.Equal(t, 4, agg.totalCount)
+	assert.Equal(t, 3, agg.successCount)
+	assert.Equal(t, 30.0, agg.result()[0])
+	assert.Equal(t, 20.0, agg.result()[1])
+	assert.Equal(t, 40.0, agg.result()[2])
+	assert.Equal(t, 25.0, agg.result()[3])
+}
+
+func TestProbeTaskAddSampleLockedTrimsRawSamplesButKeepsBucketHistory(t *testing.T) {
+	now := time.Date(2026, time.April, 21, 12, 0, 0, 0, time.UTC)
+	task := &probeTask{}
+
+	task.addSampleLocked(probeSample{responseMs: 10, timestamp: now.Add(-10 * time.Minute)})
+	task.addSampleLocked(probeSample{responseMs: 20, timestamp: now})
+
+	require.Len(t, task.samples, 1)
+	assert.Equal(t, 20.0, task.samples[0].responseMs)
+
+	agg := task.aggregateLocked(10*time.Minute, now)
+	require.True(t, agg.hasData())
+	assert.Equal(t, 2, agg.totalCount)
+	assert.Equal(t, 2, agg.successCount)
+	assert.Equal(t, 15.0, agg.result()[0])
+	assert.Equal(t, 10.0, agg.result()[1])
+	assert.Equal(t, 20.0, agg.result()[2])
+	assert.Equal(t, 0.0, agg.result()[3])
+}
+
+func TestProbeManagerGetResultsIncludesHourResponseRange(t *testing.T) {
+	now := time.Now().UTC()
+	task := &probeTask{config: probe.Config{ID: "probe-1"}}
+	task.addSampleLocked(probeSample{responseMs: 10, timestamp: now.Add(-30 * time.Minute)})
+	task.addSampleLocked(probeSample{responseMs: 20, timestamp: now.Add(-9 * time.Minute)})
+	task.addSampleLocked(probeSample{responseMs: 40, timestamp: now.Add(-5 * time.Minute)})
+	task.addSampleLocked(probeSample{responseMs: -1, timestamp: now.Add(-90 * time.Second)})
+	task.addSampleLocked(probeSample{responseMs: 30, timestamp: now.Add(-30 * time.Second)})
+
+	pm := &ProbeManager{probes: map[string]*probeTask{"icmp:example.com": task}}
+
+	results := pm.GetResults(uint16(time.Minute / time.Millisecond))
+	result, ok := results["probe-1"]
+	require.True(t, ok)
+	require.Len(t, result, 5)
+	assert.Equal(t, 30.0, result[0])
+	assert.Equal(t, 25.0, result[1])
+	assert.Equal(t, 10.0, result[2])
+	assert.Equal(t, 40.0, result[3])
+	assert.Equal(t, 20.0, result[4])
+}
+
+func TestProbeManagerGetResultsIncludesLossOnlyHourData(t *testing.T) {
+	now := time.Now().UTC()
+	task := &probeTask{config: probe.Config{ID: "probe-1"}}
+	task.addSampleLocked(probeSample{responseMs: -1, timestamp: now.Add(-30 * time.Second)})
+	task.addSampleLocked(probeSample{responseMs: -1, timestamp: now.Add(-10 * time.Second)})
+
+	pm := &ProbeManager{probes: map[string]*probeTask{"icmp:example.com": task}}
+
+	results := pm.GetResults(uint16(time.Minute / time.Millisecond))
+	result, ok := results["probe-1"]
+	require.True(t, ok)
+	require.Len(t, result, 5)
+	assert.Equal(t, 0.0, result[0])
+	assert.Equal(t, 0.0, result[1])
+	assert.Equal(t, 0.0, result[2])
+	assert.Equal(t, 0.0, result[3])
+	assert.Equal(t, 100.0, result[4])
+}
+
+func TestProbeConfigResultKeyUsesSyncedID(t *testing.T) {
+	cfg := probe.Config{ID: "probe-1", Target: "1.1.1.1", Protocol: "icmp", Interval: 10}
+	assert.Equal(t, "probe-1", cfg.ID)
+}
+
+func TestProbeManagerSyncProbesSkipsConfigsWithoutStableID(t *testing.T) {
+	validCfg := probe.Config{ID: "probe-1", Target: "https://example.com", Protocol: "http", Interval: 10}
+	invalidCfg := probe.Config{Target: "1.1.1.1", Protocol: "icmp", Interval: 10}
+
+	pm := newProbeManager()
+	pm.SyncProbes([]probe.Config{validCfg, invalidCfg})
+	defer pm.Stop()
+
+	_, validExists := pm.probes[validCfg.ID]
+	_, invalidExists := pm.probes[invalidCfg.ID]
+	assert.True(t, validExists)
+	assert.False(t, invalidExists)
+}
+
+func TestProbeManagerSyncProbesStopsRemovedTasksButKeepsExisting(t *testing.T) {
+	keepCfg := probe.Config{ID: "probe-1", Target: "https://example.com", Protocol: "http", Interval: 10}
+	removeCfg := probe.Config{ID: "probe-2", Target: "1.1.1.1", Protocol: "icmp", Interval: 10}
+
+	keptTask := &probeTask{config: keepCfg, cancel: make(chan struct{})}
+	removedTask := &probeTask{config: removeCfg, cancel: make(chan struct{})}
+	pm := &ProbeManager{
+		probes: map[string]*probeTask{
+			keepCfg.ID:   keptTask,
+			removeCfg.ID: removedTask,
+		},
+	}
+
+	pm.SyncProbes([]probe.Config{keepCfg})
+
+	assert.Same(t, keptTask, pm.probes[keepCfg.ID])
+	_, exists := pm.probes[removeCfg.ID]
+	assert.False(t, exists)
+
+	select {
+	case <-removedTask.cancel:
+	default:
+		t.Fatal("expected removed probe task to be cancelled")
+	}
+
+	select {
+	case <-keptTask.cancel:
+		t.Fatal("expected existing probe task to remain active")
+	default:
+	}
+}
+
+func TestProbeHTTP(t *testing.T) {
+	t.Run("success", func(t *testing.T) {
+		server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(http.StatusNoContent)
+		}))
+		defer server.Close()
+
+		responseMs := probeHTTP(server.Client(), server.URL)
+		assert.GreaterOrEqual(t, responseMs, 0.0)
+	})
+
+	t.Run("server error", func(t *testing.T) {
+		server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			http.Error(w, "boom", http.StatusInternalServerError)
+		}))
+		defer server.Close()
+
+		assert.Equal(t, -1.0, probeHTTP(server.Client(), server.URL))
+	})
+}
+
+func TestProbeTCP(t *testing.T) {
+	t.Run("success", func(t *testing.T) {
+		listener, err := net.Listen("tcp", "127.0.0.1:0")
+		require.NoError(t, err)
+		defer listener.Close()
+
+		accepted := make(chan struct{})
+		go func() {
+			defer close(accepted)
+			conn, err := listener.Accept()
+			if err == nil {
+				_ = conn.Close()
+			}
+		}()
+
+		port := uint16(listener.Addr().(*net.TCPAddr).Port)
+		responseMs := probeTCP("127.0.0.1", port)
+		assert.GreaterOrEqual(t, responseMs, 0.0)
+		<-accepted
+	})
+
+	t.Run("connection failure", func(t *testing.T) {
+		listener, err := net.Listen("tcp", "127.0.0.1:0")
+		require.NoError(t, err)
+
+		port := uint16(listener.Addr().(*net.TCPAddr).Port)
+		require.NoError(t, listener.Close())
+
+		assert.Equal(t, -1.0, probeTCP("127.0.0.1", port))
+	})
+}

go.mod

@@ -20,6 +20,7 @@ require (
 	github.com/stretchr/testify v1.11.1
 	golang.org/x/crypto v0.49.0
 	golang.org/x/exp v0.0.0-20260312153236-7ab1446f8b90
+	golang.org/x/net v0.52.0
 	golang.org/x/sys v0.42.0
 	gopkg.in/yaml.v3 v3.0.1
 	howett.net/plist v1.0.1
@@ -56,7 +57,6 @@ require (
 	github.com/x448/float16 v0.8.4 // indirect
 	github.com/yusufpapurcu/wmi v1.2.4 // indirect
 	golang.org/x/image v0.38.0 // indirect
-	golang.org/x/net v0.52.0 // indirect
 	golang.org/x/oauth2 v0.36.0 // indirect
 	golang.org/x/sync v0.20.0 // indirect
 	golang.org/x/term v0.41.0 // indirect

internal/cmd/agent/agent.go

@@ -195,6 +195,6 @@ func main() {
 	}
 
 	if err := a.Start(serverConfig); err != nil {
-		log.Fatal("Failed to start server: ", err)
+		log.Fatal("Failed to start: ", err)
 	}
 }

internal/common/common-ws.go

@@ -22,6 +22,8 @@ const (
 	GetSmartData
 	// Request detailed systemd service info from agent
 	GetSystemdInfo
+	// Sync network probe configuration to agent
+	SyncNetworkProbes
 	// Add new actions here...
 )
 

internal/entities/probe/probe.go

@@ -0,0 +1,24 @@
+package probe
+
+// Config defines a network probe task sent from hub to agent.
+type Config struct {
+	// ID is the stable network_probes record ID generated by the hub.
+	ID       string `cbor:"0,keyasint"`
+	Target   string `cbor:"1,keyasint"`
+	Protocol string `cbor:"2,keyasint"` // "icmp", "tcp", or "http"
+	Port     uint16 `cbor:"3,keyasint,omitempty"`
+	Interval uint16 `cbor:"4,keyasint"` // seconds
+}
+
+// Result holds aggregated probe results for a single target.
+//
+// 0: avg response in ms
+//
+// 1: average response over the last hour in ms
+//
+// 2: min response over the last hour in ms
+//
+// 3: max response over the last hour in ms
+//
+// 4: packet loss percentage over the last hour (0-100)
+type Result []float64

internal/entities/system/system.go

@@ -7,6 +7,7 @@ import (
 	"time"
 
 	"github.com/henrygd/beszel/internal/entities/container"
+	"github.com/henrygd/beszel/internal/entities/probe"
 	"github.com/henrygd/beszel/internal/entities/systemd"
 )
 
@@ -174,9 +175,10 @@ type Details struct {
 
 // Final data structure to return to the hub
 type CombinedData struct {
-	Stats           Stats              `json:"stats" cbor:"0,keyasint"`
-	Info            Info               `json:"info" cbor:"1,keyasint"`
-	Containers      []*container.Stats `json:"container" cbor:"2,keyasint"`
-	SystemdServices []*systemd.Service `json:"systemd,omitempty" cbor:"3,keyasint,omitempty"`
-	Details         *Details           `cbor:"4,keyasint,omitempty"`
+	Stats           Stats                   `json:"stats" cbor:"0,keyasint"`
+	Info            Info                    `json:"info" cbor:"1,keyasint"`
+	Containers      []*container.Stats      `json:"container" cbor:"2,keyasint"`
+	SystemdServices []*systemd.Service      `json:"systemd,omitempty" cbor:"3,keyasint,omitempty"`
+	Details         *Details                `cbor:"4,keyasint,omitempty"`
+	Probes          map[string]probe.Result `cbor:"5,keyasint,omitempty"`
 }

internal/hub/collections.go

@@ -78,7 +78,7 @@ func setCollectionAuthSettings(app core.App) error {
 		return err
 	}
 
-	if err := applyCollectionRules(app, []string{"containers", "container_stats", "system_stats", "systemd_services"}, collectionRules{
+	if err := applyCollectionRules(app, []string{"containers", "container_stats", "system_stats", "systemd_services", "network_probe_stats"}, collectionRules{
 		list: &systemScopedReadRule,
 	}); err != nil {
 		return err
@@ -92,7 +92,7 @@ func setCollectionAuthSettings(app core.App) error {
 		return err
 	}
 
-	if err := applyCollectionRules(app, []string{"fingerprints"}, collectionRules{
+	if err := applyCollectionRules(app, []string{"fingerprints", "network_probes"}, collectionRules{
 		list:   &systemScopedReadRule,
 		view:   &systemScopedReadRule,
 		create: &systemScopedWriteRule,

internal/hub/hub.go

@@ -81,6 +81,7 @@ func (h *Hub) StartHub() error {
 		}
 		// register middlewares
 		h.registerMiddlewares(e)
+		// bind events that aren't set up in different
 		// register api routes
 		if err := h.registerApiRoutes(e); err != nil {
 			return err
@@ -109,6 +110,8 @@ func (h *Hub) StartHub() error {
 	h.App.OnRecordCreate("users").BindFunc(h.um.InitializeUserRole)
 	h.App.OnRecordCreate("user_settings").BindFunc(h.um.InitializeUserSettings)
 
+	bindNetworkProbesEvents(h)
+
 	pb, ok := h.App.(*pocketbase.PocketBase)
 	if !ok {
 		return errors.New("not a pocketbase app")

internal/hub/probes.go

@@ -0,0 +1,62 @@
+package hub
+
+import (
+	"strconv"
+
+	"github.com/henrygd/beszel/internal/entities/probe"
+	"github.com/henrygd/beszel/internal/hub/systems"
+	"github.com/pocketbase/pocketbase/core"
+)
+
+// generateProbeID creates a stable hash ID for a probe based on its configuration and the system it belongs to.
+func generateProbeID(systemId string, config probe.Config) string {
+	intervalStr := strconv.FormatUint(uint64(config.Interval), 10)
+	portStr := strconv.FormatUint(uint64(config.Port), 10)
+	return systems.MakeStableHashId(systemId, config.Protocol, config.Target, portStr, intervalStr)
+}
+
+func bindNetworkProbesEvents(h *Hub) {
+	// on create, make sure the id is set to a stable hash
+	h.OnRecordCreate("network_probes").BindFunc(func(e *core.RecordEvent) error {
+		systemID := e.Record.GetString("system")
+		config := &probe.Config{
+			Target:   e.Record.GetString("target"),
+			Protocol: e.Record.GetString("protocol"),
+			Port:     uint16(e.Record.GetInt("port")),
+			Interval: uint16(e.Record.GetInt("interval")),
+		}
+		id := generateProbeID(systemID, *config)
+		e.Record.Set("id", id)
+		return e.Next()
+	})
+
+	// sync probe to agent on creation
+	h.OnRecordAfterCreateSuccess("network_probes").BindFunc(func(e *core.RecordEvent) error {
+		systemID := e.Record.GetString("system")
+		h.syncProbesToAgent(systemID)
+		return e.Next()
+	})
+	// sync probe to agent on delete
+	h.OnRecordAfterDeleteSuccess("network_probes").BindFunc(func(e *core.RecordEvent) error {
+		systemID := e.Record.GetString("system")
+		h.syncProbesToAgent(systemID)
+		return e.Next()
+	})
+	// TODO: if enabled changes, sync to agent
+}
+
+// syncProbesToAgent fetches enabled probes for a system and sends them to the agent.
+func (h *Hub) syncProbesToAgent(systemID string) {
+	system, err := h.sm.GetSystem(systemID)
+	if err != nil {
+		return
+	}
+
+	configs := h.sm.GetProbeConfigsForSystem(systemID)
+
+	go func() {
+		if err := system.SyncNetworkProbes(configs); err != nil {
+			h.Logger().Warn("failed to sync probes to agent", "system", systemID, "err", err)
+		}
+	}()
+}

internal/hub/probes_test.go

@@ -0,0 +1,79 @@
+package hub
+
+import (
+	"testing"
+
+	"github.com/henrygd/beszel/internal/entities/probe"
+	"github.com/stretchr/testify/assert"
+)
+
+func TestGenerateProbeID(t *testing.T) {
+	tests := []struct {
+		name     string
+		systemID string
+		config   probe.Config
+		expected string
+	}{
+		{
+			name:     "HTTP probe on example.com",
+			systemID: "sys123",
+			config: probe.Config{
+				Protocol: "http",
+				Target:   "example.com",
+				Port:     80,
+				Interval: 60,
+			},
+			expected: "d5f27931",
+		},
+		{
+			name:     "HTTP probe on example.com with different system ID",
+			systemID: "sys1234",
+			config: probe.Config{
+				Protocol: "http",
+				Target:   "example.com",
+				Port:     80,
+				Interval: 60,
+			},
+			expected: "6f8b17f1",
+		},
+		{
+			name:     "Same probe, different interval",
+			systemID: "sys1234",
+			config: probe.Config{
+				Protocol: "http",
+				Target:   "example.com",
+				Port:     80,
+				Interval: 120,
+			},
+			expected: "6d4baf8",
+		},
+		{
+			name:     "ICMP probe on 1.1.1.1",
+			systemID: "sys456",
+			config: probe.Config{
+				Protocol: "icmp",
+				Target:   "1.1.1.1",
+				Port:     0,
+				Interval: 10,
+			},
+			expected: "80b5836b",
+		}, {
+			name:     "ICMP probe on 1.1.1.1 with different system ID",
+			systemID: "sys4567",
+			config: probe.Config{
+				Protocol: "icmp",
+				Target:   "1.1.1.1",
+				Port:     0,
+				Interval: 10,
+			},
+			expected: "a6652680",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := generateProbeID(tt.systemID, tt.config)
+			assert.Equal(t, tt.expected, got, "generateProbeID() = %v, want %v", got, tt.expected)
+		})
+	}
+}

internal/hub/systems/system.go

@@ -6,6 +6,7 @@ import (
 	"errors"
 	"fmt"
 	"hash/fnv"
+	"log/slog"
 	"math/rand"
 	"net"
 	"strings"
@@ -18,6 +19,7 @@ import (
 	"github.com/henrygd/beszel/internal/hub/ws"
 
 	"github.com/henrygd/beszel/internal/entities/container"
+	"github.com/henrygd/beszel/internal/entities/probe"
 	"github.com/henrygd/beszel/internal/entities/smart"
 	"github.com/henrygd/beszel/internal/entities/system"
 	"github.com/henrygd/beszel/internal/entities/systemd"
@@ -29,6 +31,7 @@ import (
 	"github.com/lxzan/gws"
 	"github.com/pocketbase/dbx"
 	"github.com/pocketbase/pocketbase/core"
+	"github.com/pocketbase/pocketbase/tools/types"
 	"golang.org/x/crypto/ssh"
 )
 
@@ -238,6 +241,12 @@ func (sys *System) createRecords(data *system.CombinedData) (*core.Record, error
 			}
 		}
 
+		if data.Probes != nil {
+			if err := updateNetworkProbesRecords(txApp, data.Probes, sys.Id); err != nil {
+				return err
+			}
+		}
+
 		// update system record (do this last because it triggers alerts and we need above records to be inserted first)
 		systemRecord.Set("status", up)
 		systemRecord.Set("info", data.Info)
@@ -289,7 +298,7 @@ func createSystemdStatsRecords(app core.App, data []*systemd.Service, systemId s
 	for i, service := range data {
 		suffix := fmt.Sprintf("%d", i)
 		valueStrings = append(valueStrings, fmt.Sprintf("({:id%[1]s}, {:system}, {:name%[1]s}, {:state%[1]s}, {:sub%[1]s}, {:cpu%[1]s}, {:cpuPeak%[1]s}, {:memory%[1]s}, {:memPeak%[1]s}, {:updated})", suffix))
-		params["id"+suffix] = makeStableHashId(systemId, service.Name)
+		params["id"+suffix] = MakeStableHashId(systemId, service.Name)
 		params["name"+suffix] = service.Name
 		params["state"+suffix] = service.State
 		params["sub"+suffix] = service.Sub
@@ -306,6 +315,95 @@ func createSystemdStatsRecords(app core.App, data []*systemd.Service, systemId s
 	return err
 }
 
+func updateNetworkProbesRecords(app core.App, data map[string]probe.Result, systemId string) error {
+	if len(data) == 0 {
+		return nil
+	}
+	var err error
+	collectionName := "network_probes"
+
+	// If realtime updates are active, we save via PocketBase records to trigger realtime events.
+	// Otherwise we can do a more efficient direct update via SQL
+	realtimeActive := utils.RealtimeActiveForCollection(app, collectionName, func(filterQuery string) bool {
+		slog.Info("Checking realtime subscription filter for network probes", "filterQuery", filterQuery)
+		return !strings.Contains(filterQuery, "system") || strings.Contains(filterQuery, systemId)
+	})
+
+	var db dbx.Builder
+	var nowString string
+	var updateQuery *dbx.Query
+	if !realtimeActive {
+		db = app.DB()
+		nowString = time.Now().UTC().Format(types.DefaultDateLayout)
+		sql := fmt.Sprintf("UPDATE %s SET resAvg={:res}, resMin1h={:resMin1h}, resMax1h={:resMax1h}, resAvg1h={:resAvg1h}, loss1h={:loss1h}, updated={:updated} WHERE id={:id}", collectionName)
+		updateQuery = db.NewQuery(sql)
+	}
+
+	// insert network probe stats records
+	switch realtimeActive {
+	case true:
+		collection, _ := app.FindCachedCollectionByNameOrId("network_probe_stats")
+		record := core.NewRecord(collection)
+		record.Set("system", systemId)
+		record.Set("stats", data)
+		record.Set("type", "1m")
+		err = app.SaveNoValidate(record)
+	default:
+		if dataJSON, marshalErr := json.Marshal(data); marshalErr == nil {
+			sql := "INSERT INTO network_probe_stats (system, stats, type, created) VALUES ({:system}, {:stats}, {:type}, {:created})"
+			insertQuery := db.NewQuery(sql)
+			_, err = insertQuery.Bind(dbx.Params{
+				"system":  systemId,
+				"stats":   dataJSON,
+				"type":    "1m",
+				"created": nowString,
+			}).Execute()
+		}
+	}
+	if err != nil {
+		app.Logger().Error("Failed to update probe stats", "system", systemId, "err", err)
+	}
+
+	// update network_probes records
+	for id, values := range data {
+		switch realtimeActive {
+		case true:
+			var record *core.Record
+			record, err = app.FindRecordById(collectionName, id)
+			if err == nil {
+				record.Set("res", probeMetric(values, 0))
+				record.Set("resAvg1h", probeMetric(values, 1))
+				record.Set("resMin1h", probeMetric(values, 2))
+				record.Set("resMax1h", probeMetric(values, 3))
+				record.Set("loss1h", probeMetric(values, 4))
+				err = app.SaveNoValidate(record)
+			}
+		default:
+			_, err = updateQuery.Bind(dbx.Params{
+				"id":       id,
+				"res":      probeMetric(values, 0),
+				"resAvg1h": probeMetric(values, 1),
+				"resMin1h": probeMetric(values, 2),
+				"resMax1h": probeMetric(values, 3),
+				"loss1h":   probeMetric(values, 4),
+				"updated":  nowString,
+			}).Execute()
+		}
+		if err != nil {
+			app.Logger().Warn("Failed to update probe", "system", systemId, "probe", id, "err", err)
+		}
+	}
+
+	return nil
+}
+
+func probeMetric(values probe.Result, index int) float64 {
+	if index < len(values) {
+		return values[index]
+	}
+	return 0
+}
+
 // createContainerRecords creates container records
 func createContainerRecords(app core.App, data []*container.Stats, systemId string) error {
 	if len(data) == 0 {
@@ -540,7 +638,7 @@ func (sys *System) FetchSmartDataFromAgent() (map[string]smart.SmartData, error)
 	return result, err
 }
 
-func makeStableHashId(strings ...string) string {
+func MakeStableHashId(strings ...string) string {
 	hash := fnv.New32a()
 	for _, str := range strings {
 		hash.Write([]byte(str))

internal/hub/systems/system_manager.go

@@ -7,6 +7,7 @@ import (
 
 	"github.com/henrygd/beszel/internal/hub/ws"
 
+	"github.com/henrygd/beszel/internal/entities/probe"
 	"github.com/henrygd/beszel/internal/entities/system"
 	"github.com/henrygd/beszel/internal/hub/expirymap"
 
@@ -15,6 +16,7 @@ import (
 	"github.com/henrygd/beszel"
 
 	"github.com/blang/semver"
+	"github.com/pocketbase/dbx"
 	"github.com/pocketbase/pocketbase/core"
 	"github.com/pocketbase/pocketbase/tools/store"
 	"golang.org/x/crypto/ssh"
@@ -317,6 +319,17 @@ func (sm *SystemManager) AddWebSocketSystem(systemId string, agentVersion semver
 	if err := sm.AddRecord(systemRecord, system); err != nil {
 		return err
 	}
+
+	// Sync network probes to the newly connected agent
+	go func() {
+		configs := sm.GetProbeConfigsForSystem(systemId)
+		if len(configs) > 0 {
+			if err := system.SyncNetworkProbes(configs); err != nil {
+				sm.hub.Logger().Warn("failed to sync probes to agent", "system", systemId, "err", err)
+			}
+		}
+	}()
+
 	return nil
 }
 
@@ -329,6 +342,16 @@ func (sm *SystemManager) resetFailedSmartFetchState(systemID string) {
 	}
 }
 
+// GetProbeConfigsForSystem returns all enabled probe configs for a system.
+func (sm *SystemManager) GetProbeConfigsForSystem(systemID string) []probe.Config {
+	var configs []probe.Config
+	_ = sm.hub.DB().
+		NewQuery("SELECT id, target, protocol, port, interval FROM network_probes WHERE system = {:system} AND enabled = true").
+		Bind(dbx.Params{"system": systemID}).
+		All(&configs)
+	return configs
+}
+
 // createSSHClientConfig initializes the SSH client configuration for connecting to an agent's server
 func (sm *SystemManager) createSSHClientConfig() error {
 	privateKey, err := sm.hub.GetSSHKey("")

internal/hub/systems/system_probes.go

@@ -0,0 +1,57 @@
+package systems
+
+import (
+	"context"
+	"time"
+
+	"github.com/henrygd/beszel/internal/common"
+	"github.com/henrygd/beszel/internal/entities/probe"
+)
+
+// SyncNetworkProbes sends probe configurations to the agent.
+func (sys *System) SyncNetworkProbes(configs []probe.Config) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	var result string
+	return sys.request(ctx, common.SyncNetworkProbes, configs, &result)
+}
+
+// FetchNetworkProbeResults fetches probe results from the agent.
+// func (sys *System) FetchNetworkProbeResults() (map[string]probe.Result, error) {
+// 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+// 	defer cancel()
+// 	var results map[string]probe.Result
+// 	err := sys.request(ctx, common.GetNetworkProbeResults, nil, &results)
+// 	return results, err
+// }
+
+// hasEnabledProbes returns true if this system has any enabled network probes.
+// func (sys *System) hasEnabledProbes() bool {
+// 	count, err := sys.manager.hub.CountRecords("network_probes",
+// 		dbx.NewExp("system = {:system} AND enabled = true", dbx.Params{"system": sys.Id}))
+// 	return err == nil && count > 0
+// }
+
+// fetchAndSaveProbeResults fetches probe results and saves them to the database.
+// func (sys *System) fetchAndSaveProbeResults() {
+// 	hub := sys.manager.hub
+
+// 	results, err := sys.FetchNetworkProbeResults()
+// 	if err != nil || len(results) == 0 {
+// 		return
+// 	}
+
+// 	collection, err := hub.FindCachedCollectionByNameOrId("network_probe_stats")
+// 	if err != nil {
+// 		return
+// 	}
+
+// 	record := core.NewRecord(collection)
+// 	record.Set("system", sys.Id)
+// 	record.Set("stats", results)
+// 	record.Set("type", "1m")
+
+// 	if err := hub.SaveNoValidate(record); err != nil {
+// 		hub.Logger().Warn("failed to save probe stats", "system", sys.Id, "err", err)
+// 	}
+// }

internal/hub/systems/system_smart.go

@@ -84,7 +84,7 @@ func (sys *System) saveSmartDevices(smartData map[string]smart.SmartData) error
 
 func (sys *System) upsertSmartDeviceRecord(collection *core.Collection, deviceKey string, device smart.SmartData) error {
 	hub := sys.manager.hub
-	recordID := makeStableHashId(sys.Id, deviceKey)
+	recordID := MakeStableHashId(sys.Id, deviceKey)
 
 	record, err := hub.FindRecordById(collection, recordID)
 	if err != nil {

internal/hub/systems/system_systemd_test.go

@@ -14,9 +14,9 @@ func TestGetSystemdServiceId(t *testing.T) {
 		serviceName := "nginx.service"
 
 		// Call multiple times and ensure same result
-		id1 := makeStableHashId(systemId, serviceName)
-		id2 := makeStableHashId(systemId, serviceName)
-		id3 := makeStableHashId(systemId, serviceName)
+		id1 := MakeStableHashId(systemId, serviceName)
+		id2 := MakeStableHashId(systemId, serviceName)
+		id3 := MakeStableHashId(systemId, serviceName)
 
 		assert.Equal(t, id1, id2)
 		assert.Equal(t, id2, id3)
@@ -29,10 +29,10 @@ func TestGetSystemdServiceId(t *testing.T) {
 		serviceName1 := "nginx.service"
 		serviceName2 := "apache.service"
 
-		id1 := makeStableHashId(systemId1, serviceName1)
-		id2 := makeStableHashId(systemId2, serviceName1)
-		id3 := makeStableHashId(systemId1, serviceName2)
-		id4 := makeStableHashId(systemId2, serviceName2)
+		id1 := MakeStableHashId(systemId1, serviceName1)
+		id2 := MakeStableHashId(systemId2, serviceName1)
+		id3 := MakeStableHashId(systemId1, serviceName2)
+		id4 := MakeStableHashId(systemId2, serviceName2)
 
 		// All IDs should be different
 		assert.NotEqual(t, id1, id2)
@@ -56,14 +56,14 @@ func TestGetSystemdServiceId(t *testing.T) {
 		}
 
 		for _, tc := range testCases {
-			id := makeStableHashId(tc.systemId, tc.serviceName)
+			id := MakeStableHashId(tc.systemId, tc.serviceName)
 			// FNV-32 produces 8 hex characters
 			assert.Len(t, id, 8, "ID should be 8 characters for systemId='%s', serviceName='%s'", tc.systemId, tc.serviceName)
 		}
 	})
 
 	t.Run("hexadecimal output", func(t *testing.T) {
-		id := makeStableHashId("test-system", "test-service")
+		id := MakeStableHashId("test-system", "test-service")
 		assert.NotEmpty(t, id)
 
 		// Should only contain hexadecimal characters

internal/hub/utils/utils.go

@@ -1,7 +1,11 @@
 // Package utils provides utility functions for the hub.
 package utils
 
-import "os"
+import (
+	"os"
+
+	"github.com/pocketbase/pocketbase/core"
+)
 
 // GetEnv retrieves an environment variable with a "BESZEL_HUB_" prefix, or falls back to the unprefixed key.
 func GetEnv(key string) (value string, exists bool) {
@@ -10,3 +14,26 @@ func GetEnv(key string) (value string, exists bool) {
 	}
 	return os.LookupEnv(key)
 }
+
+// realtimeActiveForCollection checks if there are active WebSocket subscriptions for the given collection.
+func RealtimeActiveForCollection(app core.App, collectionName string, validateFn func(filterQuery string) bool) bool {
+	broker := app.SubscriptionsBroker()
+	if broker.TotalClients() == 0 {
+		return false
+	}
+	for _, client := range broker.Clients() {
+		subs := client.Subscriptions(collectionName)
+		if len(subs) > 0 {
+			if validateFn == nil {
+				return true
+			}
+			for k := range subs {
+				filter := subs[k].Query["filter"]
+				if validateFn(filter) {
+					return true
+				}
+			}
+		}
+	}
+	return false
+}

internal/migrations/0_collections_snapshot_0_19_0_dev_1.go

@@ -1699,6 +1699,223 @@ func init() {
 		"type": "base",
 		"updateRule": null,
 		"viewRule": null
+	},
+	{
+		"id": "np_probes_001",
+		"listRule": null,
+		"viewRule": null,
+		"createRule": null,
+		"updateRule": null,
+		"deleteRule": null,
+		"name": "network_probes",
+		"type": "base",
+		"fields": [
+			{
+				"autogeneratePattern": "[a-z0-9]{15}",
+				"hidden": false,
+				"id": "text3208210256",
+				"max": 15,
+				"min": 15,
+				"name": "id",
+				"pattern": "^[a-z0-9]+$",
+				"presentable": false,
+				"primaryKey": true,
+				"required": true,
+				"system": true,
+				"type": "text"
+			},
+			{
+				"cascadeDelete": true,
+				"collectionId": "2hz5ncl8tizk5nx",
+				"hidden": false,
+				"id": "np_system",
+				"maxSelect": 1,
+				"minSelect": 0,
+				"name": "system",
+				"presentable": false,
+				"required": true,
+				"system": false,
+				"type": "relation"
+			},
+			{
+				"hidden": false,
+				"id": "np_name",
+				"max": 200,
+				"min": 0,
+				"name": "name",
+				"pattern": "",
+				"presentable": false,
+				"primaryKey": false,
+				"required": false,
+				"system": false,
+				"type": "text"
+			},
+			{
+				"hidden": false,
+				"id": "np_target",
+				"max": 500,
+				"min": 1,
+				"name": "target",
+				"pattern": "",
+				"presentable": false,
+				"primaryKey": false,
+				"required": true,
+				"system": false,
+				"type": "text"
+			},
+			{
+				"hidden": false,
+				"id": "np_protocol",
+				"maxSelect": 1,
+				"name": "protocol",
+				"presentable": false,
+				"required": true,
+				"system": false,
+				"type": "select",
+				"values": ["icmp", "tcp", "http"]
+			},
+			{
+				"hidden": false,
+				"id": "np_port",
+				"max": 65535,
+				"min": 0,
+				"name": "port",
+				"onlyInt": true,
+				"presentable": false,
+				"required": false,
+				"system": false,
+				"type": "number"
+			},
+			{
+				"hidden": false,
+				"id": "np_interval",
+				"max": 3600,
+				"min": 1,
+				"name": "interval",
+				"onlyInt": true,
+				"presentable": false,
+				"required": true,
+				"system": false,
+				"type": "number"
+			},
+			{
+				"hidden": false,
+				"id": "np_enabled",
+				"name": "enabled",
+				"presentable": false,
+				"required": false,
+				"system": false,
+				"type": "bool"
+			},
+			{
+				"hidden": false,
+				"id": "autodate2990389176",
+				"name": "created",
+				"onCreate": true,
+				"onUpdate": false,
+				"presentable": false,
+				"system": false,
+				"type": "autodate"
+			},
+			{
+				"hidden": false,
+				"id": "autodate3332085495",
+				"name": "updated",
+				"onCreate": true,
+				"onUpdate": true,
+				"presentable": false,
+				"system": false,
+				"type": "autodate"
+			}
+		],
+		"indexes": [
+			"CREATE INDEX ` + "`" + `idx_np_system_enabled` + "`" + ` ON ` + "`" + `network_probes` + "`" + ` (\n  ` + "`" + `system` + "`" + `,\n  ` + "`" + `enabled` + "`" + `\n)"
+		],
+		"system": false
+	},
+	{
+		"id": "np_stats_001",
+		"listRule": null,
+		"viewRule": null,
+		"createRule": null,
+		"updateRule": null,
+		"deleteRule": null,
+		"name": "network_probe_stats",
+		"type": "base",
+		"fields": [
+			{
+				"autogeneratePattern": "[a-z0-9]{15}",
+				"hidden": false,
+				"id": "text3208210256",
+				"max": 15,
+				"min": 15,
+				"name": "id",
+				"pattern": "^[a-z0-9]+$",
+				"presentable": false,
+				"primaryKey": true,
+				"required": true,
+				"system": true,
+				"type": "text"
+			},
+			{
+				"cascadeDelete": true,
+				"collectionId": "2hz5ncl8tizk5nx",
+				"hidden": false,
+				"id": "nps_system",
+				"maxSelect": 1,
+				"minSelect": 0,
+				"name": "system",
+				"presentable": false,
+				"required": true,
+				"system": false,
+				"type": "relation"
+			},
+			{
+				"hidden": false,
+				"id": "nps_stats",
+				"maxSize": 2000000,
+				"name": "stats",
+				"presentable": false,
+				"required": true,
+				"system": false,
+				"type": "json"
+			},
+			{
+				"hidden": false,
+				"id": "nps_type",
+				"maxSelect": 1,
+				"name": "type",
+				"presentable": false,
+				"required": true,
+				"system": false,
+				"type": "select",
+				"values": ["1m", "10m", "20m", "120m", "480m"]
+			},
+			{
+				"hidden": false,
+				"id": "autodate2990389176",
+				"name": "created",
+				"onCreate": true,
+				"onUpdate": false,
+				"presentable": false,
+				"system": false,
+				"type": "autodate"
+			},
+			{
+				"hidden": false,
+				"id": "autodate3332085495",
+				"name": "updated",
+				"onCreate": true,
+				"onUpdate": true,
+				"presentable": false,
+				"system": false,
+				"type": "autodate"
+			}
+		],
+		"indexes": [
+			"CREATE INDEX ` + "`" + `idx_nps_system_type_created` + "`" + ` ON ` + "`" + `network_probe_stats` + "`" + ` (\n  ` + "`" + `system` + "`" + `,\n  ` + "`" + `type` + "`" + `,\n  ` + "`" + `created` + "`" + `\n)"
+		],
+		"system": false
 	}
 ]`
 

internal/records/probe_averaging_test.go

@@ -0,0 +1,58 @@
+//go:build testing
+
+package records_test
+
+import (
+	"testing"
+
+	"github.com/henrygd/beszel/internal/records"
+	"github.com/henrygd/beszel/internal/tests"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestAverageProbeStats(t *testing.T) {
+	hub, err := tests.NewTestHub(t.TempDir())
+	require.NoError(t, err)
+	defer hub.Cleanup()
+
+	rm := records.NewRecordManager(hub)
+	user, err := tests.CreateUser(hub, "probe-avg@example.com", "testtesttest")
+	require.NoError(t, err)
+	system, err := tests.CreateRecord(hub, "systems", map[string]any{
+		"name":   "probe-avg-system",
+		"host":   "localhost",
+		"port":   "45876",
+		"status": "up",
+		"users":  []string{user.Id},
+	})
+	require.NoError(t, err)
+
+	recordA, err := tests.CreateRecord(hub, "network_probe_stats", map[string]any{
+		"system": system.Id,
+		"type":   "1m",
+		"stats":  `{"icmp:1.1.1.1":[10,80,8,14,1]}`,
+	})
+	require.NoError(t, err)
+	recordB, err := tests.CreateRecord(hub, "network_probe_stats", map[string]any{
+		"system": system.Id,
+		"type":   "1m",
+		"stats":  `{"icmp:1.1.1.1":[40,100,9,50,5]}`,
+	})
+	require.NoError(t, err)
+
+	result := rm.AverageProbeStats(hub.DB(), records.RecordIds{
+		{Id: recordA.Id},
+		{Id: recordB.Id},
+	})
+
+	stats, ok := result["icmp:1.1.1.1"]
+	require.True(t, ok)
+	require.Len(t, stats, 5)
+	assert.Equal(t, 25.0, stats[0])
+	assert.Equal(t, 90.0, stats[1])
+	assert.Equal(t, 8.0, stats[2])
+	assert.Equal(t, 50.0, stats[3])
+	assert.Equal(t, 3.0, stats[4])
+}

internal/records/records.go

@@ -8,6 +8,7 @@ import (
 	"time"
 
 	"github.com/henrygd/beszel/internal/entities/container"
+	"github.com/henrygd/beszel/internal/entities/probe"
 	"github.com/henrygd/beszel/internal/entities/system"
 
 	"github.com/pocketbase/dbx"
@@ -70,7 +71,7 @@ func (rm *RecordManager) CreateLongerRecords() {
 	// wrap the operations in a transaction
 	rm.app.RunInTransaction(func(txApp core.App) error {
 		var err error
-		collections := [2]*core.Collection{}
+		collections := [3]*core.Collection{}
 		collections[0], err = txApp.FindCachedCollectionByNameOrId("system_stats")
 		if err != nil {
 			return err
@@ -79,6 +80,10 @@ func (rm *RecordManager) CreateLongerRecords() {
 		if err != nil {
 			return err
 		}
+		collections[2], err = txApp.FindCachedCollectionByNameOrId("network_probe_stats")
+		if err != nil {
+			return err
+		}
 		var systems RecordIds
 		db := txApp.DB()
 
@@ -138,8 +143,9 @@ func (rm *RecordManager) CreateLongerRecords() {
 					case "system_stats":
 						longerRecord.Set("stats", rm.AverageSystemStats(db, recordIds))
 					case "container_stats":
-
 						longerRecord.Set("stats", rm.AverageContainerStats(db, recordIds))
+					case "network_probe_stats":
+						longerRecord.Set("stats", rm.AverageProbeStats(db, recordIds))
 					}
 					if err := txApp.SaveNoValidate(longerRecord); err != nil {
 						log.Println("failed to save longer record", "err", err)
@@ -500,6 +506,80 @@ func AverageContainerStatsSlice(records [][]container.Stats) []container.Stats {
 	return result
 }
 
+// AverageProbeStats averages probe stats across multiple records.
+// For each probe key: avg of average fields, min of mins, and max of maxes.
+func (rm *RecordManager) AverageProbeStats(db dbx.Builder, records RecordIds) map[string]probe.Result {
+	type probeValues struct {
+		sums   probe.Result
+		counts []int
+	}
+
+	query := db.NewQuery("SELECT stats FROM network_probe_stats WHERE id = {:id}")
+
+	// accumulate sums for each probe key across records
+	sums := make(map[string]*probeValues)
+	var row StatsRecord
+	for _, rec := range records {
+		row.Stats = row.Stats[:0]
+		query.Bind(dbx.Params{"id": rec.Id}).One(&row)
+		var rawStats map[string]probe.Result
+		if err := json.Unmarshal(row.Stats, &rawStats); err != nil {
+			continue
+		}
+		for key, vals := range rawStats {
+			s, ok := sums[key]
+			if !ok {
+				s = &probeValues{sums: make(probe.Result, len(vals)), counts: make([]int, len(vals))}
+				sums[key] = s
+			}
+			if len(vals) > len(s.sums) {
+				expandedSums := make(probe.Result, len(vals))
+				copy(expandedSums, s.sums)
+				s.sums = expandedSums
+
+				expandedCounts := make([]int, len(vals))
+				copy(expandedCounts, s.counts)
+				s.counts = expandedCounts
+			}
+			for i := range vals {
+				switch i {
+				case 2: // min fields
+					if s.counts[i] == 0 || vals[i] < s.sums[i] {
+						s.sums[i] = vals[i]
+					}
+				case 3: // max fields
+					if s.counts[i] == 0 || vals[i] > s.sums[i] {
+						s.sums[i] = vals[i]
+					}
+				default: // average fields
+					s.sums[i] += vals[i]
+				}
+				s.counts[i]++
+			}
+		}
+	}
+
+	// compute final averages
+	result := make(map[string]probe.Result, len(sums))
+	for key, s := range sums {
+		if len(s.counts) == 0 {
+			continue
+		}
+		for i := range s.sums {
+			switch i {
+			case 2, 3: // min and max fields should not be averaged
+				continue
+			default:
+				if s.counts[i] > 0 {
+					s.sums[i] = twoDecimals(s.sums[i] / float64(s.counts[i]))
+				}
+			}
+		}
+		result[key] = s.sums
+	}
+	return result
+}
+
 /* Round float to two decimals */
 func twoDecimals(value float64) float64 {
 	return math.Round(value*100) / 100

internal/records/records_deletion.go

@@ -59,7 +59,7 @@ func deleteOldAlertsHistory(app core.App, countToKeep, countBeforeDeletion int)
 // Deletes system_stats records older than what is displayed in the UI
 func deleteOldSystemStats(app core.App) error {
 	// Collections to process
-	collections := [2]string{"system_stats", "container_stats"}
+	collections := [3]string{"system_stats", "container_stats", "network_probe_stats"}
 
 	// Record types and their retention periods
 	type RecordDeletionData struct {

internal/site/src/components/charts/area-chart.tsx

@@ -146,7 +146,7 @@ export default function AreaChartDefault({
 							axisLine={false}
 						/>
 					)}
-					{xAxis(chartData)}
+					{xAxis(chartData.chartTime, displayData.at(-1)?.created as number)}
 					<ChartTooltip
 						animationEasing="ease-out"
 						animationDuration={150}

internal/site/src/components/charts/line-chart.tsx

@@ -41,6 +41,7 @@ export default function LineChartDefault({
 	filter,
 	truncate = false,
 	chartProps,
+	connectNulls,
 }: {
 	chartData: ChartData
 	// biome-ignore lint/suspicious/noExplicitAny: accepts different data source types (systemStats or containerData)
@@ -62,6 +63,7 @@ export default function LineChartDefault({
 	filter?: string
 	truncate?: boolean
 	chartProps?: Omit<React.ComponentProps<typeof LineChart>, "data" | "margin">
+	connectNulls?: boolean
 }) {
 	const { yAxisWidth, updateYAxisWidth } = useYAxisWidth()
 	const { isIntersecting, ref } = useIntersectionObserver({ freeze: false })
@@ -104,7 +106,8 @@ export default function LineChartDefault({
 					isAnimationActive={false}
 					// stackId={dataPoint.stackId}
 					order={dataPoint.order || i}
-					// activeDot={dataPoint.activeDot ?? true}
+					activeDot={dataPoint.activeDot ?? true}
+					connectNulls={connectNulls}
 				/>
 			)
 		})
@@ -145,7 +148,7 @@ export default function LineChartDefault({
 							axisLine={false}
 						/>
 					)}
-					{xAxis(chartData)}
+					{xAxis(chartData.chartTime, displayData.at(-1)?.created as number)}
 					<ChartTooltip
 						animationEasing="ease-out"
 						animationDuration={150}

internal/site/src/components/navbar.tsx

@@ -8,6 +8,7 @@ import {
 	LogOutIcon,
 	LogsIcon,
 	MenuIcon,
+	NetworkIcon,
 	PlusIcon,
 	SearchIcon,
 	ServerIcon,
@@ -109,6 +110,10 @@ export default function Navbar() {
 								<HardDriveIcon className="h-4 w-4 me-2.5" strokeWidth={1.5} />
 								<span>S.M.A.R.T.</span>
 							</DropdownMenuItem>
+							<DropdownMenuItem onClick={() => navigate(getPagePath($router, "probes"))} className="flex items-center">
+								<NetworkIcon className="h-4 w-4 me-2.5" strokeWidth={1.5} />
+								<Trans>Network Probes</Trans>
+							</DropdownMenuItem>
 							<DropdownMenuItem
 								onClick={() => navigate(getPagePath($router, "settings", { name: "general" }))}
 								className="flex items-center"
@@ -180,6 +185,21 @@ export default function Navbar() {
 					</TooltipTrigger>
 					<TooltipContent>S.M.A.R.T.</TooltipContent>
 				</Tooltip>
+				<Tooltip>
+					<TooltipTrigger asChild>
+						<Link
+							href={getPagePath($router, "probes")}
+							className={cn("hidden md:grid", buttonVariants({ variant: "ghost", size: "icon" }))}
+							aria-label="Network Probes"
+							onMouseEnter={() => import("@/components/routes/probes")}
+						>
+							<NetworkIcon className="h-[1.2rem] w-[1.2rem]" strokeWidth={1.5} />
+						</Link>
+					</TooltipTrigger>
+					<TooltipContent>
+						<Trans>Network Probes</Trans>
+					</TooltipContent>
+				</Tooltip>
 				<LangToggle />
 				<ModeToggle />
 				<Tooltip>

internal/site/src/components/network-probes-table/network-probes-columns.tsx

@@ -0,0 +1,237 @@
+import type { CellContext, Column, ColumnDef } from "@tanstack/react-table"
+import { Button } from "@/components/ui/button"
+import { cn, decimalString, hourWithSeconds } from "@/lib/utils"
+import {
+	GlobeIcon,
+	TimerIcon,
+	WifiOffIcon,
+	Trash2Icon,
+	ArrowLeftRightIcon,
+	MoreHorizontalIcon,
+	ServerIcon,
+	ClockIcon,
+	NetworkIcon,
+	RefreshCwIcon,
+} from "lucide-react"
+import { t } from "@lingui/core/macro"
+import type { NetworkProbeRecord } from "@/types"
+import { DropdownMenu, DropdownMenuContent, DropdownMenuItem, DropdownMenuTrigger } from "@/components/ui/dropdown-menu"
+import { Trans } from "@lingui/react/macro"
+import { pb } from "@/lib/api"
+import { toast } from "../ui/use-toast"
+import { $allSystemsById } from "@/lib/stores"
+import { useStore } from "@nanostores/react"
+
+const protocolColors: Record<string, string> = {
+	icmp: "bg-blue-500/15 text-blue-400",
+	tcp: "bg-purple-500/15 text-purple-400",
+	http: "bg-green-500/15 text-green-400",
+}
+
+async function deleteProbe(id: string) {
+	try {
+		await pb.collection("network_probes").delete(id)
+	} catch (err: unknown) {
+		toast({ variant: "destructive", title: t`Error`, description: (err as Error)?.message })
+	}
+}
+
+export function getProbeColumns(longestName = 0, longestTarget = 0): ColumnDef<NetworkProbeRecord>[] {
+	return [
+		{
+			id: "name",
+			sortingFn: (a, b) => (a.original.name || a.original.target).localeCompare(b.original.name || b.original.target),
+			accessorFn: (record) => record.name || record.target,
+			header: ({ column }) => <HeaderButton column={column} name={t`Name`} Icon={NetworkIcon} />,
+			cell: ({ getValue }) => (
+				<div className="ms-1.5 max-w-40 block truncate tabular-nums" style={{ width: `${longestName / 1.05}ch` }}>
+					{getValue() as string}
+				</div>
+			),
+		},
+		{
+			id: "system",
+			accessorFn: (record) => record.system,
+			sortingFn: (a, b) => {
+				const allSystems = $allSystemsById.get()
+				const systemNameA = allSystems[a.original.system]?.name ?? ""
+				const systemNameB = allSystems[b.original.system]?.name ?? ""
+				return systemNameA.localeCompare(systemNameB)
+			},
+			header: ({ column }) => <HeaderButton column={column} name={t`System`} Icon={ServerIcon} />,
+			cell: ({ getValue }) => {
+				const allSystems = useStore($allSystemsById)
+				return <span className="ms-1.5 xl:w-20 block truncate">{allSystems[getValue() as string]?.name ?? ""}</span>
+			},
+		},
+		{
+			id: "target",
+			sortingFn: (a, b) => a.original.target.localeCompare(b.original.target),
+			accessorFn: (record) => record.target,
+			header: ({ column }) => <HeaderButton column={column} name={t`Target`} Icon={GlobeIcon} />,
+			cell: ({ getValue }) => (
+				<div className="ms-1.5 tabular-nums block truncate max-w-44" style={{ width: `${longestTarget / 1.05}ch` }}>
+					{getValue() as string}
+				</div>
+			),
+		},
+		{
+			id: "protocol",
+			accessorFn: (record) => record.protocol,
+			header: ({ column }) => <HeaderButton column={column} name={t`Protocol`} Icon={ArrowLeftRightIcon} />,
+			cell: ({ getValue }) => {
+				const protocol = getValue() as string
+				return (
+					<span className={cn("ms-1.5 px-2 py-0.5 rounded text-xs font-medium uppercase", protocolColors[protocol])}>
+						{protocol}
+					</span>
+				)
+			},
+		},
+		{
+			id: "interval",
+			accessorFn: (record) => record.interval,
+			header: ({ column }) => <HeaderButton column={column} name={t`Interval`} Icon={RefreshCwIcon} />,
+			cell: ({ getValue }) => <span className="ms-1.5 tabular-nums">{getValue() as number}s</span>,
+		},
+		{
+			id: "res",
+			accessorFn: (record) => record.res,
+			invertSorting: true,
+			header: ({ column }) => <HeaderButton column={column} name={t`Response`} Icon={TimerIcon} />,
+			cell: responseTimeCell,
+		},
+		{
+			id: "res1h",
+			accessorFn: (record) => record.resAvg1h,
+			invertSorting: true,
+			header: ({ column }) => <HeaderButton column={column} name={t`Avg 1h`} Icon={TimerIcon} />,
+			cell: responseTimeCell,
+		},
+		{
+			id: "max1h",
+			accessorFn: (record) => record.resMax1h,
+			invertSorting: true,
+			header: ({ column }) => <HeaderButton column={column} name={t`Max 1h`} Icon={TimerIcon} />,
+			cell: responseTimeCell,
+		},
+		{
+			id: "min1h",
+			accessorFn: (record) => record.resMin1h,
+			invertSorting: true,
+			header: ({ column }) => <HeaderButton column={column} name={t`Min 1h`} Icon={TimerIcon} />,
+			cell: responseTimeCell,
+		},
+		{
+			id: "loss",
+			accessorFn: (record) => record.loss1h,
+			invertSorting: true,
+			header: ({ column }) => <HeaderButton column={column} name={t`Loss 1h`} Icon={WifiOffIcon} />,
+			cell: ({ row }) => {
+				const { loss1h, res } = row.original
+				if (loss1h === undefined || (!res && !loss1h)) {
+					return <span className="ms-1.5 text-muted-foreground">-</span>
+				}
+				let color = "bg-green-500"
+				if (loss1h) {
+					color = loss1h > 20 ? "bg-red-500" : "bg-yellow-500"
+				}
+				return (
+					<span className="ms-1.5 tabular-nums flex gap-2 items-center">
+						<span className={cn("shrink-0 size-2 rounded-full", color)} />
+						{loss1h}%
+					</span>
+				)
+			},
+		},
+		{
+			id: "updated",
+			invertSorting: true,
+			accessorFn: (record) => record.updated,
+			header: ({ column }) => <HeaderButton column={column} name={t`Updated`} Icon={ClockIcon} />,
+			cell: ({ getValue }) => {
+				const timestamp = getValue() as number
+				return <span className="ms-1.5 tabular-nums">{hourWithSeconds(new Date(timestamp).toISOString())}</span>
+			},
+		},
+		{
+			id: "actions",
+			enableSorting: false,
+			header: () => null,
+			size: 40,
+			cell: ({ row }) => (
+				<DropdownMenu>
+					<DropdownMenuTrigger asChild>
+						<Button
+							variant="ghost"
+							size="icon"
+							className="size-10"
+							onClick={(event) => event.stopPropagation()}
+							onMouseDown={(event) => event.stopPropagation()}
+						>
+							<span className="sr-only">
+								<Trans>Open menu</Trans>
+							</span>
+							<MoreHorizontalIcon className="w-5" />
+						</Button>
+					</DropdownMenuTrigger>
+					<DropdownMenuContent align="end" onClick={(event) => event.stopPropagation()}>
+						<DropdownMenuItem
+							onClick={(event) => {
+								event.stopPropagation()
+								deleteProbe(row.original.id)
+							}}
+						>
+							<Trash2Icon className="me-2.5 size-4" />
+							<Trans>Delete</Trans>
+						</DropdownMenuItem>
+					</DropdownMenuContent>
+				</DropdownMenu>
+			),
+		},
+	]
+}
+function responseTimeCell(cell: CellContext<NetworkProbeRecord, unknown>) {
+	const val = cell.getValue() as number | undefined
+	if (!val) {
+		return <span className="ms-1.5 text-muted-foreground">-</span>
+	}
+	let color = "bg-green-500"
+	if (val > 200) {
+		color = "bg-yellow-500"
+	}
+	if (val > 2000) {
+		color = "bg-red-500"
+	}
+	return (
+		<span className="ms-1.5 tabular-nums flex gap-2 items-center">
+			<span className={cn("shrink-0 size-2 rounded-full", color)} />
+			{decimalString(val, val < 100 ? 2 : 1).toLocaleString()}ms
+		</span>
+	)
+}
+
+function HeaderButton({
+	column,
+	name,
+	Icon,
+}: {
+	column: Column<NetworkProbeRecord>
+	name: string
+	Icon: React.ElementType
+}) {
+	const isSorted = column.getIsSorted()
+	return (
+		<Button
+			className={cn(
+				"h-9 px-3 flex items-center gap-2 duration-50",
+				isSorted && "bg-accent/70 light:bg-accent text-accent-foreground/90"
+			)}
+			variant="ghost"
+			onClick={() => column.toggleSorting(column.getIsSorted() === "asc")}
+		>
+			{Icon && <Icon className="size-4" />}
+			{name}
+		</Button>
+	)
+}

internal/site/src/components/network-probes-table/network-probes-table.tsx

@@ -0,0 +1,225 @@
+import { t } from "@lingui/core/macro"
+import { Trans } from "@lingui/react/macro"
+import {
+	type ColumnFiltersState,
+	flexRender,
+	getCoreRowModel,
+	getFilteredRowModel,
+	getSortedRowModel,
+	type Row,
+	type SortingState,
+	type Table as TableType,
+	useReactTable,
+	type VisibilityState,
+} from "@tanstack/react-table"
+import { useVirtualizer, type VirtualItem } from "@tanstack/react-virtual"
+import { memo, useMemo, useRef, useState } from "react"
+import { getProbeColumns } from "@/components/network-probes-table/network-probes-columns"
+import { Card, CardHeader, CardTitle } from "@/components/ui/card"
+import { Input } from "@/components/ui/input"
+import { TableBody, TableCell, TableHead, TableHeader, TableRow } from "@/components/ui/table"
+import { isReadOnlyUser } from "@/lib/api"
+import { $allSystemsById } from "@/lib/stores"
+import { cn, getVisualStringWidth, useBrowserStorage } from "@/lib/utils"
+import type { NetworkProbeRecord } from "@/types"
+import { AddProbeDialog } from "./probe-dialog"
+
+export default function NetworkProbesTableNew({
+	systemId,
+	probes,
+}: {
+	systemId?: string
+	probes: NetworkProbeRecord[]
+}) {
+	const [sorting, setSorting] = useBrowserStorage<SortingState>(
+		`sort-np-${systemId ? 1 : 0}`,
+		[{ id: systemId ? "name" : "system", desc: false }],
+		sessionStorage
+	)
+	const [columnFilters, setColumnFilters] = useState<ColumnFiltersState>([])
+	const [columnVisibility, setColumnVisibility] = useState<VisibilityState>({})
+	const [globalFilter, setGlobalFilter] = useState("")
+
+	const { longestName, longestTarget } = useMemo(() => {
+		let longestName = 0
+		let longestTarget = 0
+		for (const p of probes) {
+			longestName = Math.max(longestName, getVisualStringWidth(p.name || p.target))
+			longestTarget = Math.max(longestTarget, getVisualStringWidth(p.target))
+		}
+		return { longestName, longestTarget }
+	}, [probes])
+
+	// Filter columns based on whether systemId is provided
+	const columns = useMemo(() => {
+		let columns = getProbeColumns(longestName, longestTarget)
+		columns = systemId ? columns.filter((col) => col.id !== "system") : columns
+		columns = isReadOnlyUser() ? columns.filter((col) => col.id !== "actions") : columns
+		return columns
+	}, [systemId, longestName, longestTarget])
+
+	const table = useReactTable({
+		data: probes,
+		columns,
+		getCoreRowModel: getCoreRowModel(),
+		getSortedRowModel: getSortedRowModel(),
+		getFilteredRowModel: getFilteredRowModel(),
+		onSortingChange: setSorting,
+		onColumnFiltersChange: setColumnFilters,
+		onColumnVisibilityChange: setColumnVisibility,
+		defaultColumn: {
+			sortUndefined: "last",
+			size: 900,
+			minSize: 0,
+		},
+		state: {
+			sorting,
+			columnFilters,
+			columnVisibility,
+			globalFilter,
+		},
+		onGlobalFilterChange: setGlobalFilter,
+		globalFilterFn: (row, _columnId, filterValue) => {
+			const probe = row.original
+			const systemName = $allSystemsById.get()[probe.system]?.name ?? ""
+			const searchString = `${probe.name}${probe.target}${probe.protocol}${systemName}`.toLocaleLowerCase()
+			return (filterValue as string)
+				.toLowerCase()
+				.split(" ")
+				.every((term) => searchString.includes(term))
+		},
+	})
+
+	const rows = table.getRowModel().rows
+	const visibleColumns = table.getVisibleLeafColumns()
+
+	return (
+		<Card className="@container w-full px-3 py-5 sm:py-6 sm:px-6">
+			<CardHeader className="p-0 mb-3 sm:mb-4">
+				<div className="grid md:flex gap-x-5 gap-y-3 w-full items-end">
+					<div className="px-2 sm:px-1">
+						<CardTitle className="mb-2">
+							<Trans>Network Probes</Trans>
+						</CardTitle>
+						<div className="text-sm text-muted-foreground flex items-center flex-wrap">
+							<Trans>Response time monitoring from agents.</Trans>
+						</div>
+					</div>
+					<div className="md:ms-auto flex items-center gap-2">
+						{probes.length > 0 && (
+							<Input
+								placeholder={t`Filter...`}
+								value={globalFilter}
+								onChange={(e) => setGlobalFilter(e.target.value)}
+								className="ms-auto px-4 w-full max-w-full md:w-64"
+							/>
+						)}
+						{!isReadOnlyUser() ? <AddProbeDialog systemId={systemId} /> : null}
+					</div>
+				</div>
+			</CardHeader>
+			<div className="rounded-md">
+				<NetworkProbesTable table={table} rows={rows} colLength={visibleColumns.length} />
+			</div>
+		</Card>
+	)
+}
+
+const NetworkProbesTable = memo(function NetworkProbeTable({
+	table,
+	rows,
+	colLength,
+}: {
+	table: TableType<NetworkProbeRecord>
+	rows: Row<NetworkProbeRecord>[]
+	colLength: number
+}) {
+	// The virtualizer will need a reference to the scrollable container element
+	const scrollRef = useRef<HTMLDivElement>(null)
+
+	const virtualizer = useVirtualizer<HTMLDivElement, HTMLTableRowElement>({
+		count: rows.length,
+		estimateSize: () => 54,
+		getScrollElement: () => scrollRef.current,
+		overscan: 5,
+	})
+	const virtualRows = virtualizer.getVirtualItems()
+
+	const paddingTop = Math.max(0, virtualRows[0]?.start ?? 0 - virtualizer.options.scrollMargin)
+	const paddingBottom = Math.max(0, virtualizer.getTotalSize() - (virtualRows[virtualRows.length - 1]?.end ?? 0))
+
+	return (
+		<div
+			className={cn(
+				"h-min max-h-[calc(100dvh-17rem)] max-w-full relative overflow-auto border rounded-md",
+				// don't set min height if there are less than 2 rows, do set if we need to display the empty state
+				(!rows.length || rows.length > 2) && "min-h-50"
+			)}
+			ref={scrollRef}
+		>
+			{/* add header height to table size */}
+			<div style={{ height: `${virtualizer.getTotalSize() + 48}px`, paddingTop, paddingBottom }}>
+				<table className="text-sm w-full h-full text-nowrap">
+					<NetworkProbeTableHead table={table} />
+					<TableBody>
+						{rows.length ? (
+							virtualRows.map((virtualRow) => {
+								const row = rows[virtualRow.index]
+								return <NetworkProbeTableRow key={row.id} row={row} virtualRow={virtualRow} />
+							})
+						) : (
+							<TableRow>
+								<TableCell colSpan={colLength} className="h-37 text-center pointer-events-none">
+									<Trans>No results.</Trans>
+								</TableCell>
+							</TableRow>
+						)}
+					</TableBody>
+				</table>
+			</div>
+		</div>
+	)
+})
+
+function NetworkProbeTableHead({ table }: { table: TableType<NetworkProbeRecord> }) {
+	return (
+		<TableHeader className="sticky top-0 z-50 w-full border-b-2">
+			{table.getHeaderGroups().map((headerGroup) => (
+				<tr key={headerGroup.id}>
+					{headerGroup.headers.map((header) => {
+						return (
+							<TableHead className="px-2" key={header.id}>
+								{header.isPlaceholder ? null : flexRender(header.column.columnDef.header, header.getContext())}
+							</TableHead>
+						)
+					})}
+				</tr>
+			))}
+		</TableHeader>
+	)
+}
+
+const NetworkProbeTableRow = memo(function NetworkProbeTableRow({
+	row,
+	virtualRow,
+}: {
+	row: Row<NetworkProbeRecord>
+	virtualRow: VirtualItem
+}) {
+	return (
+		<TableRow data-state={row.getIsSelected() && "selected"} className="transition-opacity">
+			{row.getVisibleCells().map((cell) => (
+				<TableCell
+					key={cell.id}
+					className="py-0"
+					style={{
+						width: `${cell.column.getSize()}px`,
+						height: virtualRow.size,
+					}}
+				>
+					{flexRender(cell.column.columnDef.cell, cell.getContext())}
+				</TableCell>
+			))}
+		</TableRow>
+	)
+})

internal/site/src/components/network-probes-table/probe-dialog.tsx

@@ -0,0 +1,194 @@
+import { useState } from "react"
+import { Trans, useLingui } from "@lingui/react/macro"
+import { useStore } from "@nanostores/react"
+import { pb } from "@/lib/api"
+import {
+	Dialog,
+	DialogContent,
+	DialogDescription,
+	DialogFooter,
+	DialogHeader,
+	DialogTitle,
+	DialogTrigger,
+} from "@/components/ui/dialog"
+import { Button } from "@/components/ui/button"
+import { Input } from "@/components/ui/input"
+import { Label } from "@/components/ui/label"
+import { Select, SelectContent, SelectItem, SelectTrigger, SelectValue } from "@/components/ui/select"
+import { PlusIcon } from "lucide-react"
+import { useToast } from "@/components/ui/use-toast"
+import { $systems } from "@/lib/stores"
+import * as v from "valibot"
+
+const Schema = v.object({
+	system: v.string(),
+	target: v.string(),
+	protocol: v.picklist(["icmp", "tcp", "http"]),
+	port: v.number(),
+	interval: v.pipe(v.string(), v.toNumber(), v.minValue(1), v.maxValue(3600)),
+	enabled: v.boolean(),
+	name: v.optional(v.string()),
+})
+
+export function AddProbeDialog({ systemId }: { systemId?: string }) {
+	const [open, setOpen] = useState(false)
+	const [protocol, setProtocol] = useState<string>("icmp")
+	const [target, setTarget] = useState("")
+	const [port, setPort] = useState("")
+	const [probeInterval, setProbeInterval] = useState("30")
+	const [name, setName] = useState("")
+	const [loading, setLoading] = useState(false)
+	const [selectedSystemId, setSelectedSystemId] = useState("")
+	const systems = useStore($systems)
+	const { toast } = useToast()
+	const { t } = useLingui()
+	const targetName = target.replace(/^https?:\/\//, "")
+
+	const resetForm = () => {
+		setProtocol("icmp")
+		setTarget("")
+		setPort("")
+		setProbeInterval("30")
+		setName("")
+		setSelectedSystemId("")
+	}
+
+	const handleSubmit = async (e: React.FormEvent) => {
+		e.preventDefault()
+		setLoading(true)
+
+		try {
+			const payload = v.parse(Schema, {
+				system: systemId ?? selectedSystemId,
+				target,
+				protocol,
+				port: protocol === "tcp" ? Number(port) : 0,
+				interval: probeInterval,
+				enabled: true,
+			})
+			if (name && name !== target) {
+				payload.name = name
+			}
+			await pb.collection("network_probes").create(payload)
+			resetForm()
+			setOpen(false)
+		} catch (err: unknown) {
+			toast({ variant: "destructive", title: t`Error`, description: (err as Error)?.message })
+		} finally {
+			setLoading(false)
+		}
+	}
+
+	return (
+		<Dialog open={open} onOpenChange={setOpen}>
+			<DialogTrigger asChild>
+				<Button variant="outline">
+					<PlusIcon className="size-4 me-1" />
+					<Trans>Add {{ foo: t`Probe` }}</Trans>
+				</Button>
+			</DialogTrigger>
+			<DialogContent className="max-w-md">
+				<DialogHeader>
+					<DialogTitle>
+						<Trans>Add {{ foo: t`Network Probe` }}</Trans>
+					</DialogTitle>
+					<DialogDescription>
+						<Trans>Configure response monitoring from this agent.</Trans>
+					</DialogDescription>
+				</DialogHeader>
+				<form onSubmit={handleSubmit} className="grid gap-4 tabular-nums">
+					{!systemId && (
+						<div className="grid gap-2">
+							<Label>
+								<Trans>System</Trans>
+							</Label>
+							<Select value={selectedSystemId} onValueChange={setSelectedSystemId} required>
+								<SelectTrigger>
+									<SelectValue placeholder={t`Select a system`} />
+								</SelectTrigger>
+								<SelectContent>
+									{systems.map((sys) => (
+										<SelectItem key={sys.id} value={sys.id}>
+											{sys.name}
+										</SelectItem>
+									))}
+								</SelectContent>
+							</Select>
+						</div>
+					)}
+					<div className="grid gap-2">
+						<Label>
+							<Trans>Target</Trans>
+						</Label>
+						<Input
+							value={target}
+							onChange={(e) => setTarget(e.target.value)}
+							placeholder={protocol === "http" ? "https://example.com" : "1.1.1.1"}
+							required
+						/>
+					</div>
+					<div className="grid gap-2">
+						<Label>
+							<Trans>Protocol</Trans>
+						</Label>
+
+						<Select value={protocol} onValueChange={setProtocol}>
+							<SelectTrigger>
+								<SelectValue />
+							</SelectTrigger>
+							<SelectContent>
+								<SelectItem value="icmp">ICMP</SelectItem>
+								<SelectItem value="tcp">TCP</SelectItem>
+								<SelectItem value="http">HTTP</SelectItem>
+							</SelectContent>
+						</Select>
+					</div>
+					{protocol === "tcp" && (
+						<div className="grid gap-2">
+							<Label>
+								<Trans>Port</Trans>
+							</Label>
+							<Input
+								type="number"
+								value={port}
+								onChange={(e) => setPort(e.target.value)}
+								placeholder="443"
+								min={1}
+								max={65535}
+								required
+							/>
+						</div>
+					)}
+					<div className="grid gap-2">
+						<Label>
+							<Trans>Interval (seconds)</Trans>
+						</Label>
+						<Input
+							type="number"
+							value={probeInterval}
+							onChange={(e) => setProbeInterval(e.target.value)}
+							min={1}
+							max={3600}
+							required
+						/>
+					</div>
+					<div className="grid gap-2">
+						<Label>
+							<Trans>Name (optional)</Trans>
+						</Label>
+						<Input
+							value={name}
+							onChange={(e) => setName(e.target.value)}
+							placeholder={targetName || t`e.g. Cloudflare DNS`}
+						/>
+					</div>
+					<DialogFooter>
+						<Button type="submit" disabled={loading || (!systemId && !selectedSystemId)}>
+							{loading ? <Trans>Creating...</Trans> : <Trans>Add {{ foo: t`Probe` }}</Trans>}
+						</Button>
+					</DialogFooter>
+				</form>
+			</DialogContent>
+		</Dialog>
+	)
+}

internal/site/src/components/router.tsx

@@ -4,6 +4,7 @@ const routes = {
 	home: "/",
 	containers: "/containers",
 	smart: "/smart",
+	probes: "/probes",
 	system: `/system/:id`,
 	settings: `/settings/:name?`,
 	forgot_password: `/forgot-password`,

internal/site/src/components/routes/probes.tsx

@@ -0,0 +1,25 @@
+import { useLingui } from "@lingui/react/macro"
+import { memo, useEffect } from "react"
+import NetworkProbesTableNew from "@/components/network-probes-table/network-probes-table"
+import { ActiveAlerts } from "@/components/active-alerts"
+import { FooterRepoLink } from "@/components/footer-repo-link"
+import { useNetworkProbesData } from "@/lib/use-network-probes"
+
+export default memo(() => {
+	const { t } = useLingui()
+	const { probes } = useNetworkProbesData({})
+
+	useEffect(() => {
+		document.title = `${t`Network Probes`} / Beszel`
+	}, [t])
+
+	return (
+		<>
+			<div className="grid gap-4">
+				<ActiveAlerts />
+				<NetworkProbesTableNew probes={probes} />
+			</div>
+			<FooterRepoLink />
+		</>
+	)
+})

internal/site/src/components/routes/system.tsx

@@ -11,9 +11,9 @@ import { RootDiskCharts, ExtraFsCharts } from "./system/charts/disk-charts"
 import { BandwidthChart, ContainerNetworkChart } from "./system/charts/network-charts"
 import { TemperatureChart, BatteryChart } from "./system/charts/sensor-charts"
 import { GpuPowerChart, GpuDetailCharts } from "./system/charts/gpu-charts"
-import { LazyContainersTable, LazySmartTable, LazySystemdTable } from "./system/lazy-tables"
+import { LazyContainersTable, LazySmartTable, LazySystemdTable, LazyNetworkProbesTable } from "./system/lazy-tables"
 import { LoadAverageChart } from "./system/charts/load-average-chart"
-import { ContainerIcon, CpuIcon, HardDriveIcon, TerminalSquareIcon } from "lucide-react"
+import { ContainerIcon, CpuIcon, HardDriveIcon, NetworkIcon, TerminalSquareIcon } from "lucide-react"
 import { GpuIcon } from "../ui/icons"
 import SystemdTable from "../systemd-table/systemd-table"
 import ContainersTable from "../containers-table/containers-table"
@@ -65,7 +65,7 @@ export default memo(function SystemDetail({ id }: { id: string }) {
 	const hasGpu = hasGpuData || hasGpuPowerData
 
 	// keep tabsRef in sync for keyboard navigation
-	const tabs = ["core", "disk"]
+	const tabs = ["core", "network", "disk"]
 	if (hasGpu) tabs.push("gpu")
 	if (hasContainers) tabs.push("containers")
 	if (hasSystemd) tabs.push("services")
@@ -145,6 +145,8 @@ export default memo(function SystemDetail({ id }: { id: string }) {
 				{hasContainersTable && <LazyContainersTable systemId={system.id} />}
 
 				{hasSystemd && <LazySystemdTable systemId={system.id} />}
+
+				<LazyNetworkProbesTable systemId={system.id} systemData={systemData} />
 			</>
 		)
 	}
@@ -157,6 +159,10 @@ export default memo(function SystemDetail({ id }: { id: string }) {
 						<CpuIcon className="size-3.5" />
 						<Trans context="Core system metrics">Core</Trans>
 					</TabsTrigger>
+					<TabsTrigger value="network" className="w-full flex items-center gap-1.5">
+						<NetworkIcon className="size-3.5" />
+						<Trans>Network</Trans>
+					</TabsTrigger>
 					<TabsTrigger value="disk" className="w-full flex items-center gap-1.5">
 						<HardDriveIcon className="size-3.5" />
 						<Trans>Disk</Trans>
@@ -184,16 +190,26 @@ export default memo(function SystemDetail({ id }: { id: string }) {
 				<TabsContent value="core" forceMount className={activeTab === "core" ? "contents" : "hidden"}>
 					<div className="grid xl:grid-cols-2 gap-4">
 						<CpuChart {...coreProps} />
-						<MemoryChart {...coreProps} />
 						<LoadAverageChart chartData={chartData} grid={grid} dataEmpty={dataEmpty} />
-						<BandwidthChart {...coreProps} systemStats={systemStats} />
+						<MemoryChart {...coreProps} />
+						<SwapChart chartData={chartData} grid={grid} dataEmpty={dataEmpty} systemStats={systemStats} />
 						<TemperatureChart {...coreProps} setPageBottomExtraMargin={setPageBottomExtraMargin} />
 						<BatteryChart {...coreProps} />
-						<SwapChart chartData={chartData} grid={grid} dataEmpty={dataEmpty} systemStats={systemStats} />
 						{pageBottomExtraMargin > 0 && <div style={{ marginBottom: pageBottomExtraMargin }}></div>}
 					</div>
 				</TabsContent>
 
+				<TabsContent value="network" forceMount className={activeTab === "network" ? "contents" : "hidden"}>
+					{mountedTabs.has("network") && (
+						<>
+							<div className="grid xl:grid-cols-2 gap-4">
+								<BandwidthChart {...coreProps} systemStats={systemStats} />
+							</div>
+							<LazyNetworkProbesTable systemId={system.id} systemData={systemData} />
+						</>
+					)}
+				</TabsContent>
+
 				<TabsContent value="disk" forceMount className={activeTab === "disk" ? "contents" : "hidden"}>
 					{mountedTabs.has("disk") && (
 						<>

internal/site/src/components/routes/system/chart-data.ts

@@ -1,7 +1,6 @@
-import { timeTicks } from "d3-time"
 import { getPbTimestamp, pb } from "@/lib/api"
 import { chartTimeData } from "@/lib/utils"
-import type { ChartData, ChartTimes, ContainerStatsRecord, SystemStatsRecord } from "@/types"
+import type { ChartData, ChartTimes, ContainerStatsRecord, NetworkProbeStatsRecord, SystemStatsRecord } from "@/types"
 
 type ChartTimeData = {
 	time: number
@@ -17,27 +16,6 @@ export const cache = new Map<
 	ChartTimeData | SystemStatsRecord[] | ContainerStatsRecord[] | ChartData["containerData"]
 >()
 
-// create ticks and domain for charts
-export function getTimeData(chartTime: ChartTimes, lastCreated: number) {
-	const cached = cache.get("td") as ChartTimeData | undefined
-	if (cached && cached.chartTime === chartTime) {
-		if (!lastCreated || cached.time >= lastCreated) {
-			return cached.data
-		}
-	}
-
-	// const buffer = chartTime === "1m" ? 400 : 20_000
-	const now = new Date(Date.now())
-	const startTime = chartTimeData[chartTime].getOffset(now)
-	const ticks = timeTicks(startTime, now, chartTimeData[chartTime].ticks ?? 12).map((date) => date.getTime())
-	const data = {
-		ticks,
-		domain: [chartTimeData[chartTime].getOffset(now).getTime(), now.getTime()],
-	}
-	cache.set("td", { time: now.getTime(), data, chartTime })
-	return data
-}
-
 /** Append new records onto prev with gap detection. Converts string `created` values to ms timestamps in place.
  * Pass `maxLen` to cap the result length in one copy instead of slicing again after the call. */
 export function appendData<T extends { created: string | number | null }>(
@@ -66,12 +44,12 @@ export function appendData<T extends { created: string | number | null }>(
 	return result
 }
 
-export async function getStats<T extends SystemStatsRecord | ContainerStatsRecord>(
+export async function getStats<T extends SystemStatsRecord | ContainerStatsRecord | NetworkProbeStatsRecord>(
 	collection: string,
 	systemId: string,
-	chartTime: ChartTimes
+	chartTime: ChartTimes,
+	cachedStats?: { created: string | number | null }[]
 ): Promise<T[]> {
-	const cachedStats = cache.get(`${systemId}_${chartTime}_${collection}`) as T[] | undefined
 	const lastCached = cachedStats?.at(-1)?.created as number
 	return await pb.collection<T>(collection).getFullList({
 		filter: pb.filter("system={:id} && created > {:created} && type={:type}", {

internal/site/src/components/routes/system/charts/probes-charts.tsx

@@ -0,0 +1,152 @@
+import LineChartDefault from "@/components/charts/line-chart"
+import type { DataPoint } from "@/components/charts/line-chart"
+import { toFixedFloat, decimalString } from "@/lib/utils"
+import { useLingui } from "@lingui/react/macro"
+import { ChartCard, FilterBar } from "../chart-card"
+import type { ChartData, NetworkProbeRecord, NetworkProbeStatsRecord } from "@/types"
+import { useMemo } from "react"
+import { atom } from "nanostores"
+import { useStore } from "@nanostores/react"
+
+const $filter = atom("")
+
+type ProbeChartProps = {
+	probeStats: NetworkProbeStatsRecord[]
+	grid?: boolean
+	probes: NetworkProbeRecord[]
+	chartData: ChartData
+	empty: boolean
+}
+
+type ProbeChartBaseProps = ProbeChartProps & {
+	valueIndex: number
+	title: string
+	description: string
+	tickFormatter: (value: number) => string
+	contentFormatter: ({ value }: { value: number | string }) => string | number
+	domain?: [number | "auto", number | "auto"]
+}
+
+function ProbeChart({
+	probeStats,
+	grid,
+	probes,
+	chartData,
+	empty,
+	valueIndex,
+	title,
+	description,
+	tickFormatter,
+	contentFormatter,
+	domain,
+}: ProbeChartBaseProps) {
+	const filter = useStore($filter)
+
+	const { dataPoints, visibleKeys } = useMemo(() => {
+		const sortedProbes = [...probes].sort((a, b) => b.resAvg1h - a.resAvg1h)
+		const count = sortedProbes.length
+		const points: DataPoint<NetworkProbeStatsRecord>[] = []
+		const visibleIDs: string[] = []
+		const filterTerms = filter
+			? filter
+					.toLowerCase()
+					.split(" ")
+					.filter((term) => term.length > 0)
+			: []
+		for (let i = 0; i < count; i++) {
+			const p = sortedProbes[i]
+			const label = p.name || p.target
+			const filtered = filterTerms.length > 0 && !filterTerms.some((term) => label.toLowerCase().includes(term))
+			if (filtered) {
+				continue
+			}
+			visibleIDs.push(p.id)
+			points.push({
+				order: i,
+				label,
+				dataKey: (record: NetworkProbeStatsRecord) => record.stats?.[p.id]?.[valueIndex] ?? "-",
+				color: count <= 5 ? i + 1 : `hsl(${(i * 360) / count}, var(--chart-saturation), var(--chart-lightness))`,
+			})
+		}
+		return { dataPoints: points, visibleKeys: visibleIDs }
+	}, [probes, filter, valueIndex])
+
+	const filteredProbeStats = useMemo(() => {
+		if (!visibleKeys.length) return probeStats
+		return probeStats.filter((record) => visibleKeys.some((id) => record.stats?.[id] != null))
+	}, [probeStats, visibleKeys])
+
+	const legend = dataPoints.length < 10
+
+	return (
+		<ChartCard
+			legend={legend}
+			cornerEl={<FilterBar store={$filter} />}
+			empty={empty}
+			title={title}
+			description={description}
+			grid={grid}
+		>
+			<LineChartDefault
+				chartData={chartData}
+				customData={filteredProbeStats}
+				dataPoints={dataPoints}
+				domain={domain ?? ["auto", "auto"]}
+				connectNulls
+				tickFormatter={tickFormatter}
+				contentFormatter={contentFormatter}
+				legend={legend}
+				filter={filter}
+			/>
+		</ChartCard>
+	)
+}
+
+export function ResponseChart({ probeStats, grid, probes, chartData, empty }: ProbeChartProps) {
+	const { t } = useLingui()
+
+	return (
+		<ProbeChart
+			probeStats={probeStats}
+			grid={grid}
+			probes={probes}
+			chartData={chartData}
+			empty={empty}
+			valueIndex={0}
+			title={t`Response`}
+			description={t`Average response time (ms)`}
+			tickFormatter={(value) => `${toFixedFloat(value, value >= 10 ? 0 : 1)} ms`}
+			contentFormatter={({ value }) => {
+				if (typeof value !== "number") {
+					return value
+				}
+				return `${decimalString(value, 2)} ms`
+			}}
+		/>
+	)
+}
+
+export function LossChart({ probeStats, grid, probes, chartData, empty }: ProbeChartProps) {
+	const { t } = useLingui()
+
+	return (
+		<ProbeChart
+			probeStats={probeStats}
+			grid={grid}
+			probes={probes}
+			chartData={chartData}
+			empty={empty}
+			valueIndex={4}
+			title={t`Loss`}
+			description={t`Packet loss (%)`}
+			domain={[0, 100]}
+			tickFormatter={(value) => `${toFixedFloat(value, value >= 10 ? 0 : 1)}%`}
+			contentFormatter={({ value }) => {
+				if (typeof value !== "number") {
+					return value
+				}
+				return `${decimalString(value, 2)}%`
+			}}
+		/>
+	)
+}

internal/site/src/components/routes/system/lazy-tables.tsx

@@ -1,6 +1,11 @@
 import { lazy } from "react"
 import { useIntersectionObserver } from "@/lib/use-intersection-observer"
 import { cn } from "@/lib/utils"
+import { ResponseChart, LossChart } from "./charts/probes-charts"
+import type { SystemData } from "./use-system-data"
+import { $chartTime } from "@/lib/stores"
+import { useStore } from "@nanostores/react"
+import { useNetworkProbesData } from "@/lib/use-network-probes"
 
 const ContainersTable = lazy(() => import("../../containers-table/containers-table"))
 
@@ -34,3 +39,46 @@ export function LazySystemdTable({ systemId }: { systemId: string }) {
 		</div>
 	)
 }
+
+const NetworkProbesTable = lazy(() => import("@/components/network-probes-table/network-probes-table"))
+
+export function LazyNetworkProbesTable({ systemId, systemData }: { systemId: string; systemData: SystemData }) {
+	const { isIntersecting, ref } = useIntersectionObserver()
+
+	return (
+		<div ref={ref} className={cn(isIntersecting && "contents")}>
+			{isIntersecting && <ProbesTable systemId={systemId} systemData={systemData} />}
+		</div>
+	)
+}
+
+function ProbesTable({ systemId, systemData }: { systemId: string; systemData: SystemData }) {
+	const { grid, chartData } = systemData ?? {}
+	const chartTime = useStore($chartTime)
+
+	const { probes, probeStats } = useNetworkProbesData({ systemId, loadStats: !!chartData, chartTime })
+
+	return (
+		<>
+			<NetworkProbesTable systemId={systemId} probes={probes} />
+			{!!chartData && !!probes.length && (
+				<div className="grid xl:grid-cols-2 gap-4">
+					<ResponseChart
+						probeStats={probeStats}
+						grid={grid}
+						probes={probes}
+						chartData={chartData}
+						empty={!probeStats.length}
+					/>
+					<LossChart
+						probeStats={probeStats}
+						grid={grid}
+						probes={probes}
+						chartData={chartData}
+						empty={!probeStats.length}
+					/>
+				</div>
+			)}
+		</>
+	)
+}

internal/site/src/components/routes/system/use-system-data.ts

@@ -26,7 +26,7 @@ import type {
 	SystemStatsRecord,
 } from "@/types"
 import { $router, navigate } from "../../router"
-import { appendData, cache, getStats, getTimeData, makeContainerData, makeContainerPoint } from "./chart-data"
+import { appendData, cache, getStats, makeContainerData, makeContainerPoint } from "./chart-data"
 
 export type SystemData = ReturnType<typeof useSystemData>
 
@@ -151,16 +151,11 @@ export function useSystemData(id: string) {
 	const agentVersion = useMemo(() => parseSemVer(system?.info?.v), [system?.info?.v])
 
 	const chartData: ChartData = useMemo(() => {
-		const lastCreated = Math.max(
-			(systemStats.at(-1)?.created as number) ?? 0,
-			(containerData.at(-1)?.created as number) ?? 0
-		)
 		return {
 			systemStats,
 			containerData,
 			chartTime,
 			orientation: direction === "rtl" ? "right" : "left",
-			...getTimeData(chartTime, lastCreated),
 			agentVersion,
 		}
 	}, [systemStats, containerData, direction])
@@ -200,8 +195,8 @@ export function useSystemData(id: string) {
 		}
 
 		Promise.allSettled([
-			getStats<SystemStatsRecord>("system_stats", systemId, chartTime),
-			getStats<ContainerStatsRecord>("container_stats", systemId, chartTime),
+			getStats<SystemStatsRecord>("system_stats", systemId, chartTime, cachedSystemStats),
+			getStats<ContainerStatsRecord>("container_stats", systemId, chartTime, cachedContainerData),
 		]).then(([systemStats, containerStats]) => {
 			// If another request has been made since this one, ignore the results
 			if (requestId !== statsRequestId.current) {

internal/site/src/components/ui/chart.tsx

@@ -3,9 +3,10 @@ import { useLingui } from "@lingui/react/macro"
 import * as React from "react"
 import * as RechartsPrimitive from "recharts"
 import { chartTimeData, cn } from "@/lib/utils"
-import type { ChartData } from "@/types"
+import type { ChartTimes } from "@/types"
 import { Separator } from "./separator"
 import { AxisDomain } from "recharts/types/util/types"
+import { timeTicks } from "d3-time"
 
 // Format: { THEME_NAME: CSS_SELECTOR }
 const THEMES = { light: "", dark: ".dark" } as const
@@ -400,26 +401,37 @@ function getPayloadConfigFromPayload(config: ChartConfig, payload: unknown, key:
 	return configLabelKey in config ? config[configLabelKey] : config[key as keyof typeof config]
 }
 
-let cachedAxis: JSX.Element
-const xAxis = ({ domain, ticks, chartTime }: ChartData) => {
-	if (cachedAxis && domain[0] === cachedAxis.props.domain[0]) {
-		return cachedAxis
+let cachedAxis: {
+	time: number
+	el: JSX.Element
+}
+
+const xAxis = (chartTime: ChartTimes, lastCreationTime: number) => {
+	if (Math.abs(lastCreationTime - cachedAxis?.time) < 1000) {
+		return cachedAxis.el
 	}
-	cachedAxis = (
-		<RechartsPrimitive.XAxis
-			dataKey="created"
-			domain={domain}
-			ticks={ticks}
-			allowDataOverflow
-			type="number"
-			scale="time"
-			minTickGap={12}
-			tickMargin={8}
-			axisLine={false}
-			tickFormatter={chartTimeData[chartTime].format}
-		/>
-	)
-	return cachedAxis
+	const now = new Date(lastCreationTime + 1000)
+	const startTime = chartTimeData[chartTime].getOffset(now)
+	const ticks = timeTicks(startTime, now, chartTimeData[chartTime].ticks ?? 12).map((date) => date.getTime())
+	const domain = [chartTimeData[chartTime].getOffset(now).getTime(), now.getTime()]
+	cachedAxis = {
+		time: lastCreationTime,
+		el: (
+			<RechartsPrimitive.XAxis
+				dataKey="created"
+				domain={domain}
+				ticks={ticks}
+				allowDataOverflow
+				type="number"
+				scale="time"
+				minTickGap={12}
+				tickMargin={8}
+				axisLine={false}
+				tickFormatter={chartTimeData[chartTime].format}
+			/>
+		),
+	}
+	return cachedAxis.el
 }
 
 export {

internal/site/src/lib/use-network-probes.ts

@@ -0,0 +1,314 @@
+import { chartTimeData } from "@/lib/utils"
+import type { ChartTimes, NetworkProbeRecord, NetworkProbeStatsRecord } from "@/types"
+import { useEffect, useRef, useState } from "react"
+import { getStats, appendData } from "@/components/routes/system/chart-data"
+import { pb } from "@/lib/api"
+import { toast } from "@/components/ui/use-toast"
+import type { RecordListOptions, RecordSubscription } from "pocketbase"
+
+const cache = new Map<string, NetworkProbeStatsRecord[]>()
+
+function getCacheValue(systemId: string, chartTime: ChartTimes | "rt") {
+	return cache.get(`${systemId}${chartTime}`) || []
+}
+
+function appendCacheValue(
+	systemId: string,
+	chartTime: ChartTimes | "rt",
+	newStats: NetworkProbeStatsRecord[],
+	maxPoints = 100
+) {
+	const cache_key = `${systemId}${chartTime}`
+	const existingStats = getCacheValue(systemId, chartTime)
+	if (existingStats) {
+		const { expectedInterval } = chartTimeData[chartTime]
+		const updatedStats = appendData(existingStats, newStats, expectedInterval, maxPoints)
+		cache.set(cache_key, updatedStats)
+		return updatedStats
+	} else {
+		cache.set(cache_key, newStats)
+		return newStats
+	}
+}
+
+const NETWORK_PROBE_FIELDS =
+	"id,name,system,target,protocol,port,interval,res,resMin1h,resMax1h,resAvg1h,loss1h,enabled,updated"
+
+interface UseNetworkProbesProps {
+	systemId?: string
+	loadStats?: boolean
+	chartTime?: ChartTimes
+	existingProbes?: NetworkProbeRecord[]
+}
+
+export function useNetworkProbesData(props: UseNetworkProbesProps) {
+	const { systemId, loadStats, chartTime, existingProbes } = props
+
+	const [p, setProbes] = useState<NetworkProbeRecord[]>([])
+	const [probeStats, setProbeStats] = useState<NetworkProbeStatsRecord[]>([])
+	const statsRequestId = useRef(0)
+	const pendingProbeEvents = useRef(new Map<string, RecordSubscription<NetworkProbeRecord>>())
+	const probeBatchTimeout = useRef<ReturnType<typeof setTimeout> | null>(null)
+
+	const probes = existingProbes ?? p
+
+	// clear old data when systemId changes
+	// useEffect(() => {
+	// 	return setProbes([])
+	// }, [systemId])
+
+	// initial load - fetch probes if not provided by caller
+	useEffect(() => {
+		if (!existingProbes) {
+			fetchProbes(systemId).then((probes) => setProbes(probes))
+		}
+	}, [systemId])
+
+	// Subscribe to updates if probes not provided by caller
+	useEffect(() => {
+		if (existingProbes) {
+			return
+		}
+		let unsubscribe: (() => void) | undefined
+
+		function flushPendingProbeEvents() {
+			probeBatchTimeout.current = null
+			if (!pendingProbeEvents.current.size) {
+				return
+			}
+			const events = pendingProbeEvents.current
+			pendingProbeEvents.current = new Map()
+			setProbes((currentProbes) => {
+				return applyProbeEvents(currentProbes ?? [], events.values(), systemId)
+			})
+		}
+
+		const pbOptions: RecordListOptions = { fields: NETWORK_PROBE_FIELDS }
+		if (systemId) {
+			pbOptions.filter = pb.filter("system = {:system}", { system: systemId })
+		}
+
+		;(async () => {
+			try {
+				unsubscribe = await pb.collection<NetworkProbeRecord>("network_probes").subscribe(
+					"*",
+					(event) => {
+						pendingProbeEvents.current.set(event.record.id, event)
+						if (!probeBatchTimeout.current) {
+							probeBatchTimeout.current = setTimeout(flushPendingProbeEvents, 50)
+						}
+					},
+					pbOptions
+				)
+			} catch (error) {
+				console.error("Failed to subscribe to probes", error)
+			}
+		})()
+
+		return () => {
+			if (probeBatchTimeout.current !== null) {
+				clearTimeout(probeBatchTimeout.current)
+				probeBatchTimeout.current = null
+			}
+			pendingProbeEvents.current.clear()
+			unsubscribe?.()
+		}
+	}, [systemId])
+
+	// Subscribe to new probe stats
+	useEffect(() => {
+		if (!loadStats || !systemId) {
+			return
+		}
+		let unsubscribe: (() => void) | undefined
+		const pbOptions = {
+			fields: "stats,created,type",
+			filter: pb.filter("system = {:system}", { system: systemId }),
+		}
+
+		;(async () => {
+			try {
+				unsubscribe = await pb.collection<NetworkProbeStatsRecord>("network_probe_stats").subscribe(
+					"*",
+					(event) => {
+						if (!chartTime || event.action !== "create") {
+							return
+						}
+						// if (typeof event.record.created === "string") {
+						// 	event.record.created = new Date(event.record.created).getTime()
+						// }
+						// return if not current chart time
+						// we could append to other chart times, but we would need to check the timestamps
+						// to make sure they fit in correctly, so for simplicity just ignore non-chart-time updates
+						// and fetch them via API when the user switches to that chart time
+						const chartTimeRecordType = chartTimeData[chartTime].type as ChartTimes
+						if (event.record.type !== chartTimeRecordType) {
+							// const lastCreated = getCacheValue(systemId, chartTime)?.at(-1)?.created ?? 0
+							// if (lastCreated) {
+							// 	// if the new record is close enough to the last cached record, append it to the cache so it's available immediately if the user switches to that chart time
+							// 	const { expectedInterval } = chartTimeData[chartTime]
+							// 	if (event.record.created - lastCreated < expectedInterval * 1.5) {
+							// 		console.log(
+							// 			`Caching out-of-chart-time probe stats record for chart time ${chartTime} (record type: ${event.record.type})`
+							// 		)
+							// 		const newStats = appendCacheValue(systemId, chartTime, [event.record])
+							// 		cache.set(`${systemId}${chartTime}`, newStats)
+							// 	}
+							// }
+							// console.log(`Received probe stats for non-current chart time (${event.record.type}), ignoring for now`)
+							return
+						}
+
+						// console.log("Appending new probe stats to chart:", event.record)
+						const newStats = appendCacheValue(systemId, chartTime, [event.record])
+						setProbeStats(newStats)
+					},
+					pbOptions
+				)
+			} catch (error) {
+				console.error("Failed to subscribe to probe stats:", error)
+			}
+		})()
+
+		return () => unsubscribe?.()
+	}, [systemId])
+
+	// fetch missing probe stats on load and when chart time changes
+	useEffect(() => {
+		if (!loadStats || !systemId || !chartTime || chartTime === "1m") {
+			return
+		}
+
+		const { expectedInterval } = chartTimeData[chartTime]
+		const requestId = ++statsRequestId.current
+
+		const cachedProbeStats = getCacheValue(systemId, chartTime)
+
+		// Render from cache immediately if available
+		if (cachedProbeStats.length) {
+			setProbeStats(cachedProbeStats)
+
+			// Skip the fetch if the latest cached point is recent enough that no new point is expected yet
+			const lastCreated = cachedProbeStats.at(-1)?.created
+			if (lastCreated && Date.now() - lastCreated < expectedInterval * 0.9) {
+				console.log("Using cached probe stats, skipping fetch")
+				return
+			}
+		}
+
+		getStats<NetworkProbeStatsRecord>("network_probe_stats", systemId, chartTime, cachedProbeStats).then(
+			(probeStats) => {
+				// If another request has been made since this one, ignore the results
+				if (requestId !== statsRequestId.current) {
+					return
+				}
+				const newStats = appendCacheValue(systemId, chartTime, probeStats)
+				setProbeStats(newStats)
+			}
+		)
+	}, [chartTime])
+
+	// subscribe to realtime metrics if chart time is 1m
+	useEffect(() => {
+		if (!loadStats || !systemId || chartTime !== "1m") {
+			return
+		}
+		let unsubscribe: (() => void) | undefined
+		const cache_key = `${systemId}rt`
+		pb.realtime
+			.subscribe(
+				`rt_metrics`,
+				(data: { Probes: NetworkProbeStatsRecord["stats"] }) => {
+					let prev = getCacheValue(systemId, "rt")
+					const now = Date.now()
+					// if no previous data or the last data point is older than 1min,
+					// create a new data set starting with a point 1 second ago to seed the chart data
+					if (!prev || (prev.at(-1)?.created ?? 0) < now - 60_000) {
+						prev = [{ created: now - 1000, stats: probesToStats(probes) }]
+					}
+					const stats = { created: now, stats: data.Probes } as NetworkProbeStatsRecord
+					const newStats = appendData(prev, [stats], 1000, 120)
+					setProbeStats(() => newStats)
+					cache.set(cache_key, newStats)
+				},
+				{ query: { system: systemId } }
+			)
+			.then((us) => {
+				unsubscribe = us
+			})
+		return () => unsubscribe?.()
+	}, [chartTime, systemId])
+
+	return {
+		probes,
+		probeStats,
+	}
+}
+
+function probesToStats(probes: NetworkProbeRecord[]): NetworkProbeStatsRecord["stats"] {
+	const stats: NetworkProbeStatsRecord["stats"] = {}
+	for (const probe of probes) {
+		stats[probe.id] = [probe.res, probe.resAvg1h, probe.resMin1h, probe.resMax1h, probe.loss1h]
+	}
+	return stats
+}
+
+async function fetchProbes(systemId?: string) {
+	try {
+		const res = await pb.collection<NetworkProbeRecord>("network_probes").getList(0, 2000, {
+			fields: NETWORK_PROBE_FIELDS,
+			filter: systemId ? pb.filter("system={:system}", { system: systemId }) : undefined,
+		})
+		return res.items
+	} catch (error) {
+		toast({
+			title: "Error",
+			description: (error as Error)?.message,
+			variant: "destructive",
+		})
+		return []
+	}
+}
+
+function applyProbeEvents(
+	probes: NetworkProbeRecord[],
+	events: Iterable<RecordSubscription<NetworkProbeRecord>>,
+	systemId?: string
+) {
+	// Use a map to handle updates/deletes in constant time
+	const probeById = new Map(probes.map((probe) => [probe.id, probe]))
+	const createdProbes: NetworkProbeRecord[] = []
+
+	for (const { action, record } of events) {
+		const matchesSystemScope = !systemId || record.system === systemId
+
+		if (action === "delete" || !matchesSystemScope) {
+			probeById.delete(record.id)
+			continue
+		}
+
+		if (!probeById.has(record.id)) {
+			createdProbes.push(record)
+		}
+
+		probeById.set(record.id, record)
+	}
+
+	const nextProbes: NetworkProbeRecord[] = []
+	// Prepend brand new probes (matching previous behavior)
+	for (let index = createdProbes.length - 1; index >= 0; index -= 1) {
+		nextProbes.push(createdProbes[index])
+	}
+
+	// Rebuild the final list while preserving original order for existing probes
+	for (const probe of probes) {
+		const nextProbe = probeById.get(probe.id)
+		if (!nextProbe) {
+			continue
+		}
+		nextProbes.push(nextProbe)
+		probeById.delete(probe.id)
+	}
+
+	return nextProbes
+}

internal/site/src/locales/it/it.po

@@ -8,7 +8,7 @@ msgstr ""
 "Language: it\n"
 "Project-Id-Version: beszel\n"
 "Report-Msgid-Bugs-To: \n"
-"PO-Revision-Date: 2026-04-17 09:26\n"
+"PO-Revision-Date: 2026-04-05 18:27\n"
 "Last-Translator: \n"
 "Language-Team: Italian\n"
 "Plural-Forms: nplurals=2; plural=(n != 1);\n"
@@ -57,7 +57,7 @@ msgstr "1 ora"
 #. Load average
 #: src/components/routes/system/charts/load-average-chart.tsx
 msgid "1 min"
-msgstr "1 min"
+msgstr ""
 
 #: src/lib/utils.ts
 msgid "1 minute"
@@ -74,7 +74,7 @@ msgstr "12 ore"
 #. Load average
 #: src/components/routes/system/charts/load-average-chart.tsx
 msgid "15 min"
-msgstr "15 min"
+msgstr ""
 
 #: src/lib/utils.ts
 msgid "24 hours"
@@ -87,7 +87,7 @@ msgstr "30 giorni"
 #. Load average
 #: src/components/routes/system/charts/load-average-chart.tsx
 msgid "5 min"
-msgstr "5 min"
+msgstr ""
 
 #. Table column
 #: src/components/routes/settings/quiet-hours.tsx
@@ -248,7 +248,7 @@ msgstr "Larghezza di banda"
 #. Battery label in systems table header
 #: src/components/systems-table/systems-table-columns.tsx
 msgid "Bat"
-msgstr "Batt"
+msgstr ""
 
 #: src/components/routes/system/charts/sensor-charts.tsx
 #: src/lib/alerts.ts
@@ -336,7 +336,7 @@ msgstr "Attenzione - possibile perdita di dati"
 
 #: src/components/routes/settings/general.tsx
 msgid "Celsius (°C)"
-msgstr "Celsius (°C)"
+msgstr ""
 
 #: src/components/routes/settings/general.tsx
 msgid "Change display units for metrics."
@@ -490,13 +490,13 @@ msgstr "Copia YAML"
 #: src/components/routes/system.tsx
 msgctxt "Core system metrics"
 msgid "Core"
-msgstr "Interne"
+msgstr ""
 
 #: src/components/containers-table/containers-table-columns.tsx
 #: src/components/systemd-table/systemd-table-columns.tsx
 #: src/components/systems-table/systems-table-columns.tsx
 msgid "CPU"
-msgstr "CPU"
+msgstr ""
 
 #: src/components/routes/system/cpu-sheet.tsx
 msgid "CPU Cores"
@@ -624,7 +624,7 @@ msgstr "Utilizzo del disco di {extraFsName}"
 #: src/components/routes/system/info-bar.tsx
 msgctxt "Layout display options"
 msgid "Display"
-msgstr "Display"
+msgstr ""
 
 #: src/components/routes/system/charts/cpu-charts.tsx
 msgid "Docker CPU Usage"
@@ -677,7 +677,7 @@ msgstr "Modifica {foo}"
 #: src/components/login/forgot-pass-form.tsx
 #: src/components/login/otp-forms.tsx
 msgid "Email"
-msgstr "Email"
+msgstr ""
 
 #: src/components/routes/settings/notifications.tsx
 msgid "Email notifications"
@@ -772,7 +772,7 @@ msgstr "Esporta la configurazione attuale dei tuoi sistemi."
 
 #: src/components/routes/settings/general.tsx
 msgid "Fahrenheit (°F)"
-msgstr "Fahrenheit (°F)"
+msgstr ""
 
 #: src/components/systems-table/systems-table-columns.tsx
 msgid "Failed"
@@ -824,7 +824,7 @@ msgstr "Impronta digitale"
 
 #: src/components/routes/system/smart-table.tsx
 msgid "Firmware"
-msgstr "Firmware"
+msgstr ""
 
 #: src/components/alerts/alerts-sheet.tsx
 msgid "For <0>{min}</0> {min, plural, one {minute} other {minutes}}"
@@ -858,7 +858,7 @@ msgstr "Globale"
 
 #: src/components/routes/system.tsx
 msgid "GPU"
-msgstr "GPU"
+msgstr ""
 
 #: src/components/routes/system/charts/gpu-charts.tsx
 msgid "GPU Engines"
@@ -883,7 +883,7 @@ msgstr "Stato"
 
 #: src/components/routes/settings/layout.tsx
 msgid "Heartbeat"
-msgstr "Hearthbeat"
+msgstr ""
 
 #: src/components/routes/settings/heartbeat.tsx
 msgid "Heartbeat Monitoring"
@@ -901,7 +901,7 @@ msgstr "Comando Homebrew"
 
 #: src/components/add-system.tsx
 msgid "Host / IP"
-msgstr "Host / IP"
+msgstr ""
 
 #: src/components/routes/settings/heartbeat.tsx
 msgid "HTTP Method"
@@ -1043,7 +1043,7 @@ msgstr "Istruzioni di configurazione manuale"
 #. Chart select field. Please try to keep this short.
 #: src/components/routes/system/chart-card.tsx
 msgid "Max 1 min"
-msgstr "Max 1 min"
+msgstr ""
 
 #: src/components/containers-table/containers-table-columns.tsx
 #: src/components/routes/system/info-bar.tsx
@@ -1109,7 +1109,7 @@ msgstr "Unità rete"
 #: src/components/systemd-table/systemd-table.tsx
 #: src/components/systemd-table/systemd-table.tsx
 msgid "No"
-msgstr "No"
+msgstr ""
 
 #: src/components/command-palette.tsx
 #: src/components/systemd-table/systemd-table.tsx
@@ -1196,7 +1196,7 @@ msgstr "Pagine / Impostazioni"
 #: src/components/login/auth-form.tsx
 #: src/components/login/auth-form.tsx
 msgid "Password"
-msgstr "Password"
+msgstr ""
 
 #: src/components/login/auth-form.tsx
 msgid "Password must be at least 8 characters."
@@ -1384,7 +1384,7 @@ msgstr "Riprendi"
 #: src/components/systems-table/systems-table-columns.tsx
 msgctxt "Root disk label"
 msgid "Root"
-msgstr "Root"
+msgstr ""
 
 #: src/components/routes/settings/tokens-fingerprints.tsx
 msgid "Rotate token"
@@ -1615,11 +1615,11 @@ msgstr "Temperature dei sensori di sistema"
 
 #: src/components/routes/settings/notifications.tsx
 msgid "Test <0>URL</0>"
-msgstr "Test <0>URL</0>"
+msgstr ""
 
 #: src/components/routes/settings/heartbeat.tsx
 msgid "Test heartbeat"
-msgstr "Test Heartbeat"
+msgstr ""
 
 #: src/components/routes/settings/notifications.tsx
 msgid "Test notification sent"
@@ -1665,7 +1665,7 @@ msgstr "Attiva/disattiva tema"
 #: src/components/add-system.tsx
 #: src/components/routes/settings/tokens-fingerprints.tsx
 msgid "Token"
-msgstr "Token"
+msgstr ""
 
 #: src/components/command-palette.tsx
 #: src/components/routes/settings/layout.tsx
@@ -1931,4 +1931,3 @@ msgstr "Sì"
 #: src/components/routes/settings/layout.tsx
 msgid "Your user settings have been updated."
 msgstr "Le impostazioni utente sono state aggiornate."
-

internal/site/src/locales/sr/sr.po

@@ -8,7 +8,7 @@ msgstr ""
 "Language: sr\n"
 "Project-Id-Version: beszel\n"
 "Report-Msgid-Bugs-To: \n"
-"PO-Revision-Date: 2026-04-22 14:14\n"
+"PO-Revision-Date: 2026-04-05 18:27\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"
@@ -858,7 +858,7 @@ msgstr "Глобално"
 
 #: src/components/routes/system.tsx
 msgid "GPU"
-msgstr "ГПЈ"
+msgstr ""
 
 #: src/components/routes/system/charts/gpu-charts.tsx
 msgid "GPU Engines"
@@ -1384,7 +1384,7 @@ msgstr "Настави"
 #: src/components/systems-table/systems-table-columns.tsx
 msgctxt "Root disk label"
 msgid "Root"
-msgstr "Root"
+msgstr ""
 
 #: src/components/routes/settings/tokens-fingerprints.tsx
 msgid "Rotate token"
@@ -1697,7 +1697,7 @@ msgstr "Укупни подаци poslati за сваки интерфејс"
 #: src/components/routes/system/disk-io-sheet.tsx
 msgctxt "Disk I/O"
 msgid "Total time spent on read/write (can exceed 100%)"
-msgstr "Укупно време проведено на читању/писању (може бити веће од 100%)"
+msgstr ""
 
 #. placeholder {0}: data.length
 #: src/components/systemd-table/systemd-table.tsx
@@ -1931,4 +1931,3 @@ msgstr "Да"
 #: src/components/routes/settings/layout.tsx
 msgid "Your user settings have been updated."
 msgstr "Ваша корисничка подешавања су ажурирана."
-

internal/site/src/main.tsx

@@ -30,6 +30,7 @@ const LoginPage = lazy(() => import("@/components/login/login.tsx"))
 const Home = lazy(() => import("@/components/routes/home.tsx"))
 const Containers = lazy(() => import("@/components/routes/containers.tsx"))
 const Smart = lazy(() => import("@/components/routes/smart.tsx"))
+const Probes = lazy(() => import("@/components/routes/probes.tsx"))
 const SystemDetail = lazy(() => import("@/components/routes/system.tsx"))
 const CopyToClipboardDialog = lazy(() => import("@/components/copy-to-clipboard.tsx"))
 
@@ -79,6 +80,8 @@ const App = memo(() => {
 		return <Containers />
 	} else if (page.route === "smart") {
 		return <Smart />
+	} else if (page.route === "probes") {
+		return <Probes />
 	} else if (page.route === "settings") {
 		return <Settings />
 	}

internal/site/src/types.d.ts

@@ -316,8 +316,6 @@ export interface ChartData {
 	systemStats: SystemStatsRecord[]
 	containerData: ChartDataContainer[]
 	orientation: "right" | "left"
-	ticks: number[]
-	domain: number[]
 	chartTime: ChartTimes
 }
 
@@ -546,3 +544,40 @@ export interface UpdateInfo {
 	v: string // new version
 	url: string // url to new version
 }
+
+export interface NetworkProbeRecord {
+	id: string
+	system: string
+	name: string
+	target: string
+	protocol: "icmp" | "tcp" | "http"
+	port: number
+	res: number
+	resMin1h: number
+	resMax1h: number
+	resAvg1h: number
+	loss1h: number
+	interval: number
+	enabled: boolean
+	updated: string
+}
+
+/**
+ * 0: avg 1 minute response in ms
+ *
+ * 1: avg response over 1 hour in ms
+ *
+ * 2: min response over the last hour in ms
+ *
+ * 3: max response over the last hour in ms
+ *
+ * 4: packet loss in %
+ */
+type ProbeResult = number[]
+
+export interface NetworkProbeStatsRecord {
+	id?: string
+	type?: string
+	stats: Record<string, ProbeResult>
+	created: number // unix timestamp (ms) for Recharts xAxis
+}