chore: update inactivity-actions (#1742)

.github/workflows/inactivity-actions.yml

@@ -6,6 +6,7 @@ on:
   workflow_dispatch:
 
 permissions:
+  actions: write
   issues: write
   pull-requests: write
 
@@ -48,6 +49,9 @@ jobs:
           # Action can not skip PRs, set it to 100 years to cover it.
           days-before-pr-stale: 36524
 
+          # Max issues to process before early exit. Next run resumes from cache. GH API limit: 5000.
+          operations-per-run: 1500
+
           # Labels
           stale-issue-label: 'stale'
           remove-stale-when-updated: true
@@ -56,4 +60,5 @@ jobs:
 
           # Exemptions
           exempt-assignees: true
-          exempt-milestones: true
+
+          exempt-milestones: true

Makefile

@@ -3,6 +3,40 @@ OS ?= $(shell go env GOOS)
 ARCH ?= $(shell go env GOARCH)
 # Skip building the web UI if true
 SKIP_WEB ?= false
+# Controls NVML/glibc agent build tag behavior:
+# - auto (default): enable on linux/amd64 glibc hosts
+# - true: always enable
+# - false: always disable
+NVML ?= auto
+
+# Detect glibc host for local linux/amd64 builds.
+HOST_GLIBC := $(shell \
+	if [ "$(OS)" = "linux" ] && [ "$(ARCH)" = "amd64" ]; then \
+		for p in /lib64/ld-linux-x86-64.so.2 /lib/x86_64-linux-gnu/ld-linux-x86-64.so.2 /lib/ld-linux-x86-64.so.2; do \
+			[ -e "$$p" ] && { echo true; exit 0; }; \
+		done; \
+		if command -v ldd >/dev/null 2>&1; then \
+			if ldd --version 2>&1 | tr '[:upper:]' '[:lower:]' | awk '/gnu libc|glibc/{found=1} END{exit !found}'; then \
+				echo true; \
+			else \
+				echo false; \
+			fi; \
+		else \
+			echo false; \
+		fi; \
+	else \
+		echo false; \
+	fi)
+
+# Enable glibc build tag for NVML on supported Linux builds.
+AGENT_GO_TAGS :=
+ifeq ($(NVML),true)
+AGENT_GO_TAGS := -tags glibc
+else ifeq ($(NVML),auto)
+ifeq ($(HOST_GLIBC),true)
+AGENT_GO_TAGS := -tags glibc
+endif
+endif
 
 # Set executable extension based on target OS
 EXE_EXT := $(if $(filter windows,$(OS)),.exe,)
@@ -54,7 +88,7 @@ fetch-smartctl-conditional:
 
 # Update build-agent to include conditional .NET build
 build-agent: tidy build-dotnet-conditional fetch-smartctl-conditional
-	GOOS=$(OS) GOARCH=$(ARCH) go build -o ./build/beszel-agent_$(OS)_$(ARCH)$(EXE_EXT) -ldflags "-w -s" ./internal/cmd/agent
+	GOOS=$(OS) GOARCH=$(ARCH) go build $(AGENT_GO_TAGS) -o ./build/beszel-agent_$(OS)_$(ARCH)$(EXE_EXT) -ldflags "-w -s" ./internal/cmd/agent
 
 build-hub: tidy $(if $(filter false,$(SKIP_WEB)),build-web-ui)
 	GOOS=$(OS) GOARCH=$(ARCH) go build -o ./build/beszel_$(OS)_$(ARCH)$(EXE_EXT) -ldflags "-w -s" ./internal/cmd/hub
@@ -90,9 +124,9 @@ dev-hub:
 
 dev-agent:
 	@if command -v entr >/dev/null 2>&1; then \
-		find ./internal/cmd/agent/*.go ./agent/*.go | entr -r go run github.com/henrygd/beszel/internal/cmd/agent; \
+		find ./internal/cmd/agent/*.go ./agent/*.go | entr -r go run $(AGENT_GO_TAGS) github.com/henrygd/beszel/internal/cmd/agent; \
 	else \
-		go run github.com/henrygd/beszel/internal/cmd/agent; \
+		go run $(AGENT_GO_TAGS) github.com/henrygd/beszel/internal/cmd/agent; \
 	fi
 	
 build-dotnet:

agent/docker.go

@@ -32,10 +32,6 @@ var ansiEscapePattern = regexp.MustCompile(`\x1b\[[0-9;]*[a-zA-Z]|\x1b\][^\x07]*
 const (
 	// Docker API timeout in milliseconds
 	dockerTimeoutMs = 2100
-	// Number of consecutive /containers/json failures before forcing a client reset on old Docker versions
-	dockerClientResetFailureThreshold = 3
-	// Minimum time between Docker client resets to avoid reset flapping
-	dockerClientResetCooldown = 30 * time.Second
 	// Maximum realistic network speed (5 GB/s) to detect bad deltas
 	maxNetworkSpeedBps uint64 = 5e9
 	// Maximum conceivable memory usage of a container (100TB) to detect bad memory stats
@@ -59,16 +55,12 @@ type dockerManager struct {
 	containerStatsMap   map[string]*container.Stats // Keeps track of container stats
 	validIds            map[string]struct{}         // Map of valid container ids, used to prune invalid containers from containerStatsMap
 	goodDockerVersion   bool                        // Whether docker version is at least 25.0.0 (one-shot works correctly)
-	versionChecked      bool                        // Whether docker version detection completed successfully
 	isWindows           bool                        // Whether the Docker Engine API is running on Windows
 	buf                 *bytes.Buffer               // Buffer to store and read response bodies
 	decoder             *json.Decoder               // Reusable JSON decoder that reads from buf
 	apiStats            *container.ApiStats         // Reusable API stats object
 	excludeContainers   []string                    // Patterns to exclude containers by name
 	usingPodman         bool                        // Whether the Docker Engine API is running on Podman
-	transport           *http.Transport             // Base transport used by client for connection resets
-	consecutiveListFailures int                     // Number of consecutive /containers/json request failures
-	lastClientReset     time.Time                   // Last time the Docker client connections were reset
 
 	// Cache-time-aware tracking for CPU stats (similar to cpu.go)
 	// Maps cache time intervals to container-specific CPU usage tracking
@@ -127,10 +119,8 @@ func (dm *dockerManager) shouldExcludeContainer(name string) bool {
 func (dm *dockerManager) getDockerStats(cacheTimeMs uint16) ([]*container.Stats, error) {
 	resp, err := dm.client.Get("http://localhost/containers/json")
 	if err != nil {
-		dm.handleContainerListError(err)
 		return nil, err
 	}
-	dm.consecutiveListFailures = 0
 
 	dm.apiContainerList = dm.apiContainerList[:0]
 	if err := dm.decode(resp, &dm.apiContainerList); err != nil {
@@ -214,50 +204,6 @@ func (dm *dockerManager) getDockerStats(cacheTimeMs uint16) ([]*container.Stats,
 	return stats, nil
 }
 
-func (dm *dockerManager) handleContainerListError(err error) {
-	dm.consecutiveListFailures++
-	if !dm.shouldResetDockerClient(err) {
-		return
-	}
-	dm.resetDockerClientConnections()
-}
-
-func (dm *dockerManager) shouldResetDockerClient(err error) bool {
-	if !dm.versionChecked || dm.goodDockerVersion {
-		return false
-	}
-	if dm.consecutiveListFailures < dockerClientResetFailureThreshold {
-		return false
-	}
-	if !dm.lastClientReset.IsZero() && time.Since(dm.lastClientReset) < dockerClientResetCooldown {
-		return false
-	}
-	return isDockerApiOverloadError(err)
-}
-
-func isDockerApiOverloadError(err error) bool {
-	if err == nil {
-		return false
-	}
-	if errors.Is(err, context.DeadlineExceeded) {
-		return true
-	}
-	msg := err.Error()
-	return strings.Contains(msg, "Client.Timeout exceeded") ||
-		strings.Contains(msg, "request canceled") ||
-		strings.Contains(msg, "context deadline exceeded") ||
-		strings.Contains(msg, "EOF")
-}
-
-func (dm *dockerManager) resetDockerClientConnections() {
-	if dm.transport == nil {
-		return
-	}
-	dm.transport.CloseIdleConnections()
-	dm.lastClientReset = time.Now()
-	slog.Warn("Reset Docker client connections after repeated /containers/json failures", "failures", dm.consecutiveListFailures)
-}
-
 // initializeCpuTracking initializes CPU tracking maps for a specific cache time interval
 func (dm *dockerManager) initializeCpuTracking(cacheTimeMs uint16) {
 	// Initialize cache time maps if they don't exist
@@ -607,7 +553,6 @@ func newDockerManager() *dockerManager {
 			Timeout:   timeout,
 			Transport: userAgentTransport,
 		},
-		transport:         transport,
 		containerStatsMap: make(map[string]*container.Stats),
 		sem:               make(chan struct{}, 5),
 		apiContainerList:  []*container.ApiInfo{},
@@ -666,7 +611,6 @@ func (dm *dockerManager) checkDockerVersion() {
 	if err := dm.decode(resp, &versionInfo); err != nil {
 		return
 	}
-	dm.versionChecked = true
 	// if version > 24, one-shot works correctly and we can limit concurrent operations
 	if dockerVersion, err := semver.Parse(versionInfo.Version); err == nil && dockerVersion.Major > 24 {
 		dm.goodDockerVersion = true