initial nvml collector

agent/gpu_nvml.go

@@ -2,6 +2,8 @@ package agent
 
 import (
 	"fmt"
+	"os"
+	"path/filepath"
 	"strings"
 	"time"
 	"unsafe"
@@ -73,9 +75,10 @@ type nvmlCollector struct {
 
 func (c *nvmlCollector) init() error {
 	slog.Info("NVML: Initializing")
-	libPath := getNVMLPath()
+	libPath := "libnvidia-ml.so.1"
 
-	lib, err := openLibrary(libPath)
+	// Check for standard locations if necessary, but purego/dlopen usually handles this
+	lib, err := purego.Dlopen(libPath, purego.RTLD_NOW|purego.RTLD_GLOBAL)
 	if err != nil {
 		return fmt.Errorf("failed to load %s: %w", libPath, err)
 	}
@@ -87,7 +90,8 @@ func (c *nvmlCollector) init() error {
 	purego.RegisterLibFunc(&nvmlDeviceGetHandleByIndex, lib, "nvmlDeviceGetHandleByIndex")
 	purego.RegisterLibFunc(&nvmlDeviceGetName, lib, "nvmlDeviceGetName")
 	// Try to get v2 memory info, fallback to v1 if not available
-	if hasSymbol(lib, "nvmlDeviceGetMemoryInfo_v2") {
+	_, err = purego.Dlsym(lib, "nvmlDeviceGetMemoryInfo_v2")
+	if err == nil {
 		c.isV2 = true
 		purego.RegisterLibFunc(&nvmlDeviceGetMemoryInfo, lib, "nvmlDeviceGetMemoryInfo_v2")
 	} else {
@@ -200,3 +204,34 @@ func (c *nvmlCollector) collect() {
 		slog.Info("NVML: Collected data", "gpu", gpu)
 	}
 }
+
+func (c *nvmlCollector) isGPUActive(bdf string) bool {
+	// runtime_status
+	statusPath := filepath.Join("/sys/bus/pci/devices", bdf, "power/runtime_status")
+	status, err := os.ReadFile(statusPath)
+	if err != nil {
+		slog.Info("NVML: Can't read runtime_status", "bdf", bdf, "err", err)
+		return true // Assume active if we can't read status
+	}
+	statusStr := strings.TrimSpace(string(status))
+	if statusStr != "active" && statusStr != "resuming" {
+		slog.Info("NVML: GPU is not active", "bdf", bdf, "status", statusStr)
+		return false
+	}
+
+	// power_state (D0 check)
+	// Find any drm card device power_state
+	pstatePathPattern := filepath.Join("/sys/bus/pci/devices", bdf, "drm/card*/device/power_state")
+	matches, _ := filepath.Glob(pstatePathPattern)
+	if len(matches) > 0 {
+		pstate, err := os.ReadFile(matches[0])
+		if err == nil {
+			pstateStr := strings.TrimSpace(string(pstate))
+			if pstateStr != "D0" {
+				return false
+			}
+		}
+	}
+
+	return true
+}

agent/gpu_nvml_linux.go

@@ -1,56 +0,0 @@
-//go:build linux
-
-package agent
-
-import (
-	"os"
-	"path/filepath"
-	"strings"
-
-	"github.com/ebitengine/purego"
-	"golang.org/x/exp/slog"
-)
-
-func openLibrary(name string) (uintptr, error) {
-	return purego.Dlopen(name, purego.RTLD_NOW|purego.RTLD_GLOBAL)
-}
-
-func getNVMLPath() string {
-	return "libnvidia-ml.so.1"
-}
-
-func hasSymbol(lib uintptr, symbol string) bool {
-	_, err := purego.Dlsym(lib, symbol)
-	return err == nil
-}
-
-func (c *nvmlCollector) isGPUActive(bdf string) bool {
-	// runtime_status
-	statusPath := filepath.Join("/sys/bus/pci/devices", bdf, "power/runtime_status")
-	status, err := os.ReadFile(statusPath)
-	if err != nil {
-		slog.Info("NVML: Can't read runtime_status", "bdf", bdf, "err", err)
-		return true // Assume active if we can't read status
-	}
-	statusStr := strings.TrimSpace(string(status))
-	if statusStr != "active" && statusStr != "resuming" {
-		slog.Info("NVML: GPU is not active", "bdf", bdf, "status", statusStr)
-		return false
-	}
-
-	// power_state (D0 check)
-	// Find any drm card device power_state
-	pstatePathPattern := filepath.Join("/sys/bus/pci/devices", bdf, "drm/card*/device/power_state")
-	matches, _ := filepath.Glob(pstatePathPattern)
-	if len(matches) > 0 {
-		pstate, err := os.ReadFile(matches[0])
-		if err == nil {
-			pstateStr := strings.TrimSpace(string(pstate))
-			if pstateStr != "D0" {
-				return false
-			}
-		}
-	}
-
-	return true
-}

agent/gpu_nvml_unsupported.go

@@ -1,21 +0,0 @@
-//go:build !linux && !windows
-
-package agent
-
-import "fmt"
-
-func openLibrary(name string) (uintptr, error) {
-	return 0, fmt.Errorf("nvml not supported on this platform")
-}
-
-func getNVMLPath() string {
-	return ""
-}
-
-func hasSymbol(lib uintptr, symbol string) bool {
-	return false
-}
-
-func (c *nvmlCollector) isGPUActive(bdf string) bool {
-	return true
-}

agent/gpu_nvml_windows.go

@@ -1,25 +0,0 @@
-//go:build windows
-
-package agent
-
-import (
-	"golang.org/x/sys/windows"
-)
-
-func openLibrary(name string) (uintptr, error) {
-	handle, err := windows.LoadLibrary(name)
-	return uintptr(handle), err
-}
-
-func getNVMLPath() string {
-	return "nvml.dll"
-}
-
-func hasSymbol(lib uintptr, symbol string) bool {
-	_, err := windows.GetProcAddress(windows.Handle(lib), symbol)
-	return err == nil
-}
-
-func (c *nvmlCollector) isGPUActive(bdf string) bool {
-	return true
-}

go.mod

@@ -21,7 +21,6 @@ require (
 	github.com/stretchr/testify v1.11.1
 	golang.org/x/crypto v0.45.0
 	golang.org/x/exp v0.0.0-20251125195548-87e1e737ad39
-	golang.org/x/sys v0.38.0
 	gopkg.in/yaml.v3 v3.0.1
 )
 
@@ -58,6 +57,7 @@ require (
 	golang.org/x/net v0.47.0 // indirect
 	golang.org/x/oauth2 v0.33.0 // indirect
 	golang.org/x/sync v0.18.0 // indirect
+	golang.org/x/sys v0.38.0 // indirect
 	golang.org/x/term v0.37.0 // indirect
 	golang.org/x/text v0.31.0 // indirect
 	howett.net/plist v1.0.1 // indirect

internal/site/src/components/routes/system/smart-table.tsx

@@ -93,15 +93,51 @@ export const smartColumns: ColumnDef<SmartAttribute>[] = [
 	},
 ]
 
+export type DiskInfo = {
+	id: string
+	system: string
+	device: string
+	model: string
+	capacity: string
+	status: string
+	temperature: number
+	deviceType: string
+	powerOnHours?: number
+	powerCycles?: number
+	attributes?: SmartAttribute[]
+	updated: string
+}
+
 // Function to format capacity display
 function formatCapacity(bytes: number): string {
 	const { value, unit } = formatBytes(bytes)
 	return `${toFixedFloat(value, value >= 10 ? 1 : 2)} ${unit}`
 }
 
+// Function to convert SmartDeviceRecord to DiskInfo
+function convertSmartDeviceRecordToDiskInfo(records: SmartDeviceRecord[]): DiskInfo[] {
+	const unknown = "Unknown"
+	return records.map((record) => ({
+		id: record.id,
+		system: record.system,
+		device: record.name || unknown,
+		model: record.model || unknown,
+		serialNumber: record.serial || unknown,
+		firmwareVersion: record.firmware || unknown,
+		capacity: record.capacity ? formatCapacity(record.capacity) : unknown,
+		status: record.state || unknown,
+		temperature: record.temp || 0,
+		deviceType: record.type || unknown,
+		attributes: record.attributes,
+		updated: record.updated,
+		powerOnHours: record.hours,
+		powerCycles: record.cycles,
+	}))
+}
+
 const SMART_DEVICE_FIELDS = "id,system,name,model,state,capacity,temp,type,hours,cycles,updated"
 
-export const columns: ColumnDef<SmartDeviceRecord>[] = [
+export const columns: ColumnDef<DiskInfo>[] = [
 	{
 		id: "system",
 		accessorFn: (record) => record.system,
@@ -118,12 +154,12 @@ export const columns: ColumnDef<SmartDeviceRecord>[] = [
 		},
 	},
 	{
-		accessorKey: "name",
-		sortingFn: (a, b) => a.original.name.localeCompare(b.original.name),
+		accessorKey: "device",
+		sortingFn: (a, b) => a.original.device.localeCompare(b.original.device),
 		header: ({ column }) => <HeaderButton column={column} name={t`Device`} Icon={HardDrive} />,
-		cell: ({ getValue }) => (
-			<div className="font-medium max-w-40 truncate ms-1.5" title={getValue() as string}>
-				{getValue() as string}
+		cell: ({ row }) => (
+			<div className="font-medium max-w-40 truncate ms-1.5" title={row.getValue("device")}>
+				{row.getValue("device")}
 			</div>
 		),
 	},
@@ -131,20 +167,19 @@ export const columns: ColumnDef<SmartDeviceRecord>[] = [
 		accessorKey: "model",
 		sortingFn: (a, b) => a.original.model.localeCompare(b.original.model),
 		header: ({ column }) => <HeaderButton column={column} name={t`Model`} Icon={Box} />,
-		cell: ({ getValue }) => (
-			<div className="max-w-48 truncate ms-1.5" title={getValue() as string}>
-				{getValue() as string}
+		cell: ({ row }) => (
+			<div className="max-w-48 truncate ms-1.5" title={row.getValue("model")}>
+				{row.getValue("model")}
 			</div>
 		),
 	},
 	{
 		accessorKey: "capacity",
-		invertSorting: true,
 		header: ({ column }) => <HeaderButton column={column} name={t`Capacity`} Icon={BinaryIcon} />,
-		cell: ({ getValue }) => <span className="ms-1.5">{formatCapacity(getValue() as number)}</span>,
+		cell: ({ getValue }) => <span className="ms-1.5">{getValue() as string}</span>,
 	},
 	{
-		accessorKey: "state",
+		accessorKey: "status",
 		header: ({ column }) => <HeaderButton column={column} name={t`Status`} Icon={Activity} />,
 		cell: ({ getValue }) => {
 			const status = getValue() as string
@@ -156,8 +191,8 @@ export const columns: ColumnDef<SmartDeviceRecord>[] = [
 		},
 	},
 	{
-		accessorKey: "type",
-		sortingFn: (a, b) => a.original.type.localeCompare(b.original.type),
+		accessorKey: "deviceType",
+		sortingFn: (a, b) => a.original.deviceType.localeCompare(b.original.deviceType),
 		header: ({ column }) => <HeaderButton column={column} name={t`Type`} Icon={ArrowLeftRightIcon} />,
 		cell: ({ getValue }) => (
 			<div className="ms-1.5">
@@ -168,7 +203,7 @@ export const columns: ColumnDef<SmartDeviceRecord>[] = [
 		),
 	},
 	{
-		accessorKey: "hours",
+		accessorKey: "powerOnHours",
 		invertSorting: true,
 		header: ({ column }) => (
 			<HeaderButton column={column} name={t({ message: "Power On", comment: "Power On Time" })} Icon={Clock} />
@@ -188,7 +223,7 @@ export const columns: ColumnDef<SmartDeviceRecord>[] = [
 		},
 	},
 	{
-		accessorKey: "cycles",
+		accessorKey: "powerCycles",
 		invertSorting: true,
 		header: ({ column }) => (
 			<HeaderButton column={column} name={t({ message: "Cycles", comment: "Power Cycles" })} Icon={RotateCwIcon} />
@@ -202,7 +237,7 @@ export const columns: ColumnDef<SmartDeviceRecord>[] = [
 		},
 	},
 	{
-		accessorKey: "temp",
+		accessorKey: "temperature",
 		invertSorting: true,
 		header: ({ column }) => <HeaderButton column={column} name={t`Temp`} Icon={ThermometerIcon} />,
 		cell: ({ getValue }) => {
@@ -211,14 +246,14 @@ export const columns: ColumnDef<SmartDeviceRecord>[] = [
 		},
 	},
 	// {
-	// 	accessorKey: "serial",
-	// 	sortingFn: (a, b) => a.original.serial.localeCompare(b.original.serial),
+	// 	accessorKey: "serialNumber",
+	// 	sortingFn: (a, b) => a.original.serialNumber.localeCompare(b.original.serialNumber),
 	// 	header: ({ column }) => <HeaderButton column={column} name={t`Serial Number`} Icon={HashIcon} />,
 	// 	cell: ({ getValue }) => <span className="ms-1.5">{getValue() as string}</span>,
 	// },
 	// {
-	// 	accessorKey: "firmware",
-	// 	sortingFn: (a, b) => a.original.firmware.localeCompare(b.original.firmware),
+	// 	accessorKey: "firmwareVersion",
+	// 	sortingFn: (a, b) => a.original.firmwareVersion.localeCompare(b.original.firmwareVersion),
 	// 	header: ({ column }) => <HeaderButton column={column} name={t`Firmware`} Icon={CpuIcon} />,
 	// 	cell: ({ getValue }) => <span className="ms-1.5">{getValue() as string}</span>,
 	// },
@@ -237,15 +272,7 @@ export const columns: ColumnDef<SmartDeviceRecord>[] = [
 	},
 ]
 
-function HeaderButton({
-	column,
-	name,
-	Icon,
-}: {
-	column: Column<SmartDeviceRecord>
-	name: string
-	Icon: React.ElementType
-}) {
+function HeaderButton({ column, name, Icon }: { column: Column<DiskInfo>; name: string; Icon: React.ElementType }) {
 	const isSorted = column.getIsSorted()
 	return (
 		<Button
@@ -263,7 +290,7 @@ function HeaderButton({
 }
 
 export default function DisksTable({ systemId }: { systemId?: string }) {
-	const [sorting, setSorting] = useState<SortingState>([{ id: systemId ? "name" : "system", desc: false }])
+	const [sorting, setSorting] = useState<SortingState>([{ id: systemId ? "device" : "system", desc: false }])
 	const [columnFilters, setColumnFilters] = useState<ColumnFiltersState>([])
 	const [rowSelection, setRowSelection] = useState({})
 	const [smartDevices, setSmartDevices] = useState<SmartDeviceRecord[] | undefined>(undefined)
@@ -272,95 +299,96 @@ export default function DisksTable({ systemId }: { systemId?: string }) {
 	const [rowActionState, setRowActionState] = useState<{ type: "refresh" | "delete"; id: string } | null>(null)
 	const [globalFilter, setGlobalFilter] = useState("")
 
-	const openSheet = (disk: SmartDeviceRecord) => {
+	const openSheet = (disk: DiskInfo) => {
 		setActiveDiskId(disk.id)
 		setSheetOpen(true)
 	}
 
-	// Fetch smart devices
-	useEffect(() => {
-		const controller = new AbortController()
-
+	// Fetch smart devices from collection (without attributes to save bandwidth)
+	const fetchSmartDevices = useCallback(() => {
 		pb.collection<SmartDeviceRecord>("smart_devices")
 			.getFullList({
 				filter: systemId ? pb.filter("system = {:system}", { system: systemId }) : undefined,
 				fields: SMART_DEVICE_FIELDS,
-				signal: controller.signal,
 			})
-			.then(setSmartDevices)
-			.catch((err) => {
-				if (!err.isAbort) {
-					setSmartDevices([])
-				}
+			.then((records) => {
+				setSmartDevices(records)
 			})
-
-		return () => controller.abort()
+			.catch(() => setSmartDevices([]))
 	}, [systemId])
 
-	// Subscribe to updates
+	// Fetch smart devices when component mounts or systemId changes
+	useEffect(() => {
+		fetchSmartDevices()
+	}, [fetchSmartDevices])
+
+	// Subscribe to live updates so rows add/remove without manual refresh/filtering
 	useEffect(() => {
 		let unsubscribe: (() => void) | undefined
 		const pbOptions = systemId
 			? { fields: SMART_DEVICE_FIELDS, filter: pb.filter("system = {:system}", { system: systemId }) }
 			: { fields: SMART_DEVICE_FIELDS }
 
-			; (async () => {
-				try {
-					unsubscribe = await pb.collection("smart_devices").subscribe(
-						"*",
-						(event) => {
-							const record = event.record as SmartDeviceRecord
-							setSmartDevices((currentDevices) => {
-								const devices = currentDevices ?? []
-								const matchesSystemScope = !systemId || record.system === systemId
+		;(async () => {
+			try {
+				unsubscribe = await pb.collection("smart_devices").subscribe(
+					"*",
+					(event) => {
+						const record = event.record as SmartDeviceRecord
+						setSmartDevices((currentDevices) => {
+							const devices = currentDevices ?? []
+							const matchesSystemScope = !systemId || record.system === systemId
 
-								if (event.action === "delete") {
-									return devices.filter((device) => device.id !== record.id)
-								}
+							if (event.action === "delete") {
+								return devices.filter((device) => device.id !== record.id)
+							}
 
-								if (!matchesSystemScope) {
-									// Record moved out of scope; ensure it disappears locally.
-									return devices.filter((device) => device.id !== record.id)
-								}
+							if (!matchesSystemScope) {
+								// Record moved out of scope; ensure it disappears locally.
+								return devices.filter((device) => device.id !== record.id)
+							}
 
-								const existingIndex = devices.findIndex((device) => device.id === record.id)
-								if (existingIndex === -1) {
-									return [record, ...devices]
-								}
+							const existingIndex = devices.findIndex((device) => device.id === record.id)
+							if (existingIndex === -1) {
+								return [record, ...devices]
+							}
 
-								const next = [...devices]
-								next[existingIndex] = record
-								return next
-							})
-						},
-						pbOptions
-					)
-				} catch (error) {
-					console.error("Failed to subscribe to SMART device updates:", error)
-				}
-			})()
+							const next = [...devices]
+							next[existingIndex] = record
+							return next
+						})
+					},
+					pbOptions
+				)
+			} catch (error) {
+				console.error("Failed to subscribe to SMART device updates:", error)
+			}
+		})()
 
 		return () => {
 			unsubscribe?.()
 		}
 	}, [systemId])
 
-	const handleRowRefresh = useCallback(async (disk: SmartDeviceRecord) => {
-		if (!disk.system) return
-		setRowActionState({ type: "refresh", id: disk.id })
-		try {
-			await pb.send("/api/beszel/smart/refresh", {
-				method: "POST",
-				query: { system: disk.system },
-			})
-		} catch (error) {
-			console.error("Failed to refresh SMART device:", error)
-		} finally {
-			setRowActionState((state) => (state?.id === disk.id ? null : state))
-		}
-	}, [])
+	const handleRowRefresh = useCallback(
+		async (disk: DiskInfo) => {
+			if (!disk.system) return
+			setRowActionState({ type: "refresh", id: disk.id })
+			try {
+				await pb.send("/api/beszel/smart/refresh", {
+					method: "POST",
+					query: { system: disk.system },
+				})
+			} catch (error) {
+				console.error("Failed to refresh SMART device:", error)
+			} finally {
+				setRowActionState((state) => (state?.id === disk.id ? null : state))
+			}
+		},
+		[fetchSmartDevices]
+	)
 
-	const handleDeleteDevice = useCallback(async (disk: SmartDeviceRecord) => {
+	const handleDeleteDevice = useCallback(async (disk: DiskInfo) => {
 		setRowActionState({ type: "delete", id: disk.id })
 		try {
 			await pb.collection("smart_devices").delete(disk.id)
@@ -372,7 +400,7 @@ export default function DisksTable({ systemId }: { systemId?: string }) {
 		}
 	}, [])
 
-	const actionColumn = useMemo<ColumnDef<SmartDeviceRecord>>(
+	const actionColumn = useMemo<ColumnDef<DiskInfo>>(
 		() => ({
 			id: "actions",
 			enableSorting: false,
@@ -440,8 +468,13 @@ export default function DisksTable({ systemId }: { systemId?: string }) {
 		return [...baseColumns, actionColumn]
 	}, [systemId, actionColumn])
 
+	// Convert SmartDeviceRecord to DiskInfo
+	const diskData = useMemo(() => {
+		return smartDevices ? convertSmartDeviceRecordToDiskInfo(smartDevices) : []
+	}, [smartDevices])
+
 	const table = useReactTable({
-		data: smartDevices || ([] as SmartDeviceRecord[]),
+		data: diskData,
 		columns: tableColumns,
 		onSortingChange: setSorting,
 		onColumnFiltersChange: setColumnFilters,
@@ -459,10 +492,10 @@ export default function DisksTable({ systemId }: { systemId?: string }) {
 		globalFilterFn: (row, _columnId, filterValue) => {
 			const disk = row.original
 			const systemName = $allSystemsById.get()[disk.system]?.name ?? ""
-			const device = disk.name ?? ""
+			const device = disk.device ?? ""
 			const model = disk.model ?? ""
-			const status = disk.state ?? ""
-			const type = disk.type ?? ""
+			const status = disk.status ?? ""
+			const type = disk.deviceType ?? ""
 			const searchString = `${systemName} ${device} ${model} ${status} ${type}`.toLowerCase()
 			return (filterValue as string)
 				.toLowerCase()
@@ -472,7 +505,7 @@ export default function DisksTable({ systemId }: { systemId?: string }) {
 	})
 
 	// Hide the table on system pages if there's no data, but always show on global page
-	if (systemId && !smartDevices?.length && !columnFilters.length) {
+	if (systemId && !diskData.length && !columnFilters.length) {
 		return null
 	}