[Feature] Add detailed CPU metrics (User, System, IOWait, Steal) with per-core monitoring (#1356)

* Add user, system io wait

* add per cpu core

* add total

agent/cpu.go

@@ -8,6 +8,7 @@ import (
 )
 
 var lastCpuTimes = make(map[uint16]cpu.TimesStat)
+var lastPerCoreCpuTimes = make(map[uint16][]cpu.TimesStat)
 
 // init initializes the CPU monitoring by storing the initial CPU times
 // for the default 60-second cache interval.
@@ -15,6 +16,18 @@ func init() {
 	if times, err := cpu.Times(false); err == nil {
 		lastCpuTimes[60000] = times[0]
 	}
+	if perCoreTimes, err := cpu.Times(true); err == nil {
+		lastPerCoreCpuTimes[60000] = perCoreTimes
+	}
+}
+
+// CpuMetrics contains detailed CPU usage breakdown
+type CpuMetrics struct {
+	Total  float64
+	User   float64
+	System float64
+	Iowait float64
+	Steal  float64
 }
 
 // getCpuPercent calculates the CPU usage percentage using cached previous measurements.
@@ -34,6 +47,92 @@ func getCpuPercent(cacheTimeMs uint16) (float64, error) {
 	return delta, nil
 }
 
+// getCpuMetrics calculates detailed CPU usage metrics using cached previous measurements.
+// It returns percentages for total, user, system, iowait, and steal time.
+func getCpuMetrics(cacheTimeMs uint16) (CpuMetrics, error) {
+	times, err := cpu.Times(false)
+	if err != nil || len(times) == 0 {
+		return CpuMetrics{}, err
+	}
+	// if cacheTimeMs is not in lastCpuTimes, use 60000 as fallback lastCpuTime
+	if _, ok := lastCpuTimes[cacheTimeMs]; !ok {
+		lastCpuTimes[cacheTimeMs] = lastCpuTimes[60000]
+	}
+
+	t1 := lastCpuTimes[cacheTimeMs]
+	t2 := times[0]
+
+	t1All, t1Busy := getAllBusy(t1)
+	t2All, t2Busy := getAllBusy(t2)
+
+	totalDelta := t2All - t1All
+	if totalDelta <= 0 {
+		return CpuMetrics{}, nil
+	}
+
+	metrics := CpuMetrics{
+		Total:  clampPercent((t2Busy - t1Busy) / totalDelta * 100),
+		User:   clampPercent((t2.User - t1.User) / totalDelta * 100),
+		System: clampPercent((t2.System - t1.System) / totalDelta * 100),
+		Iowait: clampPercent((t2.Iowait - t1.Iowait) / totalDelta * 100),
+		Steal:  clampPercent((t2.Steal - t1.Steal) / totalDelta * 100),
+	}
+
+	lastCpuTimes[cacheTimeMs] = times[0]
+	return metrics, nil
+}
+
+// clampPercent ensures the percentage is between 0 and 100
+func clampPercent(value float64) float64 {
+	return math.Min(100, math.Max(0, value))
+}
+
+// getPerCoreCpuMetrics calculates per-core CPU usage metrics.
+// Returns a map where the key is "cpu0", "cpu1", etc. and the value is an array of [user, system, iowait, steal] percentages.
+func getPerCoreCpuMetrics(cacheTimeMs uint16) (map[string][4]float64, error) {
+	perCoreTimes, err := cpu.Times(true)
+	if err != nil || len(perCoreTimes) == 0 {
+		return nil, err
+	}
+
+	// Initialize cache if needed
+	if _, ok := lastPerCoreCpuTimes[cacheTimeMs]; !ok {
+		lastPerCoreCpuTimes[cacheTimeMs] = lastPerCoreCpuTimes[60000]
+	}
+
+	lastTimes := lastPerCoreCpuTimes[cacheTimeMs]
+	result := make(map[string][4]float64)
+
+	// Calculate metrics for each core
+	for i, currentTime := range perCoreTimes {
+		if i >= len(lastTimes) {
+			break
+		}
+
+		t1 := lastTimes[i]
+		t2 := currentTime
+
+		t1All, _ := getAllBusy(t1)
+		t2All, _ := getAllBusy(t2)
+
+		totalDelta := t2All - t1All
+		if totalDelta <= 0 {
+			continue
+		}
+
+		// Store as [user, system, iowait, steal]
+		result[currentTime.CPU] = [4]float64{
+			clampPercent((t2.User - t1.User) / totalDelta * 100),
+			clampPercent((t2.System - t1.System) / totalDelta * 100),
+			clampPercent((t2.Iowait - t1.Iowait) / totalDelta * 100),
+			clampPercent((t2.Steal - t1.Steal) / totalDelta * 100),
+		}
+	}
+
+	lastPerCoreCpuTimes[cacheTimeMs] = perCoreTimes
+	return result, nil
+}
+
 // calculateBusy calculates the CPU busy percentage between two time points.
 // It computes the ratio of busy time to total time elapsed between t1 and t2,
 // returning a percentage clamped between 0 and 100.

agent/docker.go

@@ -54,7 +54,7 @@ type dockerManager struct {
 	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
-	containerExclude    []string                    // Patterns to exclude containers by name (supports wildcards)
+	excludeContainers   []string                    // Patterns to exclude containers by name
 
 	// Cache-time-aware tracking for CPU stats (similar to cpu.go)
 	// Maps cache time intervals to container-specific CPU usage tracking
@@ -96,13 +96,12 @@ func (d *dockerManager) dequeue() {
 	}
 }
 
-// shouldExcludeContainer checks if a container name matches any exclusion pattern using path.Match
+// shouldExcludeContainer checks if a container name matches any exclusion pattern
 func (dm *dockerManager) shouldExcludeContainer(name string) bool {
-	if len(dm.containerExclude) == 0 {
+	if len(dm.excludeContainers) == 0 {
 		return false
 	}
-	for _, pattern := range dm.containerExclude {
+	for _, pattern := range dm.excludeContainers {
-		// Use path.Match for wildcard support
 		if match, _ := path.Match(pattern, name); match {
 			return true
 		}
@@ -138,15 +137,9 @@ func (dm *dockerManager) getDockerStats(cacheTimeMs uint16) ([]*container.Stats,
 	for _, ctr := range dm.apiContainerList {
 		ctr.IdShort = ctr.Id[:12]
 
-		// Extract container name and check if it should be excluded
-		name := ctr.Names[0]
-		if len(name) > 0 && name[0] == '/' {
-			name = name[1:]
-		}
-
 		// Skip this container if it matches the exclusion pattern
-		if dm.shouldExcludeContainer(name) {
+		if dm.shouldExcludeContainer(ctr.Names[0][1:]) {
-			slog.Debug("Excluding container", "name", name, "patterns", dm.containerExclude)
+			slog.Debug("Excluding container", "name", ctr.Names[0][1:])
 			continue
 		}
 
@@ -532,20 +525,17 @@ func newDockerManager(a *Agent) *dockerManager {
 		userAgent: "Docker-Client/",
 	}
 
-	// Read container exclusion patterns from environment variable (comma-separated, supports wildcards)
+	// Read container exclusion patterns from environment variable
-	var containerExclude []string
+	var excludeContainers []string
-	if excludeStr, set := GetEnv("CONTAINER_EXCLUDE"); set && excludeStr != "" {
+	if excludeStr, set := GetEnv("EXCLUDE_CONTAINERS"); set && excludeStr != "" {
-		// Split by comma and trim whitespace
+		parts := strings.SplitSeq(excludeStr, ",")
-		parts := strings.Split(excludeStr, ",")
+		for part := range parts {
-		for _, part := range parts {
 			trimmed := strings.TrimSpace(part)
 			if trimmed != "" {
-				containerExclude = append(containerExclude, trimmed)
+				excludeContainers = append(excludeContainers, trimmed)
 			}
 		}
-		if len(containerExclude) > 0 {
+		slog.Info("EXCLUDE_CONTAINERS", "patterns", excludeContainers)
-			slog.Info("Container exclusion patterns set", "patterns", containerExclude)
-		}
 	}
 
 	manager := &dockerManager{
@@ -557,7 +547,7 @@ func newDockerManager(a *Agent) *dockerManager {
 		sem:               make(chan struct{}, 5),
 		apiContainerList:  []*container.ApiInfo{},
 		apiStats:          &container.ApiStats{},
-		containerExclude:  containerExclude,
+		excludeContainers: excludeContainers,
 
 		// Initialize cache-time-aware tracking structures
 		lastCpuContainer:    make(map[uint16]map[string]uint64),

agent/docker_test.go

@@ -1196,7 +1196,7 @@ func TestShouldExcludeContainer(t *testing.T) {
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
 			dm := &dockerManager{
-				containerExclude: tt.patterns,
+				excludeContainers: tt.patterns,
 			}
 			result := dm.shouldExcludeContainer(tt.containerName)
 			assert.Equal(t, tt.expected, result)

agent/system.go

@@ -83,12 +83,21 @@ func (a *Agent) getSystemStats(cacheTimeMs uint16) system.Stats {
 		systemStats.Battery[1] = batteryState
 	}
 
-	// cpu percent
+	// cpu metrics
-	cpuPercent, err := getCpuPercent(cacheTimeMs)
+	cpuMetrics, err := getCpuMetrics(cacheTimeMs)
 	if err == nil {
-		systemStats.Cpu = twoDecimals(cpuPercent)
+		systemStats.Cpu = twoDecimals(cpuMetrics.Total)
+		systemStats.CpuUser = twoDecimals(cpuMetrics.User)
+		systemStats.CpuSystem = twoDecimals(cpuMetrics.System)
+		systemStats.CpuIowait = twoDecimals(cpuMetrics.Iowait)
+		systemStats.CpuSteal = twoDecimals(cpuMetrics.Steal)
 	} else {
-		slog.Error("Error getting cpu percent", "err", err)
+		slog.Error("Error getting cpu metrics", "err", err)
+	}
+
+	// per-core cpu metrics
+	if perCoreCpuMetrics, err := getPerCoreCpuMetrics(cacheTimeMs); err == nil && len(perCoreCpuMetrics) > 0 {
+		systemStats.CpuCores = perCoreCpuMetrics
 	}
 
 	// load average

internal/entities/system/system.go

@@ -11,7 +11,12 @@ import (
 type Stats struct {
 	Cpu            float64             `json:"cpu" cbor:"0,keyasint"`
 	MaxCpu         float64             `json:"cpum,omitempty" cbor:"1,keyasint,omitempty"`
-	Mem            float64             `json:"m" cbor:"2,keyasint"`
+	CpuUser        float64                `json:"cpuu,omitempty" cbor:"33,keyasint,omitempty"`
+	CpuSystem      float64                `json:"cpus,omitempty" cbor:"34,keyasint,omitempty"`
+	CpuIowait      float64                `json:"cpui,omitempty" cbor:"35,keyasint,omitempty"`
+	CpuSteal       float64                `json:"cpust,omitempty" cbor:"36,keyasint,omitempty"`
+	CpuCores       map[string][4]float64  `json:"cpuc,omitempty" cbor:"37,keyasint,omitempty"` // [user, system, iowait, steal] per core
+	Mem            float64                `json:"m" cbor:"2,keyasint"`
 	MemUsed        float64             `json:"mu" cbor:"3,keyasint"`
 	MemPct         float64             `json:"mp" cbor:"4,keyasint"`
 	MemBuffCache   float64             `json:"mb" cbor:"5,keyasint"`

internal/site/src/components/charts/hooks.ts

@@ -118,6 +118,28 @@ export function useNetworkInterfaces(interfaces: SystemStats["ni"]) {
 				dataKey: ({ stats }: SystemStatsRecord) => stats?.ni?.[key]?.[index],
 				color: `hsl(${220 + (((sortedKeys.indexOf(key) * 360) / sortedKeys.length) % 360)}, 70%, 50%)`,
 
+				opacity: 0.3,
+			}))
+		},
+	}
+}
+
+// Assures consistent colors for CPU cores
+export function useCpuCores(cores: SystemStats["cpuc"]) {
+	const keys = Object.keys(cores ?? {})
+	// Sort cores by name (cpu0, cpu1, cpu2, etc.)
+	const sortedKeys = keys.sort((a, b) => {
+		const numA = Number.parseInt(a.replace("cpu", ""))
+		const numB = Number.parseInt(b.replace("cpu", ""))
+		return numA - numB
+	})
+	return {
+		length: sortedKeys.length,
+		data: (index = 0) => {
+			return sortedKeys.map((key) => ({
+				label: key,
+				dataKey: ({ stats }: SystemStatsRecord) => stats?.cpuc?.[key]?.[index],
+				color: `hsl(${(((sortedKeys.indexOf(key) * 360) / sortedKeys.length) % 360)}, 70%, 50%)`,
 				opacity: 0.3,
 			}))
 		},

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

@@ -73,6 +73,7 @@ import { Select, SelectContent, SelectItem, SelectTrigger, SelectValue } from ".
 import { Separator } from "../ui/separator"
 import { Tooltip, TooltipContent, TooltipProvider, TooltipTrigger } from "../ui/tooltip"
 import NetworkSheet from "./system/network-sheet"
+import CpuCoresSheet from "./system/cpu-cores-sheet"
 import LineChartDefault from "../charts/line-chart"
 
 
@@ -585,18 +586,49 @@ export default memo(function SystemDetail({ id }: { id: string }) {
 						grid={grid}
 						title={t`CPU Usage`}
 						description={t`Average system-wide CPU utilization`}
-						cornerEl={maxValSelect}
+						cornerEl={
+							<>
+								{maxValSelect}
+								<CpuCoresSheet chartData={chartData} dataEmpty={dataEmpty} grid={grid} maxValues={maxValues} />
+							</>
+						}
+						legend={true}
 					>
 						<AreaChartDefault
 							chartData={chartData}
 							maxToggled={maxValues}
+							legend={true}
 							dataPoints={[
 								{
-									label: t`CPU Usage`,
+									label: t`Total`,
 									dataKey: ({ stats }) => (showMax ? stats?.cpum : stats?.cpu),
 									color: 1,
 									opacity: 0.4,
 								},
+								{
+									label: t`User`,
+									dataKey: ({ stats }) => stats?.cpuu,
+									color: 2,
+									opacity: 0.3,
+								},
+								{
+									label: t`System`,
+									dataKey: ({ stats }) => stats?.cpus,
+									color: 3,
+									opacity: 0.3,
+								},
+								{
+									label: t`IOWait`,
+									dataKey: ({ stats }) => stats?.cpui,
+									color: 4,
+									opacity: 0.3,
+								},
+								{
+									label: t`Steal`,
+									dataKey: ({ stats }) => stats?.cpust,
+									color: 5,
+									opacity: 0.3,
+								},
 							]}
 							tickFormatter={(val) => `${toFixedFloat(val, 2)}%`}
 							contentFormatter={({ value }) => `${decimalString(value)}%`}

internal/site/src/components/routes/system/cpu-cores-sheet.tsx

@@ -0,0 +1,119 @@
+import { t } from "@lingui/core/macro"
+import { MoreHorizontalIcon } from "lucide-react"
+import { memo, useRef, useState } from "react"
+import AreaChartDefault from "@/components/charts/area-chart"
+import ChartTimeSelect from "@/components/charts/chart-time-select"
+import { Button } from "@/components/ui/button"
+import { Sheet, SheetContent, SheetTrigger } from "@/components/ui/sheet"
+import { DialogTitle } from "@/components/ui/dialog"
+import { decimalString, toFixedFloat } from "@/lib/utils"
+import type { ChartData, SystemStatsRecord } from "@/types"
+import { ChartCard } from "../system"
+
+export default memo(function CpuCoresSheet({
+	chartData,
+	dataEmpty,
+	grid,
+	maxValues,
+}: {
+	chartData: ChartData
+	dataEmpty: boolean
+	grid: boolean
+	maxValues: boolean
+}) {
+	const [cpuCoresOpen, setCpuCoresOpen] = useState(false)
+	const hasOpened = useRef(false)
+
+	if (cpuCoresOpen && !hasOpened.current) {
+		hasOpened.current = true
+	}
+
+	// Get list of CPU cores from the latest stats
+	const cpuCoresData = chartData.systemStats.at(-1)?.stats?.cpuc ?? {}
+	const coreNames = Object.keys(cpuCoresData).sort((a, b) => {
+		const numA = Number.parseInt(a.replace("cpu", ""))
+		const numB = Number.parseInt(b.replace("cpu", ""))
+		return numA - numB
+	})
+
+	if (coreNames.length === 0) {
+		return null
+	}
+
+	return (
+		<Sheet open={cpuCoresOpen} onOpenChange={setCpuCoresOpen}>
+			<DialogTitle className="sr-only">{t`Per-core CPU usage`}</DialogTitle>
+			<SheetTrigger asChild>
+				<Button
+					title={t`View per-core CPU`}
+					variant="outline"
+					size="icon"
+					className="shrink-0 max-sm:absolute max-sm:top-3 max-sm:end-3"
+				>
+					<MoreHorizontalIcon />
+				</Button>
+			</SheetTrigger>
+			{hasOpened.current && (
+				<SheetContent aria-describedby={undefined} className="overflow-auto w-200 !max-w-full p-4 sm:p-6">
+					<ChartTimeSelect className="w-[calc(100%-2em)]" agentVersion={chartData.agentVersion} />
+					{coreNames.map((coreName) => (
+						<ChartCard
+							key={coreName}
+							empty={dataEmpty}
+							grid={grid}
+							title={coreName.toUpperCase()}
+							description={t`CPU usage breakdown for ${coreName}`}
+							legend={true}
+							className="min-h-auto"
+						>
+							<AreaChartDefault
+								chartData={chartData}
+								maxToggled={maxValues}
+								legend={true}
+								dataPoints={[
+									{
+										label: t`Total`,
+										dataKey: ({ stats }: SystemStatsRecord) => {
+											const core = stats?.cpuc?.[coreName]
+											if (!core) return undefined
+											// Sum all metrics: user + system + iowait + steal
+											return core[0] + core[1] + core[2] + core[3]
+										},
+										color: 1,
+										opacity: 0.4,
+									},
+									{
+										label: t`User`,
+										dataKey: ({ stats }: SystemStatsRecord) => stats?.cpuc?.[coreName]?.[0],
+										color: 2,
+										opacity: 0.3,
+									},
+									{
+										label: t`System`,
+										dataKey: ({ stats }: SystemStatsRecord) => stats?.cpuc?.[coreName]?.[1],
+										color: 3,
+										opacity: 0.3,
+									},
+									{
+										label: t`IOWait`,
+										dataKey: ({ stats }: SystemStatsRecord) => stats?.cpuc?.[coreName]?.[2],
+										color: 4,
+										opacity: 0.3,
+									},
+									{
+										label: t`Steal`,
+										dataKey: ({ stats }: SystemStatsRecord) => stats?.cpuc?.[coreName]?.[3],
+										color: 5,
+										opacity: 0.3,
+									},
+								]}
+								tickFormatter={(val) => `${toFixedFloat(val, 2)}%`}
+								contentFormatter={({ value }) => `${decimalString(value)}%`}
+							/>
+						</ChartCard>
+					))}
+				</SheetContent>
+			)}
+		</Sheet>
+	)
+})

internal/site/src/types.d.ts

@@ -84,6 +84,16 @@ export interface SystemStats {
 	cpu: number
 	/** peak cpu */
 	cpum?: number
+	/** cpu user percent */
+	cpuu?: number
+	/** cpu system percent */
+	cpus?: number
+	/** cpu iowait percent */
+	cpui?: number
+	/** cpu steal percent */
+	cpust?: number
+	/** per-core cpu metrics [user, system, iowait, steal] */
+	cpuc?: Record<string, [number, number, number, number]>
 	// TODO: remove these in future release in favor of la
 	/** load average 1 minute */
 	l1?: number