refactor load average handling (#982)

- Transitioned from individual load average fields (LoadAvg1, LoadAvg5,
LoadAvg15) to a single array (LoadAvg)
- Ensure load is displayed when all zero values.
This commit is contained in:
henrygd
2025-07-25 18:07:11 -04:00
parent 6953edf59e
commit 91679b5cc0
10 changed files with 116 additions and 69 deletions

View File

@@ -80,10 +80,11 @@ func (a *Agent) getSystemStats() system.Stats {
// load average
if avgstat, err := load.Avg(); err == nil {
systemStats.LoadAvg1 = twoDecimals(avgstat.Load1)
systemStats.LoadAvg5 = twoDecimals(avgstat.Load5)
systemStats.LoadAvg15 = twoDecimals(avgstat.Load15)
slog.Debug("Load average", "5m", systemStats.LoadAvg5, "15m", systemStats.LoadAvg15)
// TODO: remove these in future release in favor of load avg array
systemStats.LoadAvg[0] = avgstat.Load1
systemStats.LoadAvg[1] = avgstat.Load5
systemStats.LoadAvg[2] = avgstat.Load15
slog.Debug("Load average", "5m", avgstat.Load5, "15m", avgstat.Load15)
} else {
slog.Error("Error getting load average", "err", err)
}
@@ -255,9 +256,11 @@ func (a *Agent) getSystemStats() system.Stats {
// update base system info
a.systemInfo.Cpu = systemStats.Cpu
a.systemInfo.LoadAvg1 = systemStats.LoadAvg1
a.systemInfo.LoadAvg5 = systemStats.LoadAvg5
a.systemInfo.LoadAvg15 = systemStats.LoadAvg15
a.systemInfo.LoadAvg = systemStats.LoadAvg
// TODO: remove these in future release in favor of load avg array
a.systemInfo.LoadAvg1 = systemStats.LoadAvg[0]
a.systemInfo.LoadAvg5 = systemStats.LoadAvg[1]
a.systemInfo.LoadAvg15 = systemStats.LoadAvg[2]
a.systemInfo.MemPct = systemStats.MemPct
a.systemInfo.DiskPct = systemStats.DiskPct
a.systemInfo.Uptime, _ = host.Uptime()

View File

@@ -47,9 +47,7 @@ type SystemAlertStats struct {
NetSent float64 `json:"ns"`
NetRecv float64 `json:"nr"`
Temperatures map[string]float32 `json:"t"`
LoadAvg1 float64 `json:"l1"`
LoadAvg5 float64 `json:"l5"`
LoadAvg15 float64 `json:"l15"`
LoadAvg [3]float64 `json:"la"`
}
type SystemAlertData struct {

View File

@@ -55,13 +55,13 @@ func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *syst
val = data.Info.DashboardTemp
unit = "°C"
case "LoadAvg1":
val = data.Info.LoadAvg1
val = data.Info.LoadAvg[0]
unit = ""
case "LoadAvg5":
val = data.Info.LoadAvg5
val = data.Info.LoadAvg[1]
unit = ""
case "LoadAvg15":
val = data.Info.LoadAvg15
val = data.Info.LoadAvg[2]
unit = ""
}
@@ -200,11 +200,11 @@ func (am *AlertManager) HandleSystemAlerts(systemRecord *core.Record, data *syst
alert.mapSums[key] += temp
}
case "LoadAvg1":
alert.val += stats.LoadAvg1
alert.val += stats.LoadAvg[0]
case "LoadAvg5":
alert.val += stats.LoadAvg5
alert.val += stats.LoadAvg[1]
case "LoadAvg15":
alert.val += stats.LoadAvg15
alert.val += stats.LoadAvg[2]
default:
continue
}

View File

@@ -31,11 +31,13 @@ type Stats struct {
Temperatures map[string]float64 `json:"t,omitempty" cbor:"20,keyasint,omitempty"`
ExtraFs map[string]*FsStats `json:"efs,omitempty" cbor:"21,keyasint,omitempty"`
GPUData map[string]GPUData `json:"g,omitempty" cbor:"22,keyasint,omitempty"`
LoadAvg1 float64 `json:"l1,omitempty" cbor:"23,keyasint,omitempty,omitzero"`
LoadAvg5 float64 `json:"l5,omitempty" cbor:"24,keyasint,omitempty,omitzero"`
LoadAvg15 float64 `json:"l15,omitempty" cbor:"25,keyasint,omitempty,omitzero"`
LoadAvg1 float64 `json:"l1,omitempty" cbor:"23,keyasint,omitempty"`
LoadAvg5 float64 `json:"l5,omitempty" cbor:"24,keyasint,omitempty"`
LoadAvg15 float64 `json:"l15,omitempty" cbor:"25,keyasint,omitempty"`
Bandwidth [2]uint64 `json:"b,omitzero" cbor:"26,keyasint,omitzero"` // [sent bytes, recv bytes]
MaxBandwidth [2]uint64 `json:"bm,omitzero" cbor:"27,keyasint,omitzero"` // [sent bytes, recv bytes]
LoadAvg [3]float64 `json:"la,omitempty" cbor:"28,keyasint"`
// TODO: remove other load fields in future release in favor of load avg array
}
type GPUData struct {
@@ -79,25 +81,27 @@ const (
)
type Info struct {
Hostname string `json:"h" cbor:"0,keyasint"`
KernelVersion string `json:"k,omitempty" cbor:"1,keyasint,omitempty"`
Cores int `json:"c" cbor:"2,keyasint"`
Threads int `json:"t,omitempty" cbor:"3,keyasint,omitempty"`
CpuModel string `json:"m" cbor:"4,keyasint"`
Uptime uint64 `json:"u" cbor:"5,keyasint"`
Cpu float64 `json:"cpu" cbor:"6,keyasint"`
MemPct float64 `json:"mp" cbor:"7,keyasint"`
DiskPct float64 `json:"dp" cbor:"8,keyasint"`
Bandwidth float64 `json:"b" cbor:"9,keyasint"`
AgentVersion string `json:"v" cbor:"10,keyasint"`
Podman bool `json:"p,omitempty" cbor:"11,keyasint,omitempty"`
GpuPct float64 `json:"g,omitempty" cbor:"12,keyasint,omitempty"`
DashboardTemp float64 `json:"dt,omitempty" cbor:"13,keyasint,omitempty"`
Os Os `json:"os" cbor:"14,keyasint"`
LoadAvg1 float64 `json:"l1,omitempty" cbor:"15,keyasint,omitempty"`
LoadAvg5 float64 `json:"l5,omitempty" cbor:"16,keyasint,omitempty"`
LoadAvg15 float64 `json:"l15,omitempty" cbor:"17,keyasint,omitempty"`
BandwidthBytes uint64 `json:"bb" cbor:"18,keyasint"`
Hostname string `json:"h" cbor:"0,keyasint"`
KernelVersion string `json:"k,omitempty" cbor:"1,keyasint,omitempty"`
Cores int `json:"c" cbor:"2,keyasint"`
Threads int `json:"t,omitempty" cbor:"3,keyasint,omitempty"`
CpuModel string `json:"m" cbor:"4,keyasint"`
Uptime uint64 `json:"u" cbor:"5,keyasint"`
Cpu float64 `json:"cpu" cbor:"6,keyasint"`
MemPct float64 `json:"mp" cbor:"7,keyasint"`
DiskPct float64 `json:"dp" cbor:"8,keyasint"`
Bandwidth float64 `json:"b" cbor:"9,keyasint"`
AgentVersion string `json:"v" cbor:"10,keyasint"`
Podman bool `json:"p,omitempty" cbor:"11,keyasint,omitempty"`
GpuPct float64 `json:"g,omitempty" cbor:"12,keyasint,omitempty"`
DashboardTemp float64 `json:"dt,omitempty" cbor:"13,keyasint,omitempty"`
Os Os `json:"os" cbor:"14,keyasint"`
LoadAvg1 float64 `json:"l1,omitempty" cbor:"15,keyasint,omitempty"`
LoadAvg5 float64 `json:"l5,omitempty" cbor:"16,keyasint,omitempty"`
LoadAvg15 float64 `json:"l15,omitempty" cbor:"17,keyasint,omitempty"`
BandwidthBytes uint64 `json:"bb" cbor:"18,keyasint"`
LoadAvg [3]float64 `json:"la,omitempty" cbor:"19,keyasint"`
// TODO: remove load fields in future release in favor of load avg array
}
// Final data structure to return to the hub

View File

@@ -203,9 +203,9 @@ func (rm *RecordManager) AverageSystemStats(db dbx.Builder, records RecordIds) *
sum.DiskWritePs += stats.DiskWritePs
sum.NetworkSent += stats.NetworkSent
sum.NetworkRecv += stats.NetworkRecv
sum.LoadAvg1 += stats.LoadAvg1
sum.LoadAvg5 += stats.LoadAvg5
sum.LoadAvg15 += stats.LoadAvg15
sum.LoadAvg[0] += stats.LoadAvg[0]
sum.LoadAvg[1] += stats.LoadAvg[1]
sum.LoadAvg[2] += stats.LoadAvg[2]
sum.Bandwidth[0] += stats.Bandwidth[0]
sum.Bandwidth[1] += stats.Bandwidth[1]
// Set peak values
@@ -285,9 +285,9 @@ func (rm *RecordManager) AverageSystemStats(db dbx.Builder, records RecordIds) *
sum.DiskWritePs = twoDecimals(sum.DiskWritePs / count)
sum.NetworkSent = twoDecimals(sum.NetworkSent / count)
sum.NetworkRecv = twoDecimals(sum.NetworkRecv / count)
sum.LoadAvg1 = twoDecimals(sum.LoadAvg1 / count)
sum.LoadAvg5 = twoDecimals(sum.LoadAvg5 / count)
sum.LoadAvg15 = twoDecimals(sum.LoadAvg15 / count)
sum.LoadAvg[0] = twoDecimals(sum.LoadAvg[0] / count)
sum.LoadAvg[1] = twoDecimals(sum.LoadAvg[1] / count)
sum.LoadAvg[2] = twoDecimals(sum.LoadAvg[2] / count)
sum.Bandwidth[0] = sum.Bandwidth[0] / uint64(count)
sum.Bandwidth[1] = sum.Bandwidth[1] / uint64(count)
// Average temperatures

View File

@@ -9,31 +9,30 @@ import {
xAxis,
} from "@/components/ui/chart"
import { useYAxisWidth, cn, formatShortDate, toFixedFloat, decimalString, chartMargin } from "@/lib/utils"
import { ChartData } from "@/types"
import { ChartData, SystemStats } from "@/types"
import { memo } from "react"
import { t } from "@lingui/core/macro"
export default memo(function LoadAverageChart({ chartData }: { chartData: ChartData }) {
const { yAxisWidth, updateYAxisWidth } = useYAxisWidth()
if (chartData.systemStats.length === 0) {
return null
}
const keys = {
l1: {
const keys: { legacy: keyof SystemStats; color: string; label: string }[] = [
{
legacy: "l1",
color: "hsl(271, 81%, 60%)", // Purple
label: t({ message: `1 min`, comment: "Load average" }),
},
l5: {
{
legacy: "l5",
color: "hsl(217, 91%, 60%)", // Blue
label: t({ message: `5 min`, comment: "Load average" }),
},
l15: {
{
legacy: "l15",
color: "hsl(25, 95%, 53%)", // Orange
label: t({ message: `15 min`, comment: "Load average" }),
},
}
]
return (
<div>
@@ -69,16 +68,22 @@ export default memo(function LoadAverageChart({ chartData }: { chartData: ChartD
/>
}
/>
{Object.entries(keys).map(([key, value]: [string, { color: string; label: string }]) => {
{keys.map(({ legacy, color, label }, i) => {
const dataKey = (value: { stats: SystemStats }) => {
if (chartData.agentVersion.patch < 1) {
return value.stats?.[legacy]
}
return value.stats?.la?.[i] ?? value.stats?.[legacy]
}
return (
<Line
key={key}
dataKey={`stats.${key}`}
name={value.label}
key={i}
dataKey={dataKey}
name={label}
type="monotoneX"
dot={false}
strokeWidth={1.5}
stroke={value.color}
stroke={color}
isAnimationActive={false}
/>
)

View File

@@ -26,6 +26,7 @@ import {
getHostDisplayValue,
getPbTimestamp,
listen,
parseSemVer,
toFixedFloat,
useLocalStorage,
} from "@/lib/utils"
@@ -191,6 +192,7 @@ export default function SystemDetail({ name }: { name: string }) {
chartTime,
orientation: direction === "rtl" ? "right" : "left",
...getTimeData(chartTime, lastCreated),
agentVersion: parseSemVer(system?.info?.v),
}
}, [systemStats, containerData, direction])
@@ -642,7 +644,7 @@ export default function SystemDetail({ name }: { name: string }) {
)}
{/* Load Average chart */}
{system.info.l1 !== undefined && (
{chartData.agentVersion?.minor >= 12 && (
<ChartCard
empty={dataEmpty}
grid={grid}

View File

@@ -73,6 +73,7 @@ import {
formatTemperature,
decimalString,
formatBytes,
parseSemVer,
} from "@/lib/utils"
import AlertsButton from "../alerts/alert-button"
import { $router, Link, navigate } from "../router"
@@ -224,8 +225,12 @@ export default function SystemsTable() {
{
id: "loadAverage",
accessorFn: ({ info }) => {
const { l1 = 0, l5 = 0, l15 = 0 } = info
return l1 + l5 + l15
const sum = info.la?.reduce((acc, curr) => acc + curr, 0)
// TODO: remove this in future release in favor of la array
if (!sum) {
return (info.l1 ?? 0) + (info.l5 ?? 0) + (info.l15 ?? 0)
}
return sum
},
name: () => t({ message: "Load Avg", comment: "Short label for load average" }),
size: 0,
@@ -233,16 +238,22 @@ export default function SystemsTable() {
header: sortableHeader,
cell(info: CellContext<SystemRecord, unknown>) {
const { info: sysInfo, status } = info.row.original
if (sysInfo.l1 === undefined) {
// agent version
const { minor, patch } = parseSemVer(sysInfo.v)
let loadAverages = sysInfo.la
// use legacy load averages if agent version is less than 12.1.0
if (!loadAverages || (minor === 12 && patch < 1)) {
loadAverages = [sysInfo.l1 ?? 0, sysInfo.l5 ?? 0, sysInfo.l15 ?? 0]
}
const max = Math.max(...loadAverages)
if (max === 0 && (status === "paused" || minor < 12)) {
return null
}
const { l1 = 0, l5 = 0, l15 = 0, t: cpuThreads = 1 } = sysInfo
const loadAverages = [l1, l5, l15]
function getDotColor() {
const max = Math.max(...loadAverages)
const normalized = max / cpuThreads
const normalized = max / (sysInfo.t ?? 1)
if (status !== "up") return "bg-primary/30"
if (normalized < 0.7) return "bg-green-500"
if (normalized < 1) return "bg-yellow-500"
@@ -252,7 +263,7 @@ export default function SystemsTable() {
return (
<div className="flex items-center gap-[.35em] w-full tabular-nums tracking-tight">
<span className={cn("inline-block size-2 rounded-full me-0.5", getDotColor())} />
{loadAverages.map((la, i) => (
{loadAverages?.map((la, i) => (
<span key={i}>{decimalString(la, la >= 10 ? 1 : 2)}</span>
))}
</div>

View File

@@ -9,6 +9,7 @@ import {
ChartTimeData,
ChartTimes,
FingerprintRecord,
SemVer,
SystemRecord,
UserSettings,
} from "@/types"
@@ -495,3 +496,14 @@ export function formatDuration(
.filter(Boolean)
.join(" ")
}
export const parseSemVer = (semVer = ""): SemVer => {
// if (semVer.startsWith("v")) {
// semVer = semVer.slice(1)
// }
if (semVer.includes("-")) {
semVer = semVer.slice(0, semVer.indexOf("-"))
}
const parts = semVer.split(".").map(Number)
return { major: parts?.[0] ?? 0, minor: parts?.[1] ?? 0, patch: parts?.[2] ?? 0 }
}

View File

@@ -50,6 +50,8 @@ export interface SystemInfo {
l5?: number
/** load average 15 minutes */
l15?: number
/** load average */
la?: [number, number, number]
/** operating system */
o?: string
/** uptime */
@@ -79,12 +81,15 @@ export interface SystemStats {
cpu: number
/** peak cpu */
cpum?: number
// TODO: remove these in future release in favor of la
/** load average 1 minute */
l1?: number
/** load average 5 minutes */
l5?: number
/** load average 15 minutes */
l15?: number
/** load average */
la?: [number, number, number]
/** total memory (gb) */
m: number
/** memory used (gb) */
@@ -234,7 +239,14 @@ type ChartDataContainer = {
[key: string]: key extends "created" ? never : ContainerStats
}
export interface SemVer {
major: number
minor: number
patch: number
}
export interface ChartData {
agentVersion: SemVer
systemStats: SystemStatsRecord[]
containerData: ChartDataContainer[]
orientation: "right" | "left"