Estimate fan duty from observed RPM maxima
This commit is contained in:
@@ -4,6 +4,7 @@ import (
|
||||
"context"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"math"
|
||||
"os"
|
||||
"os/exec"
|
||||
"path/filepath"
|
||||
@@ -56,13 +57,37 @@ type cachedPowerReading struct {
|
||||
UpdatedAt time.Time
|
||||
}
|
||||
|
||||
type fanObservationState struct {
|
||||
MaxRPM map[string]float64 `json:"max_rpm"`
|
||||
}
|
||||
|
||||
type fanPeakCandidate struct {
|
||||
FirstSeen time.Time
|
||||
RPM float64
|
||||
}
|
||||
|
||||
var (
|
||||
systemPowerCacheMu sync.Mutex
|
||||
systemPowerCache cachedPowerReading
|
||||
fanObservationMu sync.Mutex
|
||||
fanObservation fanObservationState
|
||||
fanObservationInit bool
|
||||
fanPeakCandidates = make(map[string]fanPeakCandidate)
|
||||
)
|
||||
|
||||
const systemPowerHoldTTL = 15 * time.Second
|
||||
|
||||
var fanObservationStatePath = "/var/log/bee-sat/fan-observation.json"
|
||||
|
||||
const fanObservationMinPeakHold = time.Second
|
||||
|
||||
func normalizeObservedFanMaxRPM(rpm float64) float64 {
|
||||
if rpm <= 0 {
|
||||
return 0
|
||||
}
|
||||
return math.Ceil(rpm/1000.0) * 1000.0
|
||||
}
|
||||
|
||||
// RunFanStressTest runs a two-phase GPU stress test while monitoring fan speeds,
|
||||
// temperatures, and power draw every second. Exports metrics.csv and fan-sensors.csv.
|
||||
// Designed to reproduce case-04 fan-speed lag and detect GPU thermal throttling.
|
||||
@@ -310,11 +335,13 @@ func sampleFanSpeeds() ([]FanReading, error) {
|
||||
out, err := exec.Command("ipmitool", "sdr", "type", "Fan").Output()
|
||||
if err == nil {
|
||||
if fans := parseFanSpeeds(string(out)); len(fans) > 0 {
|
||||
updateFanObservation(fans, time.Now())
|
||||
return fans, nil
|
||||
}
|
||||
}
|
||||
fans, sensorsErr := sampleFanSpeedsViaSensorsJSON()
|
||||
if len(fans) > 0 {
|
||||
updateFanObservation(fans, time.Now())
|
||||
return fans, nil
|
||||
}
|
||||
if err != nil {
|
||||
@@ -323,6 +350,119 @@ func sampleFanSpeeds() ([]FanReading, error) {
|
||||
return nil, sensorsErr
|
||||
}
|
||||
|
||||
func loadFanObservationLocked() {
|
||||
if fanObservationInit {
|
||||
return
|
||||
}
|
||||
fanObservationInit = true
|
||||
fanObservation.MaxRPM = make(map[string]float64)
|
||||
raw, err := os.ReadFile(fanObservationStatePath)
|
||||
if err != nil || len(raw) == 0 {
|
||||
return
|
||||
}
|
||||
var persisted fanObservationState
|
||||
if json.Unmarshal(raw, &persisted) != nil {
|
||||
return
|
||||
}
|
||||
for name, rpm := range persisted.MaxRPM {
|
||||
name = strings.TrimSpace(name)
|
||||
if name == "" || rpm <= 0 {
|
||||
continue
|
||||
}
|
||||
fanObservation.MaxRPM[name] = rpm
|
||||
}
|
||||
}
|
||||
|
||||
func saveFanObservationLocked() {
|
||||
if len(fanObservation.MaxRPM) == 0 {
|
||||
return
|
||||
}
|
||||
dir := filepath.Dir(fanObservationStatePath)
|
||||
if dir == "" || dir == "." {
|
||||
dir = "/var/log/bee-sat"
|
||||
}
|
||||
if err := os.MkdirAll(dir, 0755); err != nil {
|
||||
return
|
||||
}
|
||||
raw, err := json.MarshalIndent(fanObservation, "", " ")
|
||||
if err != nil {
|
||||
return
|
||||
}
|
||||
_ = os.WriteFile(fanObservationStatePath, raw, 0644)
|
||||
}
|
||||
|
||||
func updateFanObservation(fans []FanReading, now time.Time) {
|
||||
if len(fans) == 0 {
|
||||
return
|
||||
}
|
||||
fanObservationMu.Lock()
|
||||
defer fanObservationMu.Unlock()
|
||||
loadFanObservationLocked()
|
||||
changed := false
|
||||
for _, fan := range fans {
|
||||
name := strings.TrimSpace(fan.Name)
|
||||
if name == "" || fan.RPM <= 0 {
|
||||
continue
|
||||
}
|
||||
currentMax := fanObservation.MaxRPM[name]
|
||||
if fan.RPM <= currentMax {
|
||||
delete(fanPeakCandidates, name)
|
||||
continue
|
||||
}
|
||||
if cand, ok := fanPeakCandidates[name]; ok {
|
||||
if now.Sub(cand.FirstSeen) >= fanObservationMinPeakHold {
|
||||
newMax := math.Max(cand.RPM, fan.RPM)
|
||||
if newMax > currentMax {
|
||||
fanObservation.MaxRPM[name] = normalizeObservedFanMaxRPM(newMax)
|
||||
changed = true
|
||||
}
|
||||
delete(fanPeakCandidates, name)
|
||||
continue
|
||||
}
|
||||
if fan.RPM > cand.RPM {
|
||||
fanPeakCandidates[name] = fanPeakCandidate{FirstSeen: cand.FirstSeen, RPM: fan.RPM}
|
||||
}
|
||||
continue
|
||||
}
|
||||
fanPeakCandidates[name] = fanPeakCandidate{FirstSeen: now, RPM: fan.RPM}
|
||||
}
|
||||
if changed {
|
||||
saveFanObservationLocked()
|
||||
}
|
||||
}
|
||||
|
||||
func estimateFanDutyCyclePctFromObservation(fans []FanReading) (float64, bool) {
|
||||
if len(fans) == 0 {
|
||||
return 0, false
|
||||
}
|
||||
fanObservationMu.Lock()
|
||||
defer fanObservationMu.Unlock()
|
||||
loadFanObservationLocked()
|
||||
var samples []float64
|
||||
for _, fan := range fans {
|
||||
name := strings.TrimSpace(fan.Name)
|
||||
if name == "" || fan.RPM <= 0 {
|
||||
continue
|
||||
}
|
||||
maxRPM := fanObservation.MaxRPM[name]
|
||||
if maxRPM <= 0 {
|
||||
continue
|
||||
}
|
||||
pct := fan.RPM / maxRPM * 100.0
|
||||
if pct > 100 {
|
||||
pct = 100
|
||||
}
|
||||
if pct < 0 {
|
||||
pct = 0
|
||||
}
|
||||
samples = append(samples, pct)
|
||||
}
|
||||
if len(samples) == 0 {
|
||||
return 0, false
|
||||
}
|
||||
return benchmarkMean(samples), true
|
||||
}
|
||||
|
||||
// parseFanSpeeds parses "ipmitool sdr type Fan" output.
|
||||
// Handles two formats:
|
||||
//
|
||||
@@ -428,12 +568,27 @@ func sampleFanSpeedsViaSensorsJSON() ([]FanReading, error) {
|
||||
|
||||
// sampleFanDutyCyclePct reads fan PWM/duty-cycle controls from lm-sensors.
|
||||
// Returns the average duty cycle across all exposed PWM controls.
|
||||
func sampleFanDutyCyclePct() (float64, bool) {
|
||||
func sampleFanDutyCyclePct() (float64, bool, bool) {
|
||||
out, err := exec.Command("sensors", "-j").Output()
|
||||
if err != nil || len(out) == 0 {
|
||||
return 0, false
|
||||
fans, fanErr := sampleFanSpeeds()
|
||||
if fanErr != nil {
|
||||
return 0, false, false
|
||||
}
|
||||
return sampleFanDutyCyclePctFromFans(fans)
|
||||
}
|
||||
return parseFanDutyCyclePctSensorsJSON(out)
|
||||
pct, ok := parseFanDutyCyclePctSensorsJSON(out)
|
||||
return pct, ok, false
|
||||
}
|
||||
|
||||
func sampleFanDutyCyclePctFromFans(fans []FanReading) (float64, bool, bool) {
|
||||
if len(fans) == 0 {
|
||||
return 0, false, false
|
||||
}
|
||||
if pct, ok := estimateFanDutyCyclePctFromObservation(fans); ok {
|
||||
return pct, true, true
|
||||
}
|
||||
return 0, false, false
|
||||
}
|
||||
|
||||
func parseFanDutyCyclePctSensorsJSON(raw []byte) (float64, bool) {
|
||||
|
||||
Reference in New Issue
Block a user