ClusterCockpit/cc-metric-collector
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-metric-collector.git synced 2026-05-12 00:17:28 +02:00
Code Issues Packages Projects Releases Wiki Activity

Add metric 'nv_util_eff' like nvtop

Browse Source
This commit is contained in:
Thomas Roehl
2026-05-06 18:57:58 +02:00
parent 8e58072ff6
commit 0a1c7a9911
2 changed files with 31 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
collectors/nvidiaMetric.go
View File

@@ -1115,6 +1115,31 @@ func readNVLinkStats(device *NvidiaCollectorDevice, output chan lp.CCMessage) er
return nil
}
func readEfficiency(device *NvidiaCollectorDevice, output chan lp.CCMessage) error {
if !device.excludeMetrics["nv_util_eff"] {
maxPower, ret := nvml.DeviceGetEnforcedPowerLimit(device.device)
if ret == nvml.SUCCESS {
curPower, ret := nvml.DeviceGetPowerUsage(device.device)
if ret == nvml.SUCCESS {
util, ret := nvml.DeviceGetUtilizationRates(device.device)
if ret == nvml.SUCCESS {
factor := float64(curPower) / float64(maxPower)
eff := uint32(float64(util.Gpu) * factor)
if eff > 100 {
eff = 100
}
y, err := lp.NewMetric("nv_util_eff", device.tags, device.meta, eff, time.Now())
if err == nil {
y.AddTag("unit", "percent")
output <- y
}
}
}
}
}
return nil
}
func (m *NvidiaCollector) Read(interval time.Duration, output chan lp.CCMessage) {
var err error
if !m.init {
@@ -1220,6 +1245,11 @@ func (m *NvidiaCollector) Read(interval time.Duration, output chan lp.CCMessage)
if err != nil {
cclog.ComponentDebug(m.name, "readNVLinkStats for device", name, "failed")
}
err = readEfficiency(device, output)
if err != nil {
cclog.ComponentDebug(m.name, "readEfficiency for device", name, "failed")
}
}
// Actual read loop over all attached Nvidia GPUs