ClusterCockpit/cc-metric-collector
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-metric-collector.git synced 2025-04-06 13:35:55 +02:00
Code Issues Packages Projects Releases Wiki Activity

store GPU index explicitly

Browse Source
This commit is contained in:
Thomas Roehl 2022-05-23 16:29:06 +02:00
parent a8fba84dfd
commit 5d2628beb6
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
collectors/rocmsmiMetric.go
View File

@ -21,6 +21,7 @@ type RocmSmiCollectorConfig struct {
type RocmSmiCollectorDevice struct { type RocmSmiCollectorDevice struct {
device rocm_smi.DeviceHandle device rocm_smi.DeviceHandle
index int
tags map[string]string // default tags tags map[string]string // default tags
meta map[string]string // default meta information meta map[string]string // default meta information
excludeMetrics map[string]bool // copy of exclude metrics from config excludeMetrics map[string]bool // copy of exclude metrics from config
@ -150,6 +151,7 @@ func (m *RocmSmiCollector) Init(config json.RawMessage) error {
for _, e := range m.config.ExcludeMetrics { for _, e := range m.config.ExcludeMetrics {
dev.excludeMetrics[e] = true dev.excludeMetrics[e] = true
} }
dev.index = i
m.devices = append(m.devices, dev) m.devices = append(m.devices, dev)
} }
@ -167,7 +169,7 @@ func (m *RocmSmiCollector) Read(interval time.Duration, output chan lp.CCMetric)
for _, dev := range m.devices { for _, dev := range m.devices {
metrics, ret := rocm_smi.DeviceGetMetrics(dev.device) metrics, ret := rocm_smi.DeviceGetMetrics(dev.device)
if ret != rocm_smi.STATUS_SUCCESS { if ret != rocm_smi.STATUS_SUCCESS {
cclog.ComponentError(m.name, "Unable to get metrics for device at index", dev.device.Index, ":", rocm_smi.StatusStringNoError(ret)) cclog.ComponentError(m.name, "Unable to get metrics for device at index", dev.index, ":", rocm_smi.StatusStringNoError(ret))
continue continue
} }