Add container_id to metric tags

DataDog · Sep 30, 2024 · 78c1d24 · 78c1d24
1 parent 2d129c1
commit 78c1d24
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 3 deletions.
diff --git a/pkg/collector/corechecks/gpu/gpu.go b/pkg/collector/corechecks/gpu/gpu.go
@@ -110,7 +110,7 @@ func (m *Check) ensureInitialized() error {
 }
 
 // ensureProcessor ensures that there is a stats processor for the given key
-func (m *Check) ensureProcessor(key *model.StreamKey, snd sender.Sender, gpuThreads int, checkDuration time.Duration) {
+func (m *Check) ensureProcessor(key *model.StreamKey, snd sender.Sender, gpuThreads int, checkDuration time.Duration, metadata *model.StreamMetadata) {
 	if _, ok := m.statProcessors[key.Pid]; !ok {
 		m.statProcessors[key.Pid] = &StatsProcessor{
 			key: key,
@@ -123,6 +123,7 @@ func (m *Check) ensureProcessor(key *model.StreamKey, snd sender.Sender, gpuThre
 	m.statProcessors[key.Pid].measuredInterval = checkDuration
 	m.statProcessors[key.Pid].timeResolver = m.timeResolver
 	m.statProcessors[key.Pid].lastCheck = m.lastCheckTime
+	m.statProcessors[key.Pid].metadata = metadata
 }
 
 // Run executes the check
@@ -167,13 +168,13 @@ func (m *Check) Run() error {
 	usedProcessors := make(map[uint32]bool)
 
 	for _, data := range stats.CurrentData {
-		m.ensureProcessor(&data.Key, snd, gpuThreads, checkDuration)
+		m.ensureProcessor(&data.Key, snd, gpuThreads, checkDuration, &data.Metadata)
 		m.statProcessors[data.Key.Pid].processCurrentData(data)
 		usedProcessors[data.Key.Pid] = true
 	}
 
 	for _, data := range stats.PastData {
-		m.ensureProcessor(&data.Key, snd, gpuThreads, checkDuration)
+		m.ensureProcessor(&data.Key, snd, gpuThreads, checkDuration, &data.Metadata)
 		m.statProcessors[data.Key.Pid].processPastData(data)
 		usedProcessors[data.Key.Pid] = true
 	}

diff --git a/pkg/collector/corechecks/gpu/stats.go b/pkg/collector/corechecks/gpu/stats.go
@@ -65,6 +65,9 @@ type StatsProcessor struct {
 	// utilizationNormFactor is the factor to normalize the utilization by, to account for the fact that we might have more kernels enqueued than the GPU can run in parallel. This factor
 	// allows distributing the utilization over all the streams that are enqueued
 	utilizationNormFactor float64
+
+	// Metadata associated to this stream
+	metadata *model.StreamMetadata
 }
 
 // processKernelSpan processes a kernel span
@@ -127,6 +130,7 @@ func (sp *StatsProcessor) processCurrentData(data *model.StreamData) {
 func (sp *StatsProcessor) getTags() []string {
 	return []string{
 		fmt.Sprintf("pid:%d", sp.key.Pid),
+		fmt.Sprintf("container_id:%s", sp.metadata.ContainerID),
 	}
 }