k8sclusterreceiver: add init/ephemeral container metrics

[jaehanbyun]

Description

Collect k8s.container metrics for sidecar containers and ephemeral containers.

Sidecar containers (Kubernetes 1.29+): Init containers with restartPolicy: Always are now recognized as sidecars and their metrics are collected alongside regular containers.

Ephemeral containers: Debug containers added via kubectl debug are now included in metrics collection.

Regular init containers (without restartPolicy: Always) are intentionally NOT collected, as they are short-lived and complete before the main containers start.

Link to tracking issue

Fixes #42571

Testing

Unit Tests Added:

• TestSidecarMetrics: Verifies sidecar container metrics are collected while regular init containers are not
• TestEphemeralContainerMetrics: Verifies ephemeral container metrics are collected

Manual Testing with Kind cluster:

Test Pod:

apiVersion: v1
kind: Pod
metadata:
  name: sidecar-test-pod
spec:
  containers:
    - name: main-app
      image: busybox:latest
      resources:
        limits:
          cpu: "100m"
          memory: "64Mi"
  initContainers:
    - name: my-sidecar
      image: busybox:latest
      restartPolicy: Always  # This makes it a sidecar
      resources:
        limits:
          cpu: "50m"
          memory: "32Mi"
    - name: regular-init  # No restartPolicy - regular init container
      image: busybox:latest

Sidecar Container Metrics Output:

Resource attributes:
     -> k8s.pod.name: Str(sidecar-test-pod)
     -> k8s.container.name: Str(my-sidecar)
     -> container.image.name: Str(docker.io/library/busybox)
Metrics:
     -> k8s.container.cpu_limit: 0.050000
     -> k8s.container.cpu_request: 0.050000
     -> k8s.container.memory_limit: 33554432
     -> k8s.container.memory_request: 33554432
     -> k8s.container.ready: 1
     -> k8s.container.restarts: 0

Ephemeral Container (added via kubectl debug):

kubectl debug sidecar-test-pod -it --image=busybox:latest --target=main-app

Resource attributes:
     -> k8s.pod.name: Str(sidecar-test-pod)
     -> k8s.container.name: Str(debugger-4sw76)
     -> container.image.name: Str(docker.io/library/busybox)
Metrics:
     -> k8s.container.ready: 0
     -> k8s.container.restarts: 0

Documentation

Changelog entry added in .chloggen/collect-sidecar-metrics.yaml

[github-actions]

This PR was marked stale due to lack of activity. It will be closed in 14 days.

[ChrsMark]

Since the tittle states [WIP] I've converted this to draft. Feel free to mark it ready for review when it's not WIP anymore.

[jaehanbyun]

@ChrsMark Thanks! I’ll do that once it’s no longer WIP.

[github-actions]

This PR was marked stale due to lack of activity. It will be closed in 14 days.

[jaehanbyun]

@ChrsMark I'd like to discuss the current implementation.

Init Container Metrics Scope

The current implementation only collects metrics for sidecar containers (init containers with restartPolicy: Always), not regular init containers.

Rationale:

• Regular init containers are short-lived and complete before main containers start
• Sidecar containers (K8s 1.29+) run alongside main containers and benefit from monitoring
• Aligns with the original issue [receiver/k8scluster] Collect k8s.container metrics for sidecars #42571 which specifically mentions sidecars

Should we also collect metrics for all init containers, or is the sidecar-only approach preferred?

[ChrsMark]

The current implementation only collects metrics for sidecar containers (init containers with restartPolicy: Always), not regular init containers.

Hmm, I wonder if that's a fair assumption though. It could be that some init containers turned out to be problematic and hence we want to monitor them 🤔 .

I also wonder if we should make the collection of these data configurable? @povilasv @dmitryax @TylerHelmuth WDYT?