Kubernetes Monitoring

The Kubernetes collector adds cluster context — deployments, pods, events, and resource requests/limits — on top of the container metrics the agent already reports, giving you a complete picture from node to request.

Deploying the collector#

helm install applicare-agent applicare/agent \
  --namespace applicare \
  --set kubernetes.enabled=true \
  --set server=https://applicare.yourco.com \
  --set token=ARC-ENROLL-XXXXXXXX

The DaemonSet covers every node; the cluster collector watches the API server for objects and events.

Workload health#

See deployments, statefulsets, and pods with restart counts, OOMKills, pending scheduling, and requests-vs-limits pressure. Pods map to the services they run, so a CrashLoopBackOff is shown next to the latency it causes upstream.

From symptom to cause#

When a pod misbehaves, ArcIn correlates the Kubernetes event (eviction, image pull error, node pressure) with the service degradation, so you skip the manual kubectl archaeology. For the full incident workflow see Kubernetes solutions.