K8s cluster monitoring

Tools or services used for monitoring

Probes: actively monitor the health of a container. If the probe determines that a container is no longer healthy, the probe will restart it

cAdvisor: A container resource usage and performance analysis agent

Kube-state-matrics: A service listens on Kubernetes API server to generate the metrics about the state of objects like Deployments, Nodes, Pods.

Metrics-server: A replacement of Heapster which generates the cluster-wide resource usage. CPU, file descriptors, memory, request latencies, etc.

Aspects of cluster monitoring:

  1. Node level metrics: CPU, Load, Disk, Memory, Network, Pod resource, etc

  2. Orchestration level metrics: deployment, pod, replica status, etc

  3. Internal kube system components: detailed service metrics for the scheduler, controller manager, dns service, etc

Kubernetes monitoring architecture: details

Last updated