Logging and Monitoring

2 min read

Logging and monitoring

How do you monitor resource consumption in Kubernetes? or more importantly, what would you like to monitor?

image-20210810145340897

  • Cluster level

    • Node level
      • How many node
      • Health of the nodes
      • Performance metrics usage
        • CPU
        • Memory
        • RAM
        • Disk

  • Pod level

    • No. pods
    • Health of pods
    • performance usage
$ k top node
$ k top pod

Heapster vs metrics server

Heapster is now deprecated and a slimmed down version was formed known as the metrics server. It is only an in memory solution. You cannot see historical data.

image-20210810145407532

How are the metrics generated for the PODs on these nodes? Kubelet contains a sub component called “cAdvisor”, which is responsible for sending retrieved performance metrics from pods to metrics server.

image-20210810145437750

image-20210810145558271

Clone the metric server from github repo

$ git clone https://github.com/Kubernetes-incubator/metrics-server.git

Deploy the metric server

$ kubectl create -f metric-server/deploy/1.8+/

View the cluster performance

$ kubectl top node

View performance metrics of pod

$ kubectl top pod

image-20210810145640861

Logs

In Docker, logs can be accessed using the docker logs command, while in Kubernetes, logs are viewed with the kubectl logs command. For pods with multiple containers, it’s necessary to specify which container’s logs you want to see.

image-20210810151339375

image-20210810151350195

image-20210810151427136

To view the logs:

$ kubectl logs -f event-simulator-pod

If there are multiple containers in a pod then you must specify the name of the container explicitly in the command:

$ kubectl logs -f <pod-name> <container-name>
$ kubectl logs -f even-simulator-pod event-simulator

Monitoring Cluster Events

  • kubectl get events: This command gives an overview of events happening across the entire cluster.
  • kubectl get events -o wide: This command provides a more detailed overview of cluster-wide events.

Use kubectl get events when it’s unclear which resource an error relates to. If you know the specific resource involved, it’s better to use kubectl describe [resource] on that resource for more targeted information.

Graph View

Backlinks