note:

Chaos mesh does not currently support simulation injection of naked pods. It only supports some specific pods, such as deployment, statefulset, and daemonset.

Pod chaos allows you to simulate pod faults, specifically pod failure and pod kill. pod failure can be used to simulate a situation where a pod is down. In this case, the pod will be unavailable for a long time.

  • Pod Failure action periodically injects errors to pods. And it will cause the pod to not be created for a while. In another word, the selected pod will be unavailable in a specified period of time.

  • Pod Kill action kills the specified pod (ReplicaSet or something similar may be needed to ensure the pod will be restarted).

Pod Failure Action

Below is a sample pod failure configuration file:

  1. apiVersion: pingcap.com/v1alpha1
  2. kind: PodChaos
  3. metadata:
  4.   name: pod-failure-example
  5.   namespace: chaos-testing
  6. spec:
  7.   action: pod-failure
  8.   mode: one
  9.   duration: "30s"
  10.   selector:
  11.     namespaces:
  12.       - tidb-cluster-demo
  13.     labelSelectors:
  14.       "app.kubernetes.io/component": "tikv"
  15.   scheduler:
  16.     cron: "@every 2m"

For more sample files, see examples. You can edit them as needed.

Description:

  • action defines the specific chaos action for the pod. In this case, it is pod failure.
  • mode defines the mode to run chaos action. Supported mode: one / all / fixed / fixed-percent / random-max-percent.
  • duration defines the duration for each chaos experiment. The value of the Duration field is 30s, which indicates that pod failure will last 30 seconds.
  • selector is used to select pods that are used to inject chaos actions.
  • scheduler defines the scheduler rules for the running time of the chaos experiment. For more rule information, see https://godoc.org/github.com/robfig/cron.

Pod Kill Action

Note:

The detailed description of each field in the configuration template are consistent with that in Pod Failure.

Below is a sample pod kill configuration file:

  1. apiVersion: pingcap.com/v1alpha1
  2. kind: PodChaos
  3. metadata:
  4.   name: pod-kill-example
  5.   namespace: chaos-testing
  6. spec:
  7.   action: pod-kill
  8.   mode: one
  9.   selector:
  10.     namespaces:
  11.       - tidb-cluster-demo
  12.     labelSelectors:
  13.       "app.kubernetes.io/component": "tikv"
  14.   scheduler:
  15.     cron: "@every 1m"

For more sample files, see examples. You can edit them as needed.