在 kubernetes 中超时 pod/job

Question

Trying to use timeout and bash as entrypoint is not working as expected. The pod keeps running even after the specified timeout. Intention is to have a bash -c entry point for which I can define a timeout, after which, the pod enters an "Error" state. Once this is working, this will be used to timeout jobs like backups/cleaning that could get stuck forever otherwise.

apiVersion: v1
kind: Pod
metadata:
  labels:
    run: redis-cli
  name: redis-cli
spec:
  containers:
  - name: redis-cli
    image: xxxx/redis-cli:0.0.1
    imagePullPolicy: Always
    env:
      - name: REDIS_SERVICE
        value: redis.svc
      - name: REDIS_PORT
        value: "6666"
    command: ['/usr/bin/timeout', '20', '/bin/bash', '-c' ]
    args:
      - sleep 6000
    resources:
      limits:
        memory: "12Gi"
        cpu: "2000m"
  dnsPolicy: ClusterFirst
  restartPolicy: Never

Answer 1

You can achieve the timeout using spec.activeDeadlineSeconds.

spec:
  activeDeadlineSeconds: 10
  containers:
  - command:
    - /bin/sh
    - -c
    - sleep 6000
    image: busybox