没有这样的文件或目录 Kube.netes

Question

In one of my containers I am getting the error

Warning  Failed     7m52s (x4 over 8m33s)  kubelet            Error: failed to start container "xxx-xxx": Error response from daemon: OCI runtime create failed: container_linux.go:380: starting container process caused: exec: "/src/application/entrypoint.sh": stat /src/application/entrypoint.sh: no such file or directory: unknown

I can't figure out why this is an issue, as running my docker-compose file outside of kube.netes hasn't presented any issues.

The yaml section for this looks like:

spec:
  replicas: 1
  selector:
    matchLabels:
      io.kompose.service: xxx-xxx
  strategy:
    type: Recreate
  template:
    metadata:
      creationTimestamp: null
      labels:
        io.kompose.service: xxx-xxx
    spec:
      containers:
      - env:
        - name: BEARER_TOKEN
          valueFrom:
            configMapKeyRef:
              key: BEARER_TOKEN
              name: xxx-xxx-env-dev
        - name: PGADMIN_DEFAULT_EMAIL
          valueFrom:
            configMapKeyRef:
              key: PGADMIN_DEFAULT_EMAIL
              name: xxx-xxx-env-dev
        - name: PGADMIN_DEFAULT_PASSWORD
          valueFrom:
            configMapKeyRef:
              key: PGADMIN_DEFAULT_PASSWORD
              name: xxx-xxx-env-dev
        - name: PGADMIN_LISTEN_PORT
          valueFrom:
            configMapKeyRef:
              key: PGADMIN_LISTEN_PORT
              name: xxx-xxx-env-dev
        - name: POSTGRES_DB
          valueFrom:
            configMapKeyRef:
              key: POSTGRES_DB
              name: xxx-xxx-env-dev
        - name: POSTGRES_HOST_NAME
          valueFrom:
            configMapKeyRef:
              key: POSTGRES_HOST_NAME
              name: xxx-xxx-env-dev
        - name: POSTGRES_PASSWORD
          valueFrom:
            configMapKeyRef:
              key: POSTGRES_PASSWORD
              name: xxx-xxx-env-dev
        - name: POSTGRES_PORT
          valueFrom:
            configMapKeyRef:
              key: POSTGRES_PORT
              name: xxx-xxx-env-dev
        - name: POSTGRES_USER
          valueFrom:
            configMapKeyRef:
              key: POSTGRES_USER
              name: xxx-xxx-env-dev
        image: x/xx-xxx-xxx:prod
        name: xxx-xxx
        ports:
        - containerPort: 5000
        resources: {}
        volumeMounts:
        - mountPath: /src/application
          name: xxx-xxx-claim0
      restartPolicy: Always
      volumes:
      - name: xxx-xxx-claim0
        persistentVolumeClaim:
          claimName: xxx-xxx-claim0

It seems kube.netes can't locate the entrypoint file, what can I do to make it accessible? Is this persisten volume related? Sorry but I am a beginner at Kube.netes.

Answer 1

In docker-compose was the entrypoint.sh mounted from the host, eg $PWD/app? If yes, this setup would be hard to replicate with kube.netes, kompose does have an option to use hostPath but this would work only for a local dev setup like minikube or docker desktop.

For a dev workflow which works well with kube.netes tools like https://skaffold.dev/ or https://devspace.sh/ can be used, both have a file sync feature so that dev artifacts are synced inside the container, as it is said in some comments the entrypoint should not stay in a volume, it shuld be part of the image and the dev tools should allow the local copy to be deployed quickly during the dev+test cycles.

Another option is to keep docker-compose for dev and use k8s/helm for production, although I am not a big fan of this approach since it will add a big overhead when troubleshooting prod.

To workaround the error and be able to check the contents of the volume from inside the container add command: [ "/bin/bash", "-c", "sleep 1000000" ] to your pod definition, then use kubectl exec to get an interactive shell.