PostgreSQL Question posted in
The official documentation can be found here.

Postgresql – Kill remaining containers in cronjob pod after default node has completed

I want to verify the data in some Postgres dumps.
I’ve created a script that imports the dump into a fresh Postgres DB, and verifies the data.

This runs as a cronjob in Kubernetes.

The Cronjob contains two containers. The default container, running the script, and a postgresql container that is the database.

  jobTemplate:
    spec:
      template:
        spec:
          containers:
            - name: verifier
              image: "verifier:v1.0.2"
              env:
              # ...
            - name: postgres
              image: postgres:16.1
              env:
                - name: POSTGRES_PASSWORD
                  value: test

This executes fine, but when the verifier script is done it is killed. While the postgresql container keeps running.

The pod remains in state Not Ready

verifier-28577575-zq24s                1/2     NotReady    0               57m

I’ve tried adding activeDeadlineSeconds: 720, but it does not seem to trigger.

Is there a way to take down the remaining containers when the default container has completed execution?

Tags:

2

Answers


  1. Chosen as BEST ANSWER
    0

    This was resolved using a liveness probe. Using a shared volume between the containers, and having the postgresql container using cat file as livenessprobe. The verifier script creates a file, and then removes it when it is done executing. The postgresql container will then die when doing the next check:

    jobTemplate:
    spec:
      template:
        spec:
          containers:
            - name: verifier
              image: "verifier:v1.0.2"
              volumeMounts: 
                - name: shared-liveness-volume
                  mountPath: /shared
              env:
              # ...
            - name: postgres
              image: postgres:16.1
              env:
                - name: POSTGRES_PASSWORD
                  value: test
              volumeMounts: 
                - name: shared-liveness-volume
                  mountPath: /shared
              livenessProbe:
                exec: 
                  command: 
                    - cat
                    - /shared/keepalive
                  initialDelaySeconds: 5
                  periodSeconds: 10
          volumes:
            - name: shared-liveness-volume
              emptyDir: {}

    verifier.sh script

    #!/usr/bin/env bash
touch /shared/keepalive
# verify logic
rm /shared/keepalive

  2. 0

    To my knowldege, there is no such way. We are talking of a single pod. As long as one container is running, the pod is considered active. What I would do to solve this problem is to have my verifier image be based on postgres:16.1 and run my verification script there and shutdown Postgres after my checks.

    Actually, you do not even have to that, since you can simply mount your *.sql[.gz] into /docker-entrypoint-initdb.d – if something goes wrong, the container will fail. Using prometheus and the Job’s status, you can do the alerting.

    apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:
      volumes:
        - name: dumps
          # Obviously, you might want to use something more suitable
          # like https://github.com/yandex-cloud/k8s-csi-s3
          configMap:
            name: dumps
      containers:
      - name: pg
        image: postgres:16.1
      volumeMounts:
      - mountPath: /docker-entrypoint-initdb.d/dump.sql
        name: dumps
        subPath: theactualDump.sql
      restartPolicy: Never
    Login or Signup to reply.
Please signup or login to give your own answer.

Back To Top
Search