Cleanup #

Stack updates and deployments often leave older images, and sometimes old volumes behind. In normal operation, that isn’t a problem - occasional manual cleanup using the Quantum Console is more than enough.

**Please note:** Docker Swarm rollbacks don't work if the old images got deleted. And even for manual rollbacks, it's easier if the images stay on the endpoint for a while. So don't run automatic cleanups if you don't absolutely need it or add a sufficient delay to them.

If you use apps that start and stop a lot of docker containers (like CI/CD systems), you should consult the systems documentation first, but otherwise consider automatic cleanup, using a periodically-running container that runs docker system prune.

Example #

Use the following stack as an example: prune images from your nodes every 72 hours (72h). The interval may seem arbitary, but we suggest you make to make it longer than a weekend — fine tune to your needs.

version: "3.8"

services:
  system-prune:
    image: docker:20.10.23
    command: docker system prune --force
    deploy:
      mode: global
      replicas: 0
      restart_policy:
        delay: 48h
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:rw