Should I Run Plain Docker Compose in Production in 2026?

· ai cloud · Source ↗

TLDR

  • Docker Compose works in production in 2026 if you manually close five gaps: orphan cleanup, disk/log limits, health-check restarts, image digest pinning, and socket security.

Key Takeaways

  • Pass --remove-orphans on every docker compose up or removed services keep running on old ports consuming RAM indefinitely.
  • Cap logs in /etc/docker/daemon.json with max-size: 10m and max-file: 3; unbounded json-file logs are a top cause of disk-full outages.
  • Docker Engine reports unhealthy containers but does not restart them; add willfarrell/docker-autoheal or move to Swarm for actual self-healing.
  • Pin images by digest (myapp@sha256:...) not tag; two hosts pulling the same tag minutes apart can run different code after a re-push.
  • Mounting /var/run/docker.sock grants effective host root; use rootless Docker or docker-socket-proxy to limit blast radius.

Hacker News Comment Review

  • Discussion is thin and mostly humorous; commenters signal that the real answer is always context-dependent rather than a hard yes/no.
  • No substantive technical disagreement with the article’s recommendations surfaced in the thread.

Notable Comments

  • @TheChaplain: log-size capping was a genuine TIL, suggesting this operational detail is underknown even among practitioners.

Original | Discuss on HN