Many zombie processes and they are all mongosh from Docker

Question

Logging into my Linux server which hosts many Docker containers, I saw a warning that "There are 320 zombie processes." That seems bad! You can see which processes are involved using:

ps axo stat,ppid,pid,comm | grep -w defunct

They are all mongosh and that only comes from one of the docker containers running the MongoDB server. What is wrong here? And are these a problem?

Answer 1

The zombie processes exist because they were killed and some other process is still waiting for info from them. Not a big deal but probably unwanted.

My server is running via Docker Compose. In the compose file I had a health check which is using mongosh to ping MongoDB internally. The timing of that was too short and it was being killed:

healthcheck:
      test: "mongosh ... --eval 'quit(db.runCommand({ ping: 1 }).ok ? 0 : 1)' || exit 1"
      interval: 30s
      retries: 5
      start_period: 1s  # <-- That start_period is the problem
      timeout: 10s

Setting the start_period to 10s and restarting the container fixed everything. Zombie apocalypse averted!