Docker Interview Question 4: PID1 Signal Handling
Docker Interview Questions · Test 1 · question 4 of 10
A Node.js API registers a SIGTERM handler that closes database connections and logs "shutting down". Its Dockerfile ends with the shell-form instruction ENTRYPOINT node server.js.
Every docker stop takes about 10 seconds, the "shutting down" message is never logged, and docker exec api ps shows PID 1 is /bin/sh -c node server.js while node has a different PID. You need the container to shut down gracefully and promptly on docker stop. What should you change?
Show answer and explanation
Answer:
docker stop sends SIGTERM to the container's main process (PID 1) and, after a grace period, SIGKILL. The default grace period is 10 seconds for Linux containers, which is the delay being seen. The shell form of ENTRYPOINT starts the command as a subcommand of /bin/sh -c, which does not pass signals, so the executable is not PID 1 and never receives the SIGTERM. The shell ignores it, the timeout expires, and the container is killed without the handler running. The exec form runs node directly as PID 1, so it receives SIGTERM and can exit cleanly at once.
Option A only lengthens the wait before SIGKILL. node still never gets the signal, so shutdown becomes slower and is still not graceful.
Option B makes the stop fast but kills the process without letting the handler run, the opposite of a graceful shutdown.
Option D is a restart policy. It controls whether a container is restarted, not how signals reach the process.
References: Dockerfile reference - ENTRYPOINT; docker container stop