Exit Codes

Exit Codes

Reading a Crashed Container's Exit Code and Logs

Three facts explain almost every stopped container: its exit code, whether the kernel killed it for memory, and the last lines it logged. All three survive the crash until someone runs docker rm (or --rm was set). Crash the API with a database host that does not exist, then provoke Docker 514 's own exit codes:

A crashed API's last words, and Docker's own exit codesShell

docker run -d --name l3-crash -e PGHOST=nowhere l3-booknest-api:latest >/dev/null; sleep 4
docker ps -a --filter name=l3-crash --filter status=exited --format '{{.Names}}  {{.Status}}'
docker inspect l3-crash --format 'exit={{.State.ExitCode}} oom={{.State.OOMKilled}}'
docker logs --tail 20 l3-crash 2>&1 | head -1; docker rm l3-crash >/dev/null
docker run --rm --nosuchflag alpine:3 2>/dev/null; echo "exit=$?"
docker run --rm alpine:3 /etc 2>/dev/null; echo "exit=$?"
docker run --rm alpine:3 nosuchcmd 2>/dev/null; echo "exit=$?"

Output

l3-crash  Exited (1) 3 seconds ago
exit=1 oom=false
Failed to start BookNest API Error: getaddrinfo ENOTFOUND nowhere
exit=125
exit=126
exit=127

Exit code 1 with oom=false means the application chose to exit, and the log says why. Codes above 128 mean a signal killed the process (the code minus 128 is the signal's number), and 125-127 come from Docker itself:

Container exit codes and what they usually mean
Exit code	Meaning	Typical cause
1	Application error	Uncaught exception, failed startup
125	Docker could not run the container	Bad flag, missing image, unreachable log driver
126 / 127	Command not executable / not found	A directory, wrong architecture / a typo in CMD
137 (128 + 9)	SIGKILL	OOM killer, docker kill, stop timeout
143 (128 + 15)	SIGTERM	docker stop behind --init

docker stop sends SIGTERM and sends SIGKILL ten seconds later. The kernel ignores signals that a PID 1 has not installed a handler for, so a program that never handles SIGTERM sits out the grace period and dies with 137. BookNest's API handles it (Exec vs Shell Form); for programs that do not, docker run --init puts tini 11,240 in front. A restart policy hides a crash: the container runs again, but RestartCount has grown and docker logs still holds the crashed run's lines, so read them before you recreate the container.

MENU