How to Catch Docker OOM Exit 137 in Node.js Containers

The container stops. docker inspect shows "OOMKilled": true. The process status is 137, which is 128 plus SIGKILL (9). The Node process does not get an uncaughtException handler call. SIGKILL is not catchable.

The terminal often looks like a hang, then nothing. Charts for the API can stay green because the request never reached the server.

The cgroup memory limit is lower than the Node heap plus native memory. A login or index step loads a large payload, V8 grows the old space, and the kernel OOM killer sends SIGKILL. Exit code 137 is the shell's report of that signal. Raising --max-old-space-size past the container limit makes the kill more likely, not less.

• The exact argv, including flags like --env staging.
• Exit code 137 and the OOM classification.
• The stack you still have from the last log lines, if the runtime flushed them.
• One note from the person who ran it: what they were trying to do.

Install nothing global. The command below prints the CLI wrap, the Express middleware, and the Next.js instrumentation hook.

bashnpx --yes @aethellayer/sdk init

Wrap the failing command so stdout and stderr are kept when the exit code is fatal:

bashnpx --yes @aethellayer/sdk run -- node server.js

Exit 137 is classified as ERR_DOCKER_OOM. AethelLayer writes a local log and, when it can see a GitHub remote, prints a draft issue URL. It does not file the issue. You press send.

Set the memory limit above the heap you actually need, or lower the heap so it dies with a catchable V8 error instead of SIGKILL. Keep the probe on the CLI so the next 137 still leaves a draft.