Process Completed With Exit Code 143: SIGTERM Fix
Last updated
Process completed with exit code 143 means a process was asked to stop with SIGTERM. The number is 128 plus 15, and 15 is the SIGTERM signal. Unlike exit code 137, which is a forced kill, SIGTERM is a polite request that the process can catch and clean up after. In GitHub Actions, something told your step to stop: usually a cancelled run, a timeout, or a runner shutdown.
The exact message
Error: Process completed with exit code 143.
You may see Terminated printed just above it, and often The operation was canceled near the end of the log.
Causes, most likely first
1. The run was cancelled
Someone cancelled the run, or a workflow concurrency group with cancel-in-progress: true cancelled it because a newer run started. The step that was running at that moment ends with 143.
concurrency:
group: ci-${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
This is intended behaviour for pull request builds, but dangerous on deploy workflows. Do not cancel a deploy half way: use cancel-in-progress: false there. See The operation was canceled.
2. A job or step hit timeout-minutes
When a job or step exceeds its timeout-minutes, GitHub terminates it. Check the job's duration against the limit you set, and see timeout-minutes.
3. The runner is shutting down
On self-hosted runners, a stopped service, a reboot or a scale-down sends SIGTERM to the job. See The runner has received a shutdown signal.
4. docker stop or a container being replaced
docker stop sends SIGTERM first, and force-kills after a grace period (10 seconds by default). A container whose main process ignores SIGTERM then ends with 137, and one that handles it ends with 143 or 0.
5. Your own script sent it
kill, pkill and timeout send SIGTERM by default. A pkill node that matches more than you intended will stop the build tool as well.
Handle SIGTERM in your own scripts
To clean up when the job is stopped, trap the signal:
- run: |
cleanup() { echo "stopping"; docker compose down; }
trap cleanup TERM INT
./long-running-task.sh &
wait $!
For steps that must run even when the job is cancelled, use if: always() or if: cancelled():
- name: Stop services
if: always()
run: docker compose down
A cancelled job has a short window to run these steps, so keep cleanup fast.
How to prevent it
- Use
cancel-in-progressonly on workflows that are safe to interrupt. - Set
timeout-minuteswith headroom above the normal run time. - Make long tasks resumable, and trap SIGTERM to release locks and containers.
- On self-hosted runners, drain before restarts.
- Be specific in
pkillandkillcommands.
FAQ
What does exit code 143 mean?
The process received SIGTERM (128 + 15). It was asked to terminate, and it did.
Is exit code 143 an error in my code?
Usually not. It points to something outside the step stopping it. Check cancellation, timeouts and the runner first.
What is the difference between 137 and 143?
137 is SIGKILL, which cannot be caught and is typical of out-of-memory kills. 143 is SIGTERM, which a process can catch and handle.
Why did my passing job end in 143?
If a later step failed and the workflow cancelled the remaining work, or a sibling job in a matrix with fail-fast: true failed, running jobs are cancelled and can end in 143. Set fail-fast: false if you want them to finish.
Related
- Process Completed With Exit Code 1: How to DebugProcess completed with exit code 1" in GitHub Actions means a command failed. How to find the real error above it, debug it and fix common causes.
- Process Completed With Exit Code 127: Command Not FoundFix "Process completed with exit code 127" in GitHub Actions: command not found. Missing tools, PATH between steps, bad script paths and CRLF line endings.
- The Process '/usr/bin/git' Failed With Exit Code 128Fix "The process '/usr/bin/git' failed with exit code 128" in GitHub Actions: auth, missing refs, shallow clones and dubious ownership, with fixes.
- Exit Code 137 in GitHub Actions: Out-of-Memory FixProcess completed with exit code 137 in GitHub Actions means SIGKILL, usually out of memory. Fixes for Node, Jest, Gradle and Docker, plus runner RAM.