What to save before rerunning a failed GitHub Actions job

Before rerunning a failed GitHub Actions job, save enough evidence to know what changed after the rerun. The minimum packet is run URL, workflow name, job name, commit SHA, runner image, matrix values, failed step, first failing log line, and any artifacts or test reports produced before the failure. If the rerun passes, this packet is what keeps the original failure from disappearing into memory.

GitHub documentation says completed workflow runs can show status, logs, artifacts, and rerun options. It also describes rerunning all jobs, failed jobs, or a specific job, and the GitHub CLI can rerun with debug logging. The practical point is sequence: inspect and save the first failure, then rerun with a clear question.

A useful first-failure note has three sections. “Context” records branch, commit, workflow, runner, and matrix. “Failure” records the failed step and the earliest log line that explains a real problem rather than a later cascade. “Next check” says why the rerun is being attempted: flaky network, missing cache, stale artifact, changed secret, test-order suspicion, or more diagnostic logs.

Avoid treating a green rerun as proof that nothing happened. A flaky test, expired dependency cache, rate limit, or runner image difference can pass on the second try while still costing time later. If the job is important enough to rerun, it is important enough to leave a small trail.

What to save before rerunning a failed GitHub Actions job

// COMMENTS

ON THIS PAGE