Start with the existing run
Open the repository's Actions tab, choose the failed workflow execution and select the failed job. Expand the first failed step and read the lines before the final process exit message. A generic exit code describes the result, not the underlying cause. Reading existing logs is preferable to triggering another run merely to obtain an error.
- Keep the workflow, job, run ID and attempt number.
- For a matrix job, keep the exact failing configuration.
- If a dependency job failed first, investigate that upstream failure before a downstream cancellation.
Search carefully
GitHub's browser log search only searches expanded steps. Expand the relevant steps before concluding a message is absent. If you download a partial-rerun archive, it contains only the jobs rerun in that attempt; retain earlier attempt logs when you need the complete failure sequence.
- Keep enough preceding lines to explain the error, not the entire private log.
- Copy the line-specific link where the repository's access rules allow it.
- Remove tokens, private URLs, source code and customer identifiers before sharing excerpts.
Classify before changing
A dependency install failure, missing command and compiler error point toward different checks. An explicit permission or quota message belongs with the account holder. A test that correctly detects broken application behaviour should remain red until that behaviour is fixed; disabling the test changes the question rather than answering it.
- Do not enable verbose logging without considering what it can reveal.
- Do not rerun a workflow until its deployments and external effects are understood.
- Do not treat a cancelled or skipped job as a passing execution.
A useful first enquiry
Send the workflow and job names, trigger, runner label, consistency of the failure and a redacted excerpt containing the first error. State whether a branch run is safe. That is enough to begin scoping; credentials and confidential code are not first-contact inputs.
Sources and limits
- GitHub: workflow run logs Checked 2026-10-10.
- Failed steps can be inspected in each job.
- Log line links can identify an exact failure location.
- Search only includes expanded steps; partial-rerun logs omit jobs that did not rerun.