| ItamarWMDE | |
| Sep 7 2023, 3:14 PM |
| F37817151: image.png | |
| Sep 27 2023, 1:30 PM |
After an incident involving too many concurrent runs of the "Ship & Deploy" workflow in mismatch finder, we decided to utilize GitHub Actions' concurrency groups.
Docs: https://docs.github.com/en/actions/using-jobs/using-concurrency
Acceptance Criteria:
Task Breakdown notes:
Tested with manual testing in a staging branch (/staging/**) by pushing an empty commit after the commit with the workflow change:
Proof:
Test:
two PRs were created and made to run at the same time.
This is the result:
The run in the /staging/** branch was cancelled: https://github.com/wmde/wikidata-mismatch-finder/actions/runs/6327107547
and the run in /gh-actions/** that was running simultaneously ran until the end. https://github.com/wmde/wikidata-mismatch-finder/actions/runs/6327107547
Relevant PR for staging branch: https://github.com/wmde/wikidata-mismatch-finder/pull/683
Relevant PR for development branch: https://github.com/wmde/wikidata-mismatch-finder/pull/708
We need to watch out when we are changing the development branch, this caused a small incident where development branch was accidentally deployed to production (which caused imports to fail due to T313467 already being merged to development). I think we should remove the gh-action/* branch deployment to production, and just rely on local testing and testing on staging, to avoid future issues.