Deployments

How we ship code to production.

TL;DR

Merge to main → CI runs → auto-deploy to staging → manual approval → production.

Prerequisites

• Your PR has been reviewed and approved
• CI is green on your branch
• You've tested locally against a staging-like config

The flow

Normal pathHotfix

1. Merge the PR to main.
2. CI builds the artifact and deploys to staging automatically.
3. Smoke-test staging. Link to the staging URL is posted in #deploys.
4. Click Approve on the production deploy in GitHub Actions.
5. Watch the dashboards for 10 minutes. If anything looks off, roll back.

1. Branch from main, make the minimal fix.
2. Open a PR tagged hotfix. This skips the usual review SLA — still needs one approval, but anyone on-call can approve.
3. Merge and deploy as above, but skip staging soak time.
4. File a follow-up issue for the real fix if the hotfix was a band-aid.

Rolling back

Roll back first, debug second

If production is unhealthy, roll back before investigating. A bad deploy on production costs the team more than a quick revert ever will.

Two options:

• Redeploy the previous tag from the Actions tab. Usually takes about 2 minutes.
• Revert the merge commit on main and let CI redeploy. Use this when you also want to preserve the revert in history.

Who can deploy

TODO: list the team / role that can approve production deploys.