Production freshness #679
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: Production freshness | |
| # Ground truth for "is pretable.ai serving `main`?", asked from outside any CI | |
| # run. | |
| # | |
| # The in-run alarm (`Production deploy — did it actually happen?` in ci.yml) | |
| # catches a skipped or failed deploy the moment it happens, but it lives inside | |
| # the run, so it cannot see: | |
| # | |
| # - a `main` run cancelled outright (the alarm job is cancelled with it), | |
| # - a deploy that reported success while the production alias did not move, | |
| # - a rollback or promotion done by hand in the Vercel dashboard, | |
| # - anything that happens between runs. | |
| # | |
| # This workflow compares `main`'s head commit against the commit production was | |
| # built from (`/version.json`) and opens the same tracking issue when they | |
| # disagree. GH Actions owns deploys here — Vercel's git auto-deploy is disabled | |
| # — so `main` and production are supposed to agree at rest. | |
| on: | |
| schedule: | |
| # Every 30 minutes. A missed deploy went unnoticed for hours on 2026-08-14; | |
| # half an hour is the target for how long that can happen again. | |
| - cron: "*/30 * * * *" | |
| # So it can be run on demand right after a suspicious merge, and so this | |
| # workflow is testable without waiting for the cron. | |
| workflow_dispatch: | |
| inputs: | |
| dry_run: | |
| description: "Report the decision without touching the tracking issue" | |
| type: boolean | |
| default: false | |
| permissions: | |
| contents: read | |
| jobs: | |
| freshness: | |
| name: Production is serving `main` | |
| runs-on: ubuntu-latest | |
| permissions: | |
| contents: read | |
| issues: write | |
| steps: | |
| - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7 | |
| with: | |
| ref: main | |
| - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7 | |
| with: | |
| node-version: 24.19.0 | |
| - name: Read main's head commit | |
| id: head | |
| run: | | |
| echo "sha=$(git rev-parse HEAD)" >> "$GITHUB_OUTPUT" | |
| # Committer date, not author date: a rebased or squashed commit gets a | |
| # fresh committer date at merge time, which is when the deploy clock | |
| # actually starts. | |
| echo "iso=$(git log -1 --format=%cI)" >> "$GITHUB_OUTPUT" | |
| - name: Compare production against main | |
| env: | |
| GH_TOKEN: ${{ github.token }} | |
| SITE_URL: https://pretable.ai | |
| EXPECTED_SHA: ${{ steps.head.outputs.sha }} | |
| HEAD_COMMIT_ISO: ${{ steps.head.outputs.iso }} | |
| # A deploy takes a few minutes; anything younger than this is not yet | |
| # expected to be live. Without the grace window this would alarm on | |
| # every healthy merge that raced the schedule. | |
| GRACE_MINUTES: "25" | |
| DRY_RUN: ${{ inputs.dry_run && '1' || '0' }} | |
| run: node ./scripts/check-prod-deploy.mjs --mode=freshness |