How-to: Roll Back a Failed Deploy

We’ve all shipped a hotfix that backfired. This is the playbook we reach for when production needs a time machine.

Goal

Promote the last known-good Vercel deployment and redeploy any Supabase Edge Functions that might have shipped bad code.

Steps

  1. Roll back on Vercel.
    • UI path: open the most recent successful deployment → “Promote to Production.”
    • CLI path:
vercel rollback production
  1. Redeploy Edge Functions if ingestion was part of the bad release.
supabase functions deploy rss-fetch --project-ref $SUPABASE_PROJECT_REF --no-verify-jwt
supabase functions deploy twitter-fetch --project-ref $SUPABASE_PROJECT_REF --no-verify-jwt
  1. Smoke test production. Hit a critical endpoint and sanity-check the HUD.
curl -sS https://your-app.vercel.app/api/feed | head -c 200

Or load /hud and confirm stories render without console errors.

Pitfalls

  • Log the incident right away. Future-you will thank present-you for timestamps and root-cause breadcrumbs.
  • If the failure involved leaked env vars, rotate them before re-promoting production.

Time estimate

5–10 minutes. The longest part is usually writing the retro doc afterward.