Skip to content

Recover a failed deploy

A failed deploy may leave the previous version working. Identify what broke before changing the app, and protect existing work and data.

You are reading the v0.19.2 documentation. Open the current docs.

More formats
Raw Markdown ↗

Find the failure#

Record the account, project, URL, source revision, deployment ID, and failing user action. Read the relevant logs without exposing secrets or personal data. Check for:

  • A build failure before publication.
  • An error in published code.
  • The wrong account, project, environment, route, or domain.
  • A missing binding, permission, or enabled API; or a provider outage.
  • Data or schema changes the app cannot use.

If access is needed, run godmode help and godmode provider SLUG. Use the provider's official tools and current recovery instructions.

Restore service#

Rolling back code may leave database changes, secrets, queues, and external writes untouched.

Check what the provider's rollback restores before choosing a known working release. For a reversible configuration error, make the narrow correction. Before a destructive change, preserve evidence and test the recovery plan.

Do not repeatedly deploy the same failing version or delete resources to hide an error.

Check the recovery#

Open the production URL, confirm the intended release, and repeat the failing action. Test its most important related path too.

Record the cause, fix, deployed revision, remaining risk, and follow-up work. Keep synthetic tests separate from real customer incidents.

Fix account access#

Check identity, scope, resource selection, expiry, and provider status before replacing credentials. Saving a new key on your laptop does not update a deployed app or another host. See Credentials and native logins for replacement and refresh.

Godmode v0.19.2 · 2026-11-17 · Inspect the source · Get help