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.