Find the failure#
Record the account, project, deployment ID, source revision, and affected URL. Reproduce the failing action and read its logs with secrets and personal data removed.
| Signal | Check first |
|---|---|
| Build failed | Build logs and the last successful revision |
| Published app returns errors | Runtime logs, bindings, and data compatibility |
| URL serves the wrong build | Project, route, domain, and deployment target |
| Provider returns an error | Account, resource, permissions, and provider status |
If you need credentials, pin the task's setup and read the provider guide before using its recovery tools.
Restore service#
Fix a reversible configuration error at its source. For a rollback, choose a known working release and check what the provider restores. Database changes, queues, and external writes can outlast a code rollback.
Check the recovery#
Open the affected URL, confirm the intended release, and repeat the failing action. Test the main related flow and record the cause, fix, and remaining impact.
Fix account access#
Check which account the tool uses and whether the credential can access the intended resource. For an expired or revoked credential, follow the provider's rotation steps and update each affected host or deployment. See Credentials and native logins.