DocsTroubleshootingTroubleshooting

Troubleshooting

Start from a Lifecycle symptom and open the guide for the failed phase with safe evidence and recovery steps.

Start with the user-visible symptom. Before you know which phase failed, do not make broad cluster changes. Do not start another redeploy, edit the database, or dump secrets.

SymptomFirst checkGuide
No Lifecycle comment on a pull requestGitHub App access and repository onboardingMissing pull-request comment
Environment shows Config errorRepository configuration validationFailed deployment
Service shows Build failedConsole → BuildFailed build
Service shows Deploy failed or does not become ReadyConsole → Deploy, then PodsFailed deployment
Environment remains queued or staleLast update time and if other work progressesStuck Environment
Cannot sign in, link GitHub, or call an APIIdentity, key prefix, scopes, and request IDAccess and API errors
MCP, Agent workspace, or Site feature is missingFeature availability and administrator policyThe feature’s availability section

Safe escalation bundle

Collect:

  • Lifecycle component and version
  • Environment ID and Service name
  • current status and status message
  • source repository, branch, and short SHA when they are not private
  • job name, phase, and timestamps
  • request or correlation ID
  • the smallest related, redacted log excerpt

Do not include access tokens, one-time API keys, OAuth callback query strings, Secret values, full Environment-variable dumps, private repository contents, or unredacted Agent conversations.