Summary

After a failed gateway start, rerunning setup can fail with Corrupted cluster state until the user manually removes openshell-cluster-nemoclaw Docker volumes.

Problem

Cleanup is incomplete, and the current errors do not explain how to recover. Users are forced into manual Docker state management.

Expected behavior

Destroy and cleanup flows should leave the environment in a rerunnable state, or the error should include the exact automated recovery action.

Acceptance criteria

• Failed setup attempts do not leave behind cluster state that breaks the next run.
• gateway destroy or equivalent cleanup removes the Docker volumes required for a clean retry.
• If automatic cleanup is not possible, the error message prints the exact recovery command.
• A test covers rerunning onboarding after an intentionally failed gateway start.