seekdb Issue Triage

When to use

Use this skill when investigating a seekdb or OceanBase bug that needs a repeatable reproduction, bounded evidence, root-cause analysis, and a focused fix.

Workflow

Record the environment.
- Host OS, kernel, architecture, and container image.
- seekdb revision and package version.
- Relevant resource parameters.
- Whether the database directory was newly created or recovered.
Build a minimal reproduction.
- Preserve the original data and perform recovery experiments on copies.
- Record exact SQL and client commands.
- Use timeouts so blocked operations produce bounded evidence.
Collect evidence.
- Relevant seekdb logs.
- Transaction and lock state.
- Lock detector output and transaction state transitions.
- Log file hashes and timestamps when retaining evidence.
Separate the original trigger from the recovery failure.
- Do not claim to have identified the trigger without evidence.
- State clearly which failure the fix addresses.
Implement a focused fix.
- Preserve transaction and storage compatibility.
- Avoid changing live, exiting, or already decided transactions.
- Add retry and failure paths for durable recovery operations.
Validate.
- Build the affected target.
- Add or update a regression test.
- Reproduce the original failure before the fix.
- Re-run the same scenario after the fix.
- Verify restart behavior, data integrity, and absence of orphan locks.
Prepare the pull request.
- Explain root cause, behavior change, compatibility, validation, and limits.
- Link the related issue.
- Do not include credentials, private endpoints, or production data.

Output format

Provide:

Environment summary
Reproduction steps
Root cause
Fix summary
Tests and results
Compatibility impact
Remaining risks or unknowns

Safety

Never modify the only copy of affected data.
Redact tokens, passwords, private URLs, and personal information.
Prefer bounded timeouts over indefinite waiting.
Do not report a fix as complete without runtime verification.