Integration Scenarios:
Logging, Error Handling, and Operational Acceptance The platform MUST assign a unique run ID for each recording session and log at minimum: robot device ID, hardware/software versions, request timestamps, action types, HTTP status codes, recorder status codes, target disk directories, total byte counts, MCAP start and end timestamps, required Topic message counts, and approved cleanup execution results. Logs MUST NOT contain tokens, Authorization headers, SSH passwords, or full business request bodies.
| Symptom / Observation | Action / Resolution |
|---|---|
| HTTP 400 / 411 / 413 | Correct JSON formatting, strictly enforce 4-field metadata rules, or add missing Content-Length; the request did not reach the ROS 2 service layer |
| HTTP 401 | Verify token injection, rotation policies, and client process environment variables; MUST NOT output credentials in log files |
| HTTP 502 / ROS 2 Timeout | Command execution outcome is indeterminate; check active recording directories, completed directories, and recorder state before attempting any retry of start |
Back-to-back start calls or third-party client intervention | Suspend automation pipelines and confirm robot-level control authority; in-process locks do NOT provide distributed mutual exclusion |
stop succeeds but list does not immediately show the directory | Wait for directory relocation from mcap_recording to mcap_recorded; if the timeout expires, preserve host system logs for analysis |
| File transferred successfully but parsing fails | Do NOT delete remote source files; verify file integrity, MCAP format version, schema, and serialization definitions |
Minimum operational acceptance criteria for production integration:
- Read-only topic connectivity verified.
- All invalid/missing token requests strictly rejected (HTTP 401).
- Complete
start→stop→listworkflow passes using a unique run ID. - Exact directory file transfer via SSH/SFTP succeeds.
- MCAP start/end timestamps, Topics, and message counts pass validation.
- Disk space alerts are properly visible and monitored.
- Cleanup of exact directories executes ONLY after explicit, separate authorization.
- System recovers normally following bridge restarts and token rotations.