`We could not find recordings` error in PeerTube jobs during BBB classroom VOD conversion

[NamFra]

Bug Report

Problematic Behavior
A significant number of PeerTube jobs systematically end with error state We could not find recordings:

• the first resolution job runs normally (max-quality)
• all subsequent jobs fail with this error

This issue only affects the conversion of BBB classroom sessions into VOD.
This is the most frequent error occurring during the vod-hls-transcoding process:

• 5,534 jobs with this error,
• compared to 390 jobs for the second most frequent error reported by django_peertube_runner_connector_runnerjob (Video matching query does not exist).

While the final output seems to be generated correctly, the error state is misleading and may indicate an underlying issue in the workflow. The impact on Marsha's overall transcoding functionality is unclear.

Expected behavior/code
Ideally, PeerTube jobs should complete successfully without errors. However, based on observations, two scenarios are possible:

• If the error is cosmetic (e.g., incorrect status reporting while the conversion actually succeeds), the job status should accurately reflect the success of the conversion to avoid confusion
• If the error indicates a deeper issue (e.g., partial failure, metadata mismatch, or timing problem): the root cause should be investigated to ensure the reliability and consistency of the transcoding process

---

Steps to Reproduce

1. Record a classroom session in BBB.
2. After ending the session, click "Convert to VOD" from the Recordings section.
3. Monitor the jobs during the conversion process:
   • The first job completes successfully
   • Subsequent jobs fail and display the error state

---

Environment

• Marsha version: since 5.7.0?
• Platform: production

---

Possible Solution
The error first appeared on April 9, 2025, which coincides with:

• The announcement of the Marsha 5.7.0 release in the #marsha channel
  • Note that this release is officially dated May 16, 2025, idk why 🤔
• The deployment of this commit: ee5850a511530bdcf60ed958781e15746fa3b7b4

The error appears to originate from BigBlueButton:

• No related logs are visible in Sentry
• The error message is hardcoded in BBB's source code:
  • RecordingController.groovy#L189
  • RecordingController.groovy#L357

Is this issue potentially caused by:

• BBB's handling of recording metadata or API responses?
• A timing issue where subsequent jobs are triggered before the recording is fully processed?

---

Additional context/Screenshots

• Error occurrence and volume:

• Job states during transcoding (4-minute recording)

1. Jobs start normally:

2. Errors appear after processing:

3. Jobs are actually converted successfully but still display the error:

The video remains in an error state in the admin interface