Problem to Solve:
With Scalelite v1.6.8 and later, recordings can now be uploaded directly to S3-compatible storage. However, Scalelite still appears to require local filesystem access for playback.
This means recordings must still be copied or synced to /var/scalelite/recordings for playback to function, defeating the purpose of fully storing recordings in S3.
This creates additional operational overhead and prevents a true object-storage-only setup.
Proposed Solution
Enable direct playback of recordings from S3 without requiring local copies. Possible approaches include:
- Allow Scalelite to generate playback URLs that point directly to S3 objects.
- Use pre-signed URLs for secure access.
- Treat S3 as a valid recording source, so recordings do not need to be downloaded to local storage.
- Update the recording manifest handling to support S3 paths.
This would allow playback to occur entirely through S3, making deployments lighter and more scalable.
Considered Alternatives
- Syncing recordings from S3 back to local storage before playback.
Works, but removes most benefits of object storage. - Using an NFS or distributed filesystem.
Adds complexity and is unnecessary when S3 storage already exists. - Custom scripts to rewrite playback URLs.
Fragile and not maintainable long-term.
None of these provides a clean or scalable solution like native S3 playback would.
Additional Context
- This feature would significantly improve containerized and cloud-native deployments.
- It aligns with the new S3-upload support but removes the last dependency on local disks.
- If direct playback from S3 is already supported or partially available, clarification would be appreciated.