feat(session-replay): upload replay envelope from the native crash daemon#1809
Conversation
|
tustanivsky
force-pushed
the
feat/session-replay-prod
branch
from
154ce2f to
b437853
Compare
tustanivsky
changed the title
![sentry-warden[bot]](https://avatars.githubusercontent.com/in/2755838?s=60&v=4){kind=small}
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
![sentry[bot]](https://avatars.githubusercontent.com/in/12637?s=60&v=4){kind=small}
jpnurmi
left a comment
jpnurmi
left a comment
There was a problem hiding this comment.
Very cool to see session replays becoming properly integrated! 🎉
jpnurmi
left a comment
jpnurmi
left a comment
There was a problem hiding this comment.
LGTM 👍
Even if this feature is currently only used and tested by sentry-unreal, it wouldn't hurt to have something in test_integration_native.py to verify that a correctly structured replay envelope is sent when a dummy replays/foo.mp4 exists.
Not sure if you saw the discussion about the overhead of the continuously flushed __sentry-event file yesterday and how it will likely be replaced by IPC messaging in the future. Such a change would also impact the replay envelope, so it'd be nice to prevent potential regressions early. 🙏
|
Added a minimal set of tests in b1ab19c. Maybe creating a dummy mp4 will work to test this end-to-end in Unreal too 🤔 |
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 6cabda1. Configure here.
2 participants
This PR adds Session Replay product support to the
nativebackend. When a downstream SDK (e.g. sentry-unreal) has staged a recorded clip on disk, the crash daemon now builds areplay_videoenvelope from it and performs an upload so the clip becomes a first-class Session Replay in Sentry, linked to the crash instead of only being available as a file attachment.Key Changes
sentry__session_replay_flush_pending(): scans<database>/replays/for embedder-staged clips (areplay-<id>.jsonmetadata sidecar + its.mp4), builds thereplay_event+replay_recording+ video into areplay_videoenvelope, and hands it to the transport.trace_id, read back from<run>/__sentry-event. This makes the replay carry the same tags as the crash and correlate via the same trace.flush_pendingout-of-process, so the replay is sent in the same session as the crash (no next-launch step).replays/convention.replay_videoenvelope items are now classified under a dedicatedreplaydata category instead of falling back toerror. This stops error rate limits from dropping replays, honors a server-sidereplayrate limit, and reports discarded replays under the correct category.Known Limitations
crashpad/breakpad/inproc.Related items
#skip-changelog