We use OpenVidu CE 2.15.0 deployed using this instruction in our application and we had an video session with Individual output mode and Manual recording mode (session name live-interview-549).
When two participants were connected we started recording of the session.
During video session everything was ok, but when we download recordings zip file, we saw that one of the video file has 80 seconds start offset, even though files have the same duration, and seems like recording for both of them were started more or less at the same time. As recordings are processed automatically by our system and streams synchronize based on start and end offset we had an out of sync result at the end.
In logs we have next:
Recording started for session
2021-03-29T13:04:07.996738903Z [INFO] 2021-03-29 13:04:07,996 [http-nio-0.0.0.0-5443-exec-80] io.openvidu.server.recording.service.SingleStreamRecordingService - Starting individual (video+audio) recording live-interview-549 of session live-interview-549
Starting recording for first stream
2021-03-29T13:04:07.996824388Z [INFO] 2021-03-29 13:04:07,996 [http-nio-0.0.0.0-5443-exec-80] io.openvidu.server.recording.service.SingleStreamRecordingService - Starting single stream recorder for stream str_CUS_Tob2_con_HJu4bn0YGp in session live-interview-549
2021-03-29T13:04:08.131210308Z [INFO] 2021-03-29 13:04:08,131 [AbstractJsonRpcClientWebSocket-reqResEventExec-e2-t59809] io.openvidu.server.recording.service.SingleStreamRecordingService - Recording started event for stream str_CUS_Tob2_con_HJu4bn0YGp
So it started almost instantly, but for second stream we have next:
2021-03-29T13:04:18.414684660Z [INFO] 2021-03-29 13:04:18,414 [SessionHandler-m7m7cpovrufl4lg3rj8qt4i62l-e7068-t0] io.openvidu.server.recording.service.SingleStreamRecordingService - Starting single stream recorder for stream str_CUS_D5Me_con_BJiwh0lwJa in session live-interview-549
…
2021-03-29T13:05:28.117149165Z [INFO] 2021-03-29 13:05:28,117 [AbstractJsonRpcClientWebSocket-reqResEventExec-e2-t59797] io.openvidu.server.recording.service.SingleStreamRecordingService - Recording started event for stream str_CUS_D5Me_con_BJiwh0lwJa
So it took more then one minute, and probably that’s why we have such big offset. Our first thought was that it caused by server load, but there was not more than 30% of cpu usage, and there also was plenty of RAM and IO resources.
Could you give us any idea why this can happen and how to avoid such situations in future ?