Saw an instance where ffmpeg died recording a stream. The stream-recorder lambda restarted it, but it took a bit, so we (A) missed some gap of time, and (B) only sent the 2nd partial recording to Feeder. So it would take manual work for someone to get the maximum amount of the hour.
We can do better!
- Can we build some immediate retry logic into Oxbow itself, so if ffmpeg exits 0 early, it restarts immediately?
- Can we concat the partial files together after, so Feeder gets them?
- We'd probably still need to tell Feeder there's a gap in the audio, somehow. And maybe indicate where it is? Hrm...
Saw an instance where ffmpeg died recording a stream. The stream-recorder lambda restarted it, but it took a bit, so we (A) missed some gap of time, and (B) only sent the 2nd partial recording to Feeder. So it would take manual work for someone to get the maximum amount of the hour.
We can do better!