@pixelflowers @palomakop @andrei_jay
taking a wee look into this now.
in the job panel i can see that some transcoding jobs were failing
for some of them the messages says:
Error: ffmpeg was killed with signal SIGKILL
which seems to imply that something (peertube?) killed the transcoding job - maybe it was taking too long ?
on the 1st of jan the cpu is at 100% for around 12 hours. so my guess is that some jobs may have timed out during these times.
one issue here is the droplet that videos.scanlines runs on is too small (only 1cpu). since it does appear to be getting used more these days i think we could justify bumping it up to a 2cpu droplet. this should solve some of these issues (and transcode faster in general
the other reason i can see for jobs failing is:
ffmpeg exited with code 1: Conversion failed!
im guessing this one is related to the format of the original video that is uploaded.
here is the user : name of the transcode jobs that failed in the last week:
sigkill:
150f25fe-fa93-4dcd-b692-16e892485ed2 - andrei_jay : Ephemeris Time Block1-1
6d0ebb97-1043-4250-9cab-1805f6ac98ee - pixelflowers : 2020-12-30-02-40-20-logo-1280
conversion failed:
02cac947-55dd-465e-b663-99f0748f183e - deme : #2.mp4
dc043831-9e82-4f13-af3f-8cff12ec4091 - deme : #3.mp4
7133c865-874d-4713-aa58-f5865dd1b76b - deme : #4.mp4
9e330284-f8e8-49f7-8522-d182bca203e8 - pixelflowers : 2020-12-30-02-40-20-logo-1280