There appears to be a file descriptor leak in '/usr/lib/plexmediaserver/Plex Media Server'. I say this because over the past 2 months I've had to restart plex 3 times and every time it has been because this process hit it's max of 1024 file descriptors. When this has happend, nearly all of the file descriptors in use as reported by lsof look like this:
Plex 17186 plex 1020u sock 0,8 0t0 4197068556 can't identify protocol
Plex 17186 plex 1021u sock 0,8 0t0 4197075113 can't identify protocol
Plex 17186 plex 1022u sock 0,8 0t0 4197071252 can't identify protocol
Plex 17186 plex 1023u sock 0,8 0t0 4197069506 can't identify protocol
I can say with 100% certainly this occurs in the following versions:
0.9.12.13.1464-4ccd2ca
0.9.15.0.1621-344f193
I just upgraded to 0.9.15.1.1639-26325ea so there is a possibility it is fix in this version, but as I can't find any other reports of this and I could not find anything relevant in the change logs, I expect the problem is still present.
Unfortunately, I am going 10 to 20 days between hangs, and I don't know what is triggering these sockets, so all I can do is wait and check on the file descriptors every few days to find out if it appears to be continuing.
I suspect this has something to do with syncing content to remote devices as I started using things feature fairly heavily around the time I started having the problem. When my server is transcoding I will see those types of sockets and then the transcoding stops and they go way typically, but I suspect in some cases the transcoding is interrputed for some reason and that results in the leaked file descriptor. But I don't really know.
In any case I just wanted to throw this out there as having the service become unresponsive is very annoying. I could easily moinitor for this condition and automatically restart the service, but I figureed it was better to let other know what I was experiencing and find out if others are seeing the same thing.
Thanks!
