I wasn’t able to get this file but the same issue just occurred. Have snagged every file from the logs folder. Server went inaccessible at Jan 28, 2020 17:16:35.436 [0x7fac95783700] ERROR - Couldn’t accept new TLS connection: Too many open files
You have a flurry of sync activity. That activity is overloading your max network connections (128 by default). Normally it’s not a big deal but, with a default timeout of 60 seconds after being closed before the slot becomes available for other use, this does become a big problem.
We can either find out why there is such a flurry on the machine
-or-
customize the system’s configuration a bit to handle the higher load.
Is this the same thing happening to other plex server unavailable issue ? maybe the others also ran out of connections ? is there a way to increase this max connections as some routers can handle much more ?
It’s not a router issue that I’m aware of from what the logs show me.
I see the flurry of activity. If the router were dropping, I’d see the disconnects which I don’t.
This is a non-permanent test you can perform. It is ONLY valid until you restart the system again.
This will give some measure as to whether you’re being overwhelmed with traffic or not. What this does is reduce the close wait time from 60 seconds to 30 seconds after any process is done with the connection. (The unused slots will become available again sooner).
If this results in no positive change after a period of testing, restart the machine to return to the Linux default values.
THat’s great. Ideally I’d like to understand what’s causing the overload of sync activity and so attack the problem at cause but am also happy to configure for higer maximum connections as a workaround.
Sync-wise - there shouldn’t be too much hapening as far as I know.
I feel the first step is to confirm whether or not it’s indeed an overload of the inbound requests to the server.
This run-time tweak does just that.
In the logs, I saw several sync requests generated at once. It looked like playlists were being pushed / updated (not clear in the logs) but a lot of traffic was flowing in that short time. The traffic isn’t so much a volume problem as a transactions/minute problem.