@FIGHT said:
Crap my server crashed again, has been running ok for the last many days…
The Plex Media Server.exe process ran out of memory. Could be a memory leak or insufficient memory available to the process.
The Plex Media Server process was started 7th May at 22:55. The logs which go back to 10:45 am 13th May showed the first memory allocation failure at 12:45
May 13, 2018 12:45:07.269 [8748] ERROR - Format [JPEG] - Insufficient memory (case 4)
This was followed by many allocation failures
May 13, 2018 12:45:07.372 [9700] ERROR - Format [JPEG] - Insufficient memory (case 4)
May 13, 2018 12:45:07.397 [13032] ERROR - Format [JPEG] - DIB allocation failed, maybe caused by an invalid image size or by a lack of memory
May 13, 2018 12:45:07.479 [2644] ERROR - Format [JPEG] - DIB allocation failed, maybe caused by an invalid image size or by a lack of memory
May 13, 2018 12:45:07.610 [2644] ERROR - Format [JPEG] - DIB allocation failed, maybe caused by an invalid image size or by a lack of memory
May 13, 2018 12:45:09.228 [13120] ERROR - Format [JPEG] - DIB allocation failed, maybe caused by an invalid image size or by a lack of memory
until eventual crash at 13:13
With memory leaks and getting to a point where the process reaches the maximum available memory, one would need to use diagnostic tools such SysInternals VMMap and SysInternals Process Monitor - following similar investigations as was done in these two forum threads
http://forums.plex.tv/discussion/319786/plex-media-server-crashing-memory-allocation-failures/p1
and
http://forums.plex.tv/discussion/310320/pms-crashing-version-1-11-3-4803-windows-7/p1
I hope you will be ok with going down this investigation route to try and see if there is a memory leak and where it might be
VMMap allows you to see in a timeline chart if memory usage is climbing up - the sign of a memory leak. And when memory usage remains climbing and is going over 1Gb (max possible is 2Gb) then one can run Process Monitor for 5 minutes at a time and capture events - a couple of such captures would probably be sufficient if captured during periods when VMMap timeline shows climbing memory use.
Have a look at the timeline charts pasted in the 2 forum threads.
In the last case, VMware video drivers were suspected as leading to a threads leak (too many threads 0 each using up 1.25Mb ram) - this was occurring when Plex Media Server was testing out codecs. The problem happened to go away after upgrading a client app !
The previous case was a bug in Plex Media Server which was fixed
These procmon captures together with saving the VMMap environment and the Plex Media Server logs zip, it should be possible then to narrow the problem down
SysInternal VMMap can be downloaded from here VMMap - Sysinternals | Microsoft Learn
SysInternals Process Monitor can be downloaded from here Process Monitor - Sysinternals | Microsoft Learn
ProcMon captures would be very large and capturing of events should be limited to 5 minutes at a time. The saved capture PML files do zip very well. These should be uploaded to a private area eg dropbox and link sent to me by private message. Do not filter any events for procmon. I can do the filtering when investigating and looking at the captures
VMMap mmp saved files are not very large - and you can save the VMMap environment to mmp file once you noticed te memory used by the Plex Media Server.exe process has reached over 1Gb and you have captured a couple of process monitor captures during the climb
Best to have all evidence together and you can send me link
Let me know if you need any clarifications
Thanks
