Plex Using 100+ GB of RAM during Schedule Tasks

Plex Media Server Desktops & Laptops
63

Another update … another crash.
Plex Media Server 1.25.4.5487 was released yesterday

I installed it and I still crashed due to excessive RAM usage during scheduled tasks. I enabled the secret setting to log memory usage, so here’s my log files in hopes that something stands out for why this is happening.
Plex Media Server Logs_2022-02-02_14-01-36.zip (5.1 MB)

Here is a log from my server host showing Plex’s excessive RAM usage before the limiter is forced to kill my Plex process.
plex_oom_log.txt (19.3 KB)

Highlight from that log

Feb  2 13:20:57 rhea kernel: Tasks state (memory values in pages):
Feb  2 13:20:57 rhea kernel: [  pid  ]   uid  tgid total_vm      rss pgtables_bytes swapents oom_score_adj name
Feb  2 13:20:57 rhea kernel: [  49716]  1380 49716 25244938 16284663 130904064        0             0 PMS RunServer
Feb  2 13:20:57 rhea kernel: [  49905]  1380 49905    16812     8853   155648        0             0 Plex Script Hos
Feb  2 13:20:57 rhea kernel: [  50630]  1380 50630    14272     6405   135168        0             0 Plex Script Hos
Feb  2 13:20:57 rhea kernel: [  50688]  1380 50688    14031     6089   126976        0             0 Plex Script Hos
Feb  2 13:20:57 rhea kernel: [  50690]  1380 50690    13931     5975   131072        0             0 Plex Script Hos
Feb  2 13:20:57 rhea kernel: [  50930]  1380 50930    17093     9264   159744        0             0 Plex Script Hos
Feb  2 13:20:57 rhea kernel: [  34236]  1380 34236      640       21    40960        0             0 EasyAudioEncode
Feb  2 13:20:57 rhea kernel: oom-kill:constraint=CONSTRAINT_MEMCG,nodemask=(null),cpuset=/,mems_allowed=0-1,oom_memcg=/elusivezatchmo,task_memcg=/elusivezatchmo,task=PMS RunServer,pid=49716,uid=1380
Feb  2 13:20:57 rhea kernel: Memory cgroup out of memory: Killed process 49716 (PMS RunServer) total-vm:100979752kB, anon-rss:65138168kB, file-rss:484kB, shmem-rss:0kB, UID:1380 pgtables:127836kB oom_score_adj:0
Feb  2 13:21:07 rhea kernel: oom_reaper: reaped process 49716 (PMS RunServer), now anon-rss:0kB, file-rss:0kB, shmem-rss:0kB

  • My OS is Gentoo.

  • My server host WhatBox uses OpenRC to install RPM versions of Plex.

  • Whatbox kills my Plex instance the moment it starts to use 100+ GB of RAM out of fairness to other uses on the server. This often results in the lack of a crash report generated by Plex.

  • My current database size is around 1.35 GB

  • I’ve doubled check that all my libraries are using the latest Plex scanners and Metadata agents

  • I’ve refreshed all of my libraries with the most current agents

  • I’ve cleaned up and optimized my DB nearly 100 times since the recent updates in hopes of bringing down the database size, but the size of my database remains unchanged.

  • I’ve used instructions provide on the Plex website to check my database for errors and corruption and repair.

  • I’ve updated preferences.xml to include memory usage in the logs to help diagnose the issue.

  • It’s been this way since the big database change over release.

  • It strictly only occurs during the window of scheduled tasks. It doesn’t happen during any playback or library refresh/scan run manually. If I change the scheduled task window, I’m able to change the time of the RAM burst.

  • I don’t wish to start over with a new database because I have a large amount of entries in TV, Movies, and Music that I either had to correct incredible mismatches by Plex or put in manually due to non-existence in Plex’s meta database and to start over would mean that these would then become lost in the library (since my library is quite big at this point).

  • I’m looking for either a fix to the crashing, or a way to export and then import my metadata back so that I don’t lose any of the hard work I’ve put into my server over the past several years.

Update 2/3/2022:
Once in a great while, Plex doesn’t use an excessive amount of RAM during scheduled tasks and doesn’t trigger the limiter. To me, there’s been no recognizable pattern for this, but since it didn’t blow up last night, here’s last night’s log file as a counter balance to the log files where it does blow up.
Plex Media Server Logs_2022-02-03_16-09-18.zip (5.3 MB)

Update 2/14/2022:
It appears to be related to detect intro; which is confusing to me because when new media is scanned in and new intros are scanned, it’s never had an issue. But for some reason, during scheduled tasks it’s exploding my RAM usage. I discovered this by turning off all scheduled tasks options and turning things off and on and running with only one setting selected on at a time.