I’ve built a new server with 128GB of ram, a 7950X and a proper asus motherboard paired with a 980 Pro 2TB SSD. On there, I’m running ESXI8.0, with in there a windows 11 Pro N virtual machine. All this was built and installed 4 days ago.
Every night at 2:00-2:10AM, my VM randomly completely died. I didn’t quite understand, although I did notice a strange log in the event viewer of windows stating that:
Windows successfully diagnosed a low virtual memory condition. The following programs consumed the most virtual memory: Plex Media Scanner.exe (14028) consumed 63484837888 bytes, dotnet.exe (13800) consumed 1089007616 bytes, and MsMpEng.exe (5112) consumed 301260800 bytes.
Now, I’m running Docker with some containers, sabnzbd and Plex, so I figured it’d either be plex or docker. According to the logs Plex was consuming a whopping 60GB OF RAM!
I managed to check it out and actually get a live view of what happened. I increased the maximum pagefile size to 32gb to see what’d happen.
The background maintenance was set to run at 00:00:
Avoid putting all tracks into one big folder.
Be on the lookout for folders full of unsorted tracks
or
“Anthology”-style boxsets with several dozen numbers of discs.
You might want to set these aside for now.
and the list goes on like that. There are a couple of actual albums with just the album in them as well, all sorted by genre. That would seem (to me, at least) like Plex should be able to manage. However, since turning off the “Generate loudness data” stuff the background task finishes in literally 15 seconds and uses practically no RAM at all.
Any suggestions on what folder structure would sort this?
Also, in the MP3 subfolders are always just 1 song, but some have the most ridiculous names (even non-UTF-8). Maybe Plex is highly confused?
UTF-8 can encode every character of Unicode.
Maybe you mean Latin1? But anyway, this shouldn’t be a problem unless you use file shares in combination with ancient versions of SAMBA.