I have my PMS running as a full VM under Proxmox 8.3. So far so good and I have passed my iGPU to the VM, so it shows up under hardware transcode config option. The VM has 2 vCPUs and 4 gigs of vRAM. So far this seems to be ok for all Plex related functions, although the vRAM usage is high, sitting at about 3.6 gig of vRAM consumed. This has been steady for 3 days now.
Does the vRAM usage stay steady, even when PMS is not doing anything? I am concerned that I might not have allocated enough vRAM but I am not sure. I am about to turn on a monitoring tool to the whole VM, so I should get better stats. I wish Plex or Tautulli would keep these kind of stats, but if not my monitoring tool(Zabbix) will do it for me.
Inside the VM itself, what do free -h and vmstat -S M show? The reason I ask is that Linux will allocate some of your RAM for disk I/O buffers/caches. This will appear as “used” RAM from an outside monitoring perspective (Proxmox) but this memory is generally available to applications, if needed.
Particularly in memory-constrained environments this can make it look like you’re bumping into high RAM utilization when you may not be (because the buffer/cache RAM is still largely available to apps, if needed).
A big tell (in my experience) is to look at what the swap memory utilization is (from free). If there’s a lot of swap in use, it might be an indication that you’d want to add more (v)RAM to the VM. It’s completely normal for some swap to be used, particularly when you have a relatively small amount of RAM.
Note: The above is a very, very simplistic overview of buffers and caches. DuckDuckGo is your friend here.
