Memory ballooning and stats on main page

Feb 13, 2019
10
0
6
We have a small proxmox ve server, mainly hosting Windows VMs. Because there is only 128GB memory we want to use ballooning as described in https://pve.proxmox.com/wiki/Dynamic_Memory_Management. So we've set up the driver in the VMs and the memory usage of the VMs is shown correctly.
However the stats on the main page show the totally reserved maximum memory, not the actually used size.
Did we make a mistake in the configuration or do we just have to use another statistic?

Thanks a lot and best regards,
André

Schnappschuss_021319_102356_AM.jpg Schnappschuss_021319_102650_AM.jpg
 
you can change which columns are shown when you click the arrow that appears when you move your mouse over a column header
there are many things to show, but some of them are hidden by default

edit: mhmm sorry it seems i misunderstood
where exactly does it show the wrong information?

the node memory usage is the correct one
 
I am guessing the PM host shows correct values, as VMs use the "free" memory for caching and the memory is not really free from the hosts point of view, just from the VMs point of view because it is using it for caching (and PM gets free memory it via qemu guest agent which does not report buffered and cached as used). Check out free -h on host as well as htop and or top and from inside the VMs (pay close attention to buffered and cached memory inside VMs).

Here is how one small node of mine looks (same symptoms as yours).
PM/Host usage: 31,08 GB of used RAM in PM GUI.
VM1: 2,3 GB from 4 GB delegated.
VM2: 3,71 GB from 5 GB delegated.
VM3: 6,75 GB of 10 GB delegated.
VM4: 3,91 GB of 8 GB delegated.

So one would think the correct memory usage would be VMs + Host = 16,67 GB + max cca. 2GB for Linux/PM and max 4 GB ARC = 24 GB
But in reality is 31,08 GB because: 4 + 5 + 10 + 8 + Host = 27 GB + 6 GB = 33 GB, which is somewhat close to what PM GUI reports.