[SOLVED] Weird visual bug - RAM showing wrong usage value for VM

[crawrj]

SOLVED thanks to _gabriel - The Balloon service was missing somehow.

For me, repairing or reinstalling did not fix it. I had to run the .exe, choose change, remove Balloon, wait for it to finish, repeat and add Balloon back.

I have a weird visual bug that has appeared out of nowhere. The system has been fine for months, but suddenly, I have two of ten virtual machines showing the wrong RAM usage value on the summary page. All of my virtual machines are Server 2025. Two of them have the bug, and the other 8 are working fine. These two virtual machines will continue to grow RAM usage until they eventually hit 100%, but when I check the VMs, they are only using about 18%. My hosts are exactly the same. All VMs are basically the same, just different use cases. If I migrate the impacted VMs, the value stays the same. I have rebooted the hosts multiple times. I have reinstalled the virtio software. I have restarted the VMs. I have powered off the hosts and VMs. Nothing has fixed it. Has anyone experienced this or have any ideas? It isn't impacting performance at all, but it would be nice not to see the big red value when it isn't actually a problem, so that I know when there is a real problem.

Also, it has persisted through multiple kernel updates and other package updates. It has been happening for a few weeks, but I'm just now getting around to posting this.

Both hosts are fully updated as of today using the enterprise repository.

Current versions
PVE 9.2.4
Linux 7.0.14-11-pve (2026-08-06T23:36Z)

 

[_gabriel]

What about Cached memory usage within VM?

 

[crawrj]

What about Cached memory usage within VM?

It is in line with the other servers. Using one VM as an example, DHCP server 1 is showing 75.71% usage in GUI but 19% in VM. In VM it shows 2.9GB in use, 3.0/18.3GB Committed and 7.3GB cached. DHCP server 2, which is identical in hardware and config to 1, is showing 22.48% usage in GUI and 23% in VM. In VM it shows 3.3GB in use, 3.5/18.3GB Committed and 8.0GB cached.

 

[_gabriel]

is Windows service "Balloon" present and running in dhcp 1?

For information :
(again mixed myself, Balloon service installed with Qemu Guest agent reports used RAM not the cached).
VM "Memory usage" = "used Memory" in Windows.
But remember, Cached memory within VM is really used and reserved in the Node/Host RAM,
that's why PVE UI Node/Host "RAM usage" reports the real numbers ( Used + Cached + Overhead memory of all VMs)

 

[Impact]

Give this (especially the links there) a look. Can you share screenshots of what you see?

 

[crawrj]

is Windows service "Balloon" present and running in dhcp 1?

For information :
(again mixed myself, Windows Qemu guest agent reports used RAM not the cached).
if Qemu Guest Agent is running within Windows Guest and Qemu Guest Agent option is enabled in PVE \ VM \ Options, VM Memory usage display the same Memory used as Windows.
But remember, Cached memory within VM is really used and reserved in the Node/Host RAM, that's why PVE UI Node/Host "RAM usage" reports the real numbers ( Used + Cached + Overhead memory of all VMs)

You are a genius. I didn't think to check the service because I had reinstalled it twice during testing, one repair and one reinstall. That is the issue. Both bugged servers didn't have the balloon service at all, which is kinda blowing my mind. How did it get removed? Why did repairing and reinstalling not put it back? I had to run the exe, choose change, mark Balloon as removed, wait for it to finish, repeat the process and add Balloon then it shows up in services. So odd.