[SOLVED] The virtual machine memory usage displayed on the Proxmox VE platform is inconsistent with the actual usage of the virtual machine

[sungl]

The virtual machine has allocated a total of 32GB of memory and actually used 2.8GB. However, on the Proxmox VE platform, the graphics interface shows that the memory usage rate of this virtual machine is 21.6GB. After testing, the memory usage of this virtual machine displayed on the Proxmox VE platform will continue to increase over time.

1. The above situation will occur when the Ballooning feature is enabled or not enabled
2. The virtual machine has installed the QEMU Guest Agent service
3. The Proxmox Virtual Environment version is 8.0.3
4. The virtual machine version is CentOS7.6



At present, it is uncertain what impact this display will have, but it feels like it is a potential issue that requires assistance.

 

[LnxBil]

Please post the free output in code tags from your pve host. This is most probably due to caching.

 

[sungl]

Please post the free output in code tags from your pve host. This is most probably due to caching.

The internal cache of the virtual machine is 3.8GB, which is too far from the 21.6GB displayed on the platform. Swap is currently in an unused state.

 

[LnxBil]

The internal cache of the virtual machine is 3.8GB, which is too far from the 21.6GB displayed on the platform. Swap is currently in an unused state.

Even if you freed it, it'll be used from the PVE side nonetheless.

If you don't answer my question, please search the forum for this question... it is asked once a week and gets always the same answer. I wrote exactly that as my first answer and now regret that I changed it in order to help you.

 

[sungl]

Even if you freed it, it'll be used from the PVE side nonetheless.

If you don't answer my question, please search the forum for this question... it is asked once a week and gets always the same answer. I wrote exactly that as my first answer and now regret that I changed it in order to help you.

Thank you for your reply.

Sorry, I still want to know how to handle this issue. What I ultimately want to achieve is that the memory indicators seen by the platform are similar to the actual values used internally by the virtual machine.

 

[sungl]

Please post the free output in code tags from your pve host. This is most probably due to caching.

Thank you for your reply.

Sorry, I don't quite understand what actions are needed. I took a screenshot of the memory usage of the host computer.

 

[LnxBil]

What I ultimately want to achieve is that the memory indicators seen by the platform are similar to the actual values used internally by the virtual machine.

That is exactly whats wrong with all the post including yours. This cannot be done, because the platform/PVE is correct, not the VM. Memory is managed by PVE, so if it says the memory is used, the memory is used.

Sorry, I don't quite understand what actions are needed. I took a screenshot of the memory usage of the host computer.

Your two screenshots are not from the same machine. One is showing TiB and the other is showing GiB.

 

[gfngfn256]

Your two screenshots are not from the same machine. One is showing TiB and the other is showing GiB.

As far as I understood, the OP's original images (first post) are from the VMs stats, while the more recent post is from the PVE host, so I believe no contradiction there.

I do however totally agree to your comments:

(a) The PVE host "knows" what its talking about, & in any event this is the actual number that matters.
(b) Its probably just a caching situation & best left alone.

Yes, its almost nauseating how many times this very question gets asked on this forum. It would appear Google doesn't fully operate in all countries!

 

[LnxBil]

Yes, its almost nauseating how many times this very question gets asked on this forum. It would appear Google doesn't fully operate in all countries!

Yes, we need a page like https://www.linuxatemyram.com/ but for VMs.

 

[sungl]

As far as I understood, the OP's original images (first post) are from the VMs stats, while the more recent post is from the PVE host, so I believe no contradiction there.

I do however totally agree to your comments:

(a) The PVE host "knows" what its talking about, & in any event this is the actual number that matters.
(b) Its probably just a caching situation & best left alone.

Yes, its almost nauseating how many times this very question gets asked on this forum. It would appear Google doesn't fully operate in all countries!

Thank you for your reply.

I am unable to Google in my region. I have searched for these questions through Google and some search engines in some ways, but have not found the answer I am looking for. It may be that I searched for the wrong keywords or method.

 

[sungl]

As far as I understood, the OP's original images (first post) are from the VMs stats, while the more recent post is from the PVE host, so I believe no contradiction there.

I do however totally agree to your comments:

(a) The PVE host "knows" what its talking about, & in any event this is the actual number that matters.
(b) Its probably just a caching situation & best left alone.

Yes, its almost nauseating how many times this very question gets asked on this forum. It would appear Google doesn't fully operate in all countries!

Thank you for your reply.

 

[sungl]

Thank you for your response.

I have found many similar issues through forum search. Below is my summary, hoping to provide a reference for those who have similar issues:

1. Completely shut down swap on the virtual machine
2. Enable memory balloons on the platform side for this virtual machine
3. On the platform side, disable the disk acceleration feature for this virtual machine
4. The virtual machine needs to enable the QEMU agent

https://forum.proxmox.com/threads/memory-usage-on-proxmox-is-different-than-vm.109644/
https://forum.proxmox.com/threads/memory-usage-not-same-as-in-vm.82640/
https://forum.proxmox.com/threads/ram-usage-inconsistent-in-almalinux-vm.127418/

 

[gfngfn256]

Good you've got it working. Maybe tag mark the thread-title with [SOLVED], (upper right hand corner under title).