Memory Usage Mismatch After swapoff in Proxmox

[tyuhbh0]

I’m noticing that the memory usage shown in htop, top, and Proxmox Web UI seems incorrect or inconsistent, especially for free memory — it's much lower than expected even when VMs aren’t using much.

My PVE Version 8.4.1.

whenever i am rebooting then show fine but after an hours again Memory usage increased drastically.

root@TAK-1:~# free
total used free shared buff/cache available
Mem: 197762588 126903776 3412988 84648 70702216 70858812
Swap: 8388604 524 8388080


Please help!

 

[janus57]

Hi,

For help you need to tell more about your setup.

And have you read this[0] ?

[0] https://forum.proxmox.com/threads/f...ram-usage-of-the-vms-in-the-dashboard.165943/

Best regards,

 

[tyuhbh0]

@janus57

Not helpful. (https://forum.proxmox.com/threads/f...ram-usage-of-the-vms-in-the-dashboard.165943/)

i have 3 server cluster. 2 has exact memory. but one has showing me wrong memory.

All Vms and LXC container working good & showing exact memory as i have allotted. but PVE is showing wrong memory usage.

total memory using by all vms & LXC is around 68GB but it consuming 128 or 121GB. i don't understand what happened.

 

[janus57]

Hi,

Well, without sharing your exact setup, it's impossible to help you other than general FAQ.

Best regards,

 

[tyuhbh0]

Hi,

Well, without sharing your exact setup, it's impossible to help you other than general FAQ.

Best regards,

exact setup i didn't get you. 3 nodes clusters PVE.


8 Vms and 8 LXc containers . Thats'a it.

 

[guruevi]

You can use tools like free and ps/top to see what is consuming your memory. But there are many variables, are you using balloon memory, are you using PCIe passthrough, what CPU/memory/BIOS settings. Are you actually running out of memory, or are you just complaining about cache.

 

[tyuhbh0]

You can use tools like free and ps/top to see what is consuming your memory. But there are many variables, are you using balloon memory, are you using PCIe passthrough, what CPU/memory/BIOS settings. Are you actually running out of memory, or are you just complaining about cache.

It was running well until run command "swapoff -a". i have total 16 VMs (including containers) total memory of all VMs is around 68GB. but when i see Summary of PVE i can see i can 128Gb or 121gb Ram using.

i don't understand what happened, sometimes "actually running out of memory" without changes anything.

i am not using PCi passthrough. i don't know what is happening.

 

[Impact]

Please stop full quoting and perhaps delete the existing ones. It's just noise and redundant. What does this look like?

top -co %MEM -bn1 -w260 | head -n60
arc_summary -s arc | grep -A3 "ARC size"

 

[Impact]

Please read what you just quoted. I won't help/respond further unless you stop that. ZFS' ARC is using 90G~ of your memory. See here how to limit it.

 

[tyuhbh0]

Please read what you just quoted. I won't help/respond further unless you stop that. ZFS' ARC is using 90G~ of your memory. See here how to limit it.

Thank you. After limit of ZFS' ARC to 1Gb. it works fine.

echo "options zfs zfs_arc_max=1073741824" >> /etc/modprobe.d/zfs.conf

update-initramfs -u

reboot

 

[Johannes S]

Thank you. After limit of ZFS' ARC to 1Gb. it works fine.

I doubt this. The ARC cache exists for a reason to speed up file operations. With over 100 GB RAM you can ( and should ) use this, non-used RAM is wasted. 16 GB ( the default on newer PVE Installations ) for the ARC cache should serve you well.
Btw: The reference to the doc on how to limit the ARC cache was already in the "not helpful" first link together with an explaination why this is actually a good thing.