Proxmox draining RAM

[Otas]

Hello, i have trouble with draining ram. (including htop ss)
Any suggestion? After restart node everything ok, but wait 1 day and again take's all of ram.

 

[sb-jw]

Unused memory is wasted memory, so the Linux kernel attempts to use this memory to improve performance.

Check your RAM Usage with free -h and check for what it's used.

 

[Otas]

Check your RAM Usage with free -h and check for what it's used.

But when drain all of ram proxmox write much IO delay, and all VM's performance goes down.
When GUI show all ram was drained, free -h shows used 9gb free 500mb.

And htop shows much the same VM's its like proxmox dublicating VM's for performance, it is possible?

 

[sb-jw]

Did you overcommit your node? Can you post your HW Config and the VM RAM config?

 

[Otas]

Did you overcommit your node? Can you post your HW Config and the VM RAM config?

You mean this? No i dont overcommit, how i can do it?

 

[LnxBil]

But when drain all of ram proxmox write much IO delay, and all VM's performance goes down.

Do you mean that after you freed your memory, your IO delay increases and the performance goes does? If so, this is normal. Having free memory is a very bad thing for almost every computer system ever invented.

 

[Otas]

Do you mean that after you freed your memory, your IO delay increases and the performance goes does? If so, this is normal. Having free memory is a very bad thing for almost every computer system ever invented.

But if i have 10GB of ram and i have 2 VM's which i added for 2GB of ram, why after one day all 10GB ram drained and all performance goes down? If my VMs do nothing for all day, RAM stay good, but if my VM work with much load, seems like proxmox giving to it free RAM's and after much load VM catch only 2GB and used ~300MB but proxmox dashboard and htop shows what it drained

 

[LnxBil]

You have not shown us any output of a system that was "drained", so please do that.

 

[Otas]

You have not shown us any output of a system that was "drained", so please do that.

I am new at this so please tell me what i need to show and i will. (Including month ram usage, all less usage was node reboot)

 

[LnxBil]

Again: Used ram is good, not bad.

You say that your I/O delay goes up when the RAM is full, show it.
@sb-jw suggested that you show the whole output of free -h, not just some numbers, so please do that.

For text, it's best to paste it as text in CODE tags.

 

[Otas]

Hello,

Now i have that:
total used free shared buff/cache available
Mem: 9.7G 6.6G 2.8G 51M 203M 2.8G
Swap: 0B 0B 0B

Working 2 VM's
1 VM - 37.99% (777.98 MiB of 2.00 GiB)
2 VM - 70.91% (1.42 GiB of 2.00 GiB)

 

[LnxBil]

Thank you for using CODE tags. Everything looks normal, still no i/o delay output.

 

[Otas]

So if my 1 VM who have let to use 2GB ram eat 9GB ram is normal?
If yes, weird question, why i need to set how much ram VM need to use in first confg?

 

[Tapio Lehtonen]

From the info you have posted so far, I can not see a VM using more than 2 GB memory.
Memory is expensive so a well made OS does not let RAM sit idle, it is used for buffers, disk cache etc.. If you would create swap then you can see memory is exhausted when lots of swap is in use. The screenshots you showed do not look like memory is drained, whatever that means.
I do not fully understand what the memory issue is in this thread.
But the IO issue would be troublesome, but you have not showed any info about that except writing "much IO delay".

 

[LnxBil]

So if my 1 VM who have let to use 2GB ram eat 9GB ram is normal?

Best to read this:

https://www.linuxatemyram.com/

I do not fully understand what the memory issue is in this thread.
But the IO issue would be troublesome, but you have not showed any info about that except writing "much IO delay".

I totally agree.