Proxmox not releasing memory

[zemzema]

I check node with two VM on in, and there is that consumption of memory is very high. When you check VM that are running on than server, they are using very small amount.

Check screens.

On first picture is Host, on other two are VM, and fourth is htop memory consumption on Host.

 

[LnxBil]

You are running two 8 GB VMs on only 12 GB of RAM?

 

[zemzema]

Yes, but that machine is using around 2GB for normal run. There was a peek to 8GB, and after two second machine returned to normal, but host server didn't release memory consumption. On the end I restart hypervisor and memory was back to normal.

 

[josedavidg]

I am seeing the same issue, I am running manually these commands:

oot@proxmox:~# sync; echo 1 > /proc/sys/vm/drop_caches root@proxmox:~# sync; echo 2 > /proc/sys/vm/drop_caches root@proxmox:~# sync; echo 3 > /proc/sys/vm/drop_caches

According to the screenshot attached it drop from 6GB to 4 GB ram utilization after this. I am unsure if this will be a good practice, more likely not, but I am considering scheduling a job to release memory every hour or so until I find the root cause of such behavior.

 

[josedavidg]

I am seeing the same issue, I am running manually these commands:

oot@proxmox:~# sync; echo 1 > /proc/sys/vm/drop_caches root@proxmox:~# sync; echo 2 > /proc/sys/vm/drop_caches root@proxmox:~# sync; echo 3 > /proc/sys/vm/drop_caches

According to the screenshot attached it drop from 6GB to 4 GB ram utilization after this. I am unsure if this will be a good practice, more likely not, but I am considering scheduling a job to release memory every hour or so until I find the root cause of such behavior.

View attachment 23591

I was wrong and it does not seem to be an issue, is just a miss understanding in how the memory is being reported. Although it shows as assigned, it is on DISK CACHE. Please visit these sites: https://tldp.org/LDP/sag/html/buffer-cache.html and: https://www.linuxatemyram.com/ to understand how it works.

 

[Dunuin]

Yes, thats normal and there is no way to limit the size of that buffer cache. Linux will eat all the RAM you throw at it so overcommitting RAM isn't really working.