Memory spikes win 2022 trying to defragment a 1TB disk

[alejoga81]

Hi,

I have a mv with windows 2022 in pve 7.3-3, a week ago I added a 1TB disk, but the operating system eats all the RAM, when entering the OS I identified that what is eating the memory is a process that tries to deframe the disk



I have been adding to the RAM but it always has the same behavior, This is my disk information



The storage es type ZFS



this the disk from OS



I stopped the process of optimizing the drive to free up memory



I appreciate if someone can help me and recommend a solution, I have already visited several blogs and followed best practice guides but I have not been able to solve it.

 

[cwt]

You should enable the „Discard“ feature (below the dropdown „Default (no cache)“ and see if it‘s getting to normal RAM consumption level. Afaik you‘ll have to shutdown the VM before activating this.

 

[alejoga81]

Hi
Thanks for your respose, I enabled the Discard, The server stabilized, but I copied about 30 GB more information and within an hour it consumed all the memory again, both in the virtual machine and in the node.

 

[cwt]

There are some points which are pretty normal when running Windows VMs on ZFS:

- ZFS uses RAM for caching (ARC) (up to 50% per default)
- Windows also „eats“ all available/configured RAM as shown in the GUI, but within the VM itself the task manager should display the actual RAM consumption

From the wiki:

Limit ZFS Memory Usage​

ZFS uses 50 % of the host memory for the Adaptive Replacement Cache (ARC) by default. Allocating enough memory for the ARC is crucial for IO performance, so reduce it with caution. As a general rule of thumb, allocate at least 2 GiB Base + 1 GiB/TiB-Storage. For example, if you have a pool with 8 TiB of available storage space then you should use 10 GiB of memory for the ARC.

 

[alejoga81]

ok, but I only have a storage of 1 TB and if you notice I have the virtual machine with 20 GB, additionally in the moments that win2k22 does the disk framentation it uses 100% of the memory available in the virtual machine and also occupies the 100% of the node memory, look at the following image, the node memory is used at 100%, that is, 40 GB

 

[cwt]

It seems that it’s not an uncommon problem:

https://forum.proxmox.com/threads/windows-vm-requires-lots-of-ram-to-trim-larger-hard-disks.96288/

How does your underlying ZFS setup look like? Drive types, RAID mode, storage?

 

[totalizator]

I have observed the same with my Win10 VM that uses 1TB disk on a ZFS storage (NVME, no RAID). Memory usage spikes every time the Service Host: Disk Defragmenter becomes active.

 

[_gabriel]

edit: indeed, set discard_granularity is the correct fix.
Thanks @ky41083 for sharing link where is tip for global set discard_granularity.

btw, I wonder if the SSD model has something to do with it.

a side fix, downgrade virtio scsi to version 0.1.208 https://forum.proxmox.com/threads/r...device-system-unresponsive.139160/post-691337

 

[ky41083]

This was solved in the thread linked to in post 6