Windows VM's are Extremely Laggy

[jwsl224]

decently spec'ed proxmox host goes under severe i/o load running windows vm. all the vm's on the server become mostly unusable because one vm is running 2 firefox tabs. i have no idea how to start troubleshooting this, but it shouldn't be this had to run a VM.



this results in an unusable amount of lag.

host:
CPU: Xeon Gold 6336y
Memory: 320 GB DDR4-3200 ECC RAM
storage for this VM: ZFS Raid10, 4x Crucial T700 1TB

vm config:

agent: 1
balloon: 0
bios: seabios
boot: order=scsi0;net0
cores: 32
cpu: host
hostpci0: 0000:8a:00,pcie=1
machine: pc-q35-8.0
memory: 24000
meta: creation-qemu=8.1.5,ctime=1723473166
name: Admin
net0: virtio=BC:24:11:23:E3:3A,bridge=vmbr0,firewall=1,tag=51
numa: 0
ostype: win11
scsi0: nvme5:vm-121-disk-0,cache=writeback,discard=on,iothread=1,size=50G,ssd=1
scsihw: virtio-scsi-single
smbios1: uuid=8230bf6b-1aa2-4562-ad64-53dc1f5e57dd
sockets: 1
tpmstate0: local-zfs:vm-121-disk-0,size=4M,version=v2.0
vmgenid: a9d6de2a-6124-494a-b56a-4a10b0b05a61

also, i followed the windows vm bast practices, found here

what the heck is left to do? this host is rendered useless simply from running a windows vm.
Also, the zfs config is still at default values. so the dirty data max would be at 4GB, and the ARC would be 16 GB.

here is the zpool status of the boot pool, as well as the storage pool this vm is on, pool nvme5

pool: nvme5
 state: ONLINE
  scan: scrub repaired 0B in 00:22:44 with 0 errors on Sun Sep  8 00:46:47 2024
config:

        NAME                                                STATE     READ WRITE CKSUM
        nvme5                                               ONLINE       0     0     0
          mirror-0                                          ONLINE       0     0     0
            nvme-CT1000T700SSD3_2326E6E7EAD7                ONLINE       0     0     0
            nvme-CT1000T700SSD3_2326E6E7EAC6                ONLINE       0     0     0
          mirror-1                                          ONLINE       0     0     0
            nvme-CT1000T700SSD3_2326E6E7EBED                ONLINE       0     0     0
            nvme-uuid.9d1ea987-d8bb-4f3f-9cab-37db52eeb07a  ONLINE       0     0     0

errors: No known data errors

  pool: rpool
 state: ONLINE
  scan: scrub repaired 0B in 00:00:54 with 0 errors on Sun Sep  8 00:24:58 2024
config:

        NAME                                 STATE     READ WRITE CKSUM
        rpool                                ONLINE       0     0     0
          mirror-0                           ONLINE       0     0     0
            nvme-eui.002538bb11500c72-part3  ONLINE       0     0     0
            nvme-eui.002538bb11500c6a-part3  ONLINE       0     0     0

errors: No known data errors

 

[_gabriel]

There is an issue , try downgrade virtio-scsi driver to version 0.1.208, within Windows Device Manager.
https://forum.proxmox.com/threads/r...device-system-unresponsive.139160/post-691337

btw, curious about your fsync and fio on file in a ZFS file dataset (--direct=1 --sync=1 --rw=randwrite --bs=4K --numjobs=1 --iodepth=1)

 

[jwsl224]

jeepers.... the 0.1.208 drivers work much much much better. they also seem to have solved a bunch of the issues from this thread. meaning, even if the host is slow on writes after the cache is full, the VM's no longer completely freeze up. on top of better performance on the host. this is quite something.
questions:
am i expected to go back and manually downgrade all 18 of our windows vm's?
what do you do when you can no longer trust the official drivers? i mean....rolling back more than 5 releases....jeepers..

this has cost me two whole weeks issues without any troubleshooting progress. jeepers. i have no idea why the forum you included above didn't come up in my search on this issue. this is crazy..