Server 2022/2025 RDS Server Issues

[Jakest]

So, I'd like to start this off by saying that I firmly believe this is a Windows issue at this point, but I'm absolutely running out of places to ask, and I've had a ticket open with Microsoft for months now.

Whenever we build an RDS Server running server 2022 or server 2025, after about 2 weeks the printing subsystem breaks, entirely, right down to print to pdf, the only thing we've managed to do to fix it is rebuild the OS, or roll back to snapshots or backups before the breakage happens, It happens without any patching on the windows end.

The setup is 3 AMD Epyc 7452 servers, with separate 10g LACP networking for VM communications and storage communications, proxmox 8.4.1, though this issue started for us on earlier versions

We've had this happen to about 6 VM's at this point. We keep having to fall back to Server 2019 to have it stay stable

We use parallels RAS to connect to the RDS's
We've tried changing out the CPU, X86-64-v2-aes, X86-64-v3, and host.
The storage is attached via NFS

If anyone has seen anything similar and managed to fix it, or even has any ideas pointing in the right direction, I'd love to see if this could be resolved somehow

 

[SteveITS]

Any printer or just PDF? We ran into an error on Windows 11 back in May ish where they broke the MS PDF printer. Other things were fine but since we built the ISO with the broken code, PDF was broken and we wasted a bunch of time fixing it. Anything installed earlier and upgraded was fine. (This may not be relevant to Server)

 

[uzumo]

Does thick provisioning and disabling cached IO work?

Assuming the rollback resolves the issue, it likely means that the issue is either data corruption on the VM or VM IO is being affected.

 

[Jakest]

Any printer or just PDF? We ran into an error on Windows 11 back in May ish where they broke the MS PDF printer. Other things were fine but since we built the ISO with the broken code, PDF was broken and we wasted a bunch of time fixing it. Anything installed earlier and upgraded was fine. (This may not be relevant to Server)

Unfortunately, every single printer, including redirected printers, the whole thing just gives up printing entirely.

Does thick provisioning and disabling cached IO work?

Assuming the rollback resolves the issue, it likely means that the issue is either data corruption on the VM or VM IO is being affected.

Haven't thick provisioned, but disabling cached IO didn't solve it either. I'd also assume if we were running into IO issues, other VM's would start to show weirdness, it's worth noting that printing works fine forever on 2022 and 2025 when it's not an RDS host, and RDS works great on 2019. I also made 100% sure that our NFS traffic was TCP, to prevent any weird UDP issues.