Hello.  I have been able to setup proxmox to run my unRaid 7.0.1 server and all works as expected.  Day to day it runs very well but when I go to do a parity check/rebuild it seems like the array maxes out at 25MB/s even after running for almost an hour.  If I run the same setup on the same hardware without being virtualized in Proxmox I get much higher speeds (90MB/s at start and scaling up before I cancel).  I have a 14 TB array so you can see why this would just be torture to wait the 6+ days for party check/disk rebuild.  

 

My system is a Supermicro H11SSL-i (latest firmware) with a 7551p and 512GB of ECC memory.  It's not the fastest system but the cores and memory allow me to tinker which I love to do.

 

I am currently running all of my HDD's (24 total) on two HBAs (LSI 9300 16i, all 4 chips updated to latest firmware) and they are passed through to the unRaid VM.  Im wondering if it's the fact that the drives are on two different HBAs that's the issue?  I say this because any couple of drives doing something seem to achieve good (or at least good enough) speeds but it's only when they all are doing something it seems to cause a problem. I also ran disk speed test and the results looked fine to me.

 

Attaching my diagnostics with parity running in the VM with docker disabled to see if anyone is able to assist.

 

Thanks for your time and hope I am not doing anything too stupid here

unraid-nas-diagnostics-20250422-2012.zip

Edited April 23, 20251 yr by mattym0

Is your CPU type set to "host" in Proxmox?

At the time I did that diagnostic it wasn’t as I was messing around with different options but yesterday I had it as host and it has the same symptoms. I can rerun diagnostics with it set to host if that is helpful.

It seems you've already isolated the problem to virtualisation, as you said bare metal runs at expected speeds so I would be more interested in your proxmox vm config than the Unraid config.

A screenshot of your proxmox vm's Hardware tab would probably be the most useful

Edited April 23, 20251 yr by tjb_altf4

Sure thing. Here it is with x86-64 as the processor type.  Top 4 PCI Devices are the Two HBAs and 4 after that are 4 NVMe's used for cache.  Last PCI Device is GPU obviously.
  

 

But I also changed it to host and reran just to confirm.

 

Pegs itself right at 25MB/s on all drives every time

 

Your setup is similar to mine, a few things to try:
setting the numa flag for cpu


explicitly disabling memory ballooning (which won't normally work with pci passthrough anyway)


Possibly you also could try falling back to the default display implementation with noVNC.

You may also like to enable the guest agent on the Unraid vm for better controls and some info like IP in proxmox

 

Thanks for the info.  Implemented all of that (and the qumu agent is super useful so thank you!).  No difference in speeds.

 

Anyone else have any ideas?  I wondered if diagnostics would show anything in the logs for why the speeds were slow was the reason I had uploaded it before.  Especially as when I look at the diskspeed docker from within the unraid vm I see:

 

 

Which all look fine and at the very least faster than the 25 Im getting