I've got an Unraid server I've been using for years..  I have 6 or so docker containers that run with no problem, but I now have one VM that I'd like to run, but every time I start it, the whole unraid performance begins to degrade over the next couple hours to the point that even the webpage wont render.  I can still RDP into the Windows VM typically, and if I shut it down, the Unraid server will slowly become responsive again, although it takes an hour or so..  My server has a Ryzen 5 5600G  CPU and 16gb RAM.  The CPU  has 6 cores with 12 threads.  It doesn't matter if I only assign the VM one thread - It still has the same problem.

 

The only odd thing is that I do have 8 total virtual hard drives, each a 3tb file on a spinning drive which is a member of the array, but it doesn't matter if I am using the VM or not, and the problem doesn't start immediately, so I dont think this is related.

 

I'm not sure where to start troubleshooting with this.  Is there some CPU parameter on the VM that could be doing this?  Has anyone else experienced this?  I can't find anything similar searching the forum or just generically googling the problem on the Internet at large.

 

Thanks in advance for any help!

 

One thing I forgot to mention..  Although my secondary drives are on spinning drives, the C Drive for the OS is on an SSD backed cache pool of 2 drives in a mirror..

After posting this 2 hours ago, I brought the VM up, and now although the VM is running fine, Unraid is completely unusable..  No screen updates, took 10 minutes for "top" to come up on an ssh session which was already established..  Here's what I see..  Why would the qemu process be taking so much CPU if the VM is only configured for 4 threads (2 cores)? And, it doesn't instantly do this.. it built up over the 2 hours..  

 

So strange!

 

:  

How does it act if you assign the VM 4GB of RAM?

On 9/29/2023 at 7:09 PM, JonathanM said:

How does it act if you assign the VM 4GB of RAM?

Thanks for your reply, and sorry, somehow I didn't get notified of this post 'till now.

RAM doesn't seem to affect it.   I had made it higher and lower with no noticeable difference.  I can't say I specifically selected 4gb so I just set it to that, and started the VM.  

The VM's purpose is to back up data from various workstations around my house so I can use Backblaze on the VM to back up to the cloud, so I started the backup batch file (uses robocopy)

It's been running 10 minutes or so.  So far it's still responsive and the one core assigned to the VM has been pegged the whole time, but the other cores are inching up slowly and I'd expect it to be unresponsive on the "UnRaid side of things" soon.  I'll keep an eye on it and report results.  

Watching top, of course the qemu thread is using most, but it looks like "agent", "notify", and "shfs" threads are the other heavy hitters.

 

I am kind of assuming that while the VM is running, since the underlying VM disks are spread on several separate physical drives, it is slowing the server down as it tries to keep up with parity calculations, but I'm not sure why that should keep the gui from being responsive.

 

I had previously intentionally spread the virtual disk files across disks so as to try to minimize contention on one disk, but I think I'll try to try to consolidate the disk files onto a smaller number (maybe 1!) of disks and see if perhaps it will help the overall server perhaps at the risk of a hit to the performance of the VM itself, which is fine with me.  I dont care if a backup cycle takes a whole week to run, but I dont want it to impact my use of the rest of the UnRaid while it happens!

 

Thanks!