Windows VM start-up hanging at 100% CPU

[rowid_alex]

Hello guys,

 

I am new for unRAID and still in the trail license. (I watched the videos from spaceinvader one for  quite some time, however).  Basically my machine is Intel Core 4790K + ASUS Maximums VII + RX480/580, which I see some success stories in our forum. With PCIe ACS override, I can passthrough my RX480/580 with my Intel IGP to the windows VM (one, win7 64, one win10 64) without any problem. My unRAID version is 6.4.0-rc8q.

 

As I have four cores (8 threads) so I reserve one core for unRAID and assign rest of 3 cores (6 threads) to VM as I don't start up multiple VM in the same time.

 

The strange thing is that it is very likely the Windows VM start-up will be hanging in the windows logo, and the CPUs assigned to VM just stay with 100% utilization.

 

 

In ssh, htop I can see the CPU are used by qemu-system-x86_64 KVM processes.

 

 

I tried to generate a call stack with the process, but I find no gstack utility is delivered with unRAID installation...

 

Based on my limited knowledge of Linux, I assume this should be some spin contention code in KVM due to whatever reason.

 

I attach the diagnostics archive as below.

unraid-diagnostics-20170927-1907.zip

 

How I work around the issue:

Just like the instruction or video said, windows VM might be not able to boot when doing the installation. Change to one core sovles the problem.

For me, I just reduce the cores (threads) from 3/6 to 2/4 or 1/2 then VM can boot without any problem.

Once it is started, if I changed the VM configuration back to 3/6, sometimes it works, but most of time it doesn't boot.

 

So my questions,

1. Is this a known issue, or due to some improper configuration of my machine?

2. How to install gstack and other system utility in unRAID, e.g. like apt-get or something else?

 

Thank you.

 

Best Regards,

Alex

[rowid_alex]

Somebody can check this?

[1812]

You're running a release client and not a stable version, so that could complicate things. If you're dead set on staying on the bleeding edge, then upgrade to rc9xxx and see if that resolves it. Also, look in the 6.4 prerelease section of the forum and see if anyone else has had this issue.

 

If not, then downgrade to 6.3.5, and change your machine type to 2.5 or 2.7 (whatever it has available.)