(SOLVED) VM Win 10 Keeps Pausing or Will not start

[italeffect]

My Windows 10 VM that runs Blue Iris has been working for over a year. Recently it started only running for 5-10 minutes and then ending in a paused state.  All my attempts to fix it have not helped and often I can't even get it to start. 

It's passed 2 cores, 12 Gb of RAM, the IGPU from my 8700k and has two vdisks, both stored on unassigned devices. One SSD and one HD. It has been a great setup and has run with no issues for a long time.

In attempts to fix, I have:

recreated the VM

turned VMs on/off

Updated Unraid from 6.8.0-rc7 (which I ran since post with no issues) to 6.8.2

Turned off PCIe ACS override: Downstream and VFIO allow unsafe interrupts:Yes (no change)

Changed machine type between i4440fx-3.0 (original), i4440fx-4.2, Q35-4.2  

Changed from 2 cores to 1

Removing iGPU and booting VNC only (freezes at login or won't connect at all)

 

The one set of log messages I'm seeing after a couple minutes of launching the VM is:

 

Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop
Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop
Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop
Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop
Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop
Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop
Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop
Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop
Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop
Feb 22 15:12:08 unRAID kernel: kvm [1007]: vcpu2, guest rIP: 0xfffff80108dc1a52 kvm_set_msr_common: MSR_IA32_DEBUGCTLMSR 0x1, nop

 But not really seeing much else in terms of errors.

 

I could really use any help on what direction to go in please.

Diagnostics attached.

Thanks!

 

unraid-diagnostics-20200222-0914.zip

[JorgeB]

VMs usually pause when the device containing a vdisk runs out of space, you have one unassigned disk 100% allocated, is it used on that VM?

[italeffect]

Thanks for getting back to me. Yes the 2nd vdisk is on a 10TB HD and is the only item on the disk. I think the size is 9.9TB. Inside the windows VM it shows 2TB used out of 9.9TB. Do I need to resize this disk smaller inside the windows VM to leave some space on the disk in unassigned devices?

 

FYI The boot vdisk is on an SSD and has 50GB free inside the VM and ~300GB free looking at the drive in unassigned devices.

 

As of now I can't get the VM to boot at all, it just hangs at some point during boot. Going to see if I can boot a backup copy from a couple months ago.

[Squid]

1 minute ago, italeffect said:

2TB used out of 9.9TB

Not relevant to the size of the vdisk unless you do this 

 

3 minutes ago, italeffect said:

Yes the 2nd vdisk is on a 10TB HD and is the only item on the disk.

Since that disk is 100% full, basically you either need to do the above which keeps the vdisk the same size as what windows reports is used, or shrink the vdisk.

 

 

Basically by default, the vdisk no matter what you do will always expand to the size you've set for it.

 

IE: say from a fresh install of Windows and a new vdisk Windows says that 1TB is used.  the vdisk size on the hard drive will also be 1TB

 

Now copy 1 TB of files to the VM.  Windows will say that the used space is 2TB, and the vdisk will also be 2TB.

 

Delete those 2TB of files, windows will report 1TB used, but the vdisk is still 2TB

 

Copy over another 1TB of files and windows says 2TB used, but the vdisk now takes up 3TB  and so on until the vdisk takes up the maximum size you've specified or you run out of space on the drive storing it (which is what's happened here due to overhead etc)

 

 

Oh yeah, that drive that you think is 10TB isn't.  It's only an 8TB (7.7TB after formatting)

[italeffect]

1 hour ago, Squid said:

 

Since that disk is 100% full, basically you either need to do the above which keeps the vdisk the same size as what windows reports is used, or shrink the vdisk.

 

 

Thanks very much for the explanation. I'm surprised I haven't run into this earlier since this is my video storage vdisk for Blue Iris and I've let it fill up (6-7TB) several times over the last several months before clearing it out. 

 

I removed the 2nd vdisk and the VM booted right away. I'm assuming i need to run qemu-img resize to shrink the vdisk, but since I can't boot into windows with the disk first to shrink the file system, what are my options? Do I need to just delete the storage vdisk and start over? I have most of the data I need backed up off it so it's not really a huge loss. Or can I just run the resize command anyway since it's not the boot vdisk?

 

Is it a more sane setup to just pass the whole drive to the VM and use it that way? 

 

I found your directions on using vitio-scsi controller together with discard='unmap', so thanks for that and I'll get that setup.

 

[Squid]

3 minutes ago, italeffect said:

Is it a more sane setup to just pass the whole drive to the VM and use it that way? 

 

If the drive is dedicated just for the VM, then yeah that would make more sense to me.

[italeffect]

Thanks again for the help. For the sake of simplicity I just reformatted the 8TB disk and passed it directly to the VM. Problem solved.

[lankster2k]

I don't know if this helps but I had a similar issue on a Windows 10 VM with a GPU attached.

(Unraid 6.10-rc2 Windows 10 with nvidia gpu. SeaBIOS used otherwise always goes into bios on boot)

I made sure VM power management was set to never turn off and it would still go paused erratically.

 

Changing the machine type to a lower number seems to have solved it. It was on i440fx-6.1 but now with i440fx-3.0 no issues so far (its been 2 days). I haven't tried another other versions and I don't know what benefits I've lost by using the lower level of 3.0 but as it working ...

Cheers