[OmniOS-discuss] Ang: Re: r151014 KVM crash

[Dan McDonald]

> On Apr 6, 2015, at 10:36 AM, Johan Kragsterman <johan.kragsterman at capvert.se> wrote:
> 
> 
> Hej!
> 
> 
> I mean the KVM process crashes or get killed or whatever. The KVM virtual machine is suddenly stopped, and can't get started again, until reboot of the OmniOS machine. The Linux OS on the virtual machine doesn't crash.
> 
> I guess I need to get back to the r151014 again to find out more, but I wanted to roll back to r151012 to see the difference. And as I told you, there were no crashes on the r151012.
> 
> When I done some update jobs here on r151012, I'll get back to r151014 and try to find out more...

I'm talking with the illumos KVM folks.  They mentioned that Ivy Bridge Xeons (i.e. Xeon E5-26xx v2, where v2 means Ivy Bridge) have erratum that can cause problems.  That you do not seem to see these in r151012, however, is very odd.

There was only ONE change in the illumos-kvm kernel bits between releases, and it should have nothing to do with the errors you're seeing.

	commit 8bdb134ab2cbdd179cd0af6b4c4acf21af34fb87
	Author: Robert Mustacchi <rm at joyent.com>
	Date:   Wed Dec 10 21:22:50 2014 +0000

	    HVM-812 QEMU build fails after eventfd

That change was very tiny.

They also asked for kernel panics on your guests, but you say they aren't panicking, just stopped?

Do you have coredump from the kvm processes by any chance?  That would ALSO be useful.

Dan