Hi,

I have a test-guest image which I can't start after a host forced shutdown.

I'm using Cent OS 6.x on both host and guest using KVM

When trying to start in VMM it tries to resume the guest, but gets an error.

Is there a way to skip the resume and just boot the guest from scratch ? Or do a repair ? Using virsh i presume.

I think the problem is, that when the host went down, it tried to save the state of the guest to disk, but then i forced the host down. Testing what would happen.

If there's an issue here, it would be nice to know before production starts...

Regards.

You didn't mention what the guest OS is (Windows, Linux, FreeBSD, etc.). I assume the host is Linux, but which Distro? In fact, are you using VirtualBox, KVM or VMWare?

In any case, you may be able to download a Linux rescue CD image such as SystemRescueCD, RIP Linux, Finnix or similar and boot your VM guest with it.

Then you may be able to fsck the disk and mount it and look around for damage.

Ohh, I'm sorry.

I'm using Cent OS 6.x both as host and guest using KVM.

Can't virsh do the job ?

I had a similar problem not so long ago when the ups was configured to wait until 30% of battery capacity before shutdown. Linux was still completing a save of the guest image when it ran out of power. Since then I've configured the ups to initiate shutdown after 30 seconds.
The image was winxp pro & I did the following:
service libvirtd stop
top
kill qemu-kvm processes with -9
cd /var/lib/libvirt
rm -rf qemu
rm -rf network

Remove kvm
Re-install kvm

service libvirtd start

Removing & re-installing kvm may not be necessary but this procedure worked.

Well, I have other images which works correct, so I don't think removing kvm is a solution...

By the way, is 30 seconds enough to save the image ?

My image, 120 GB disk size, 10GB ram took alot more, then the power went down on the host in the middle of the saving guest image.

This sounds like a problem ??

If images gets corrupted, when power is gone, because it takes so much time to save the image.... then what to do ??

Is the image stored in /var/lib/libvirt/qemu/snapshot/?

If so, I would try renaming it or moving it, if not deleting it, and then trying to boot.

No, it's located under /home...

I meant the image of RAM, not the disk image.

You said it's trying to resume, which implies their is a RAM image somewhere. I don't know where it would be stored, but if you can find it, then rename or delete it.

1 members found this post helpful.

Ohh yeah... I will try to find out where it is... :-)

Found it.

/var/lib/libvirt/qemu/save

Thanks. :-)

Thomas, the 30 seconds refers to the time that the ups will begin its shutdown sequence after a power fail. It does not mean the time required to save an image which obviously takes a lot longer.

Oh, of-course, I understand.

The bad RAM image I just deleted was 1 Gb in size. The guest has configured 10 GB om ram, so I presume the RAM image to become 10 GB when saved during af host shutdown.

Is it normal it takes minutes to save the RAM image ? We are talking a brand new x3850 with 4 x 8core CPU's...

Yes it is.
You could always initiate a shutdown from the guest before the host if you want this to be faster. Means loading up a UPS app on the guest.
Network UPS Tools (NUT) will do this for you.

Okay... then it's possible to have the UPS software running on both the host AND the guest ? Reading from the same usb-port, to communicate with the UPS ?