[ovirt-users] VM get stuck randomly

[Christophe TREFOIS]

Hi,

Just to give a small update.

It seems that when installing the willy-lts-kernel backport package on 14.04, the VMs stopped getting stuck.

We have now been running 6 weeks without a single VM crashing.

Before, we had 1-2 per day.

Maybe it will be useful to somebody.

Kind regards,

—
Christophe  

> On 30 Mar 2016, at 14:23, Christophe TREFOIS <christophe.trefois at uni.lu> wrote:
> 
> Hi Kevin,
> 
> Ok thanks for the feedback.
> 
> Do you have experience with kernel panics and more specifically how to get any meaningful message if this is happening?
> 
> According to syslog on the guest, at least after reboot, there's no information that something went wrong.
> 
> Since this seems to leave the qemu world, I'm now not sure where to ask for support in this :(
> 
> Kind regards,
> 
> --
> Christophe 
> Sent from my iPhone
> 
>> On 30 Mar 2016, at 14:14, Kevin Wolf <kwolf at redhat.com> wrote:
>> 
>> Hi Christophe,
>> 
>> Am 30.03.2016 um 13:45 hat Christophe TREFOIS geschrieben:
>>> Another host went down, so I have to prepare info for this one.
>>> 
>>> I could not SSH to it anymore.
>>> Console would show login screen, but no keystrokes were registered.
>>> 
>>> I could “suspend” the VM and “run” it, but still can’t SSH to it.
>>> Before suspension, all QEMU threads were around 0%, after resuming, 3 of them hover at 100%.
>>> 
>>> Attached you could find the gdb, core dump, and other logs.
>>> 
>>> Logs: https://dl.dropboxusercontent.com/u/63261/ubuntu2-logs.tar.gz
>>> 
>>> Core Dump: https://dl.dropboxusercontent.com/u/63261/core-ubuntu2.tar.gz
>>> 
>>> Is there anything else we could provide?
>> 
>> This sounds much like it's not qemu that hangs (because then stopping
>> and resuming wouldn't work any more), but just the guest OS that is
>> running inside the VM.
>> 
>> We've had cases before where qemu was reported to hang with 100% CPU
>> usage and in the end it turned out that the guest kernel had panicked.
>> Can you check whether a guest kernel crash could be the cause? If this
>> is reproducible, maybe the easiest way would be to attach a serial
>> console to the VM and let the kernel print its messages there.
>> 
>> Kevin

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 3509 bytes
Desc: not available
URL: <http://lists.ovirt.org/pipermail/users/attachments/20160616/4d704dfd/attachment-0001.p7s>