[openstack-dev] [nova] bug 1334398 and libvirt live snapshot support

Kashyap Chamarthy kchamart at redhat.com
Mon Dec 8 10:45:36 UTC 2014


On Fri, Dec 05, 2014 at 02:12:44PM -0600, Matt Riedemann wrote:
> 
> 
> On 12/5/2014 1:32 PM, Sean Dague wrote:
> >On 12/05/2014 01:50 PM, Matt Riedemann wrote:
> >>In Juno we effectively disabled live snapshots with libvirt due to bug
> >>1334398 [1] failing the gate about 25% of the time.
> >>
> >>I was going through the Juno release notes today and saw this as a known
> >>issue, which reminded me of it and was wondering if there is anything
> >>being done about it?

As Dan Berrangé noted, it's nearly impossible to reproduce this issue
independently outside of OpenStack Gating environment. I brought this up
at the recently concluded KVM Forum earlier this October. To debug this
any further, one of the QEMU block layer developers asked if we can get
QEMU instance running on Gate run under `gdb` (IIRC, danpb suggested
this too, previously) to get further tracing details.

> >>As I recall, it *works* but it wasn't working under the stress our
> >>check/gate system puts on that code path.

FWIW, I myself couldn't reproduce it independently via libvirt alone or
via QMP (QEMU Machine Protocol) commands.

Dan's workaround ("enable it permanently, except for   
under the gate") sounds sensible to me.

> >>One thing I'm thinking is, couldn't we make this an experimental config
> >>option and by default it's disabled but we could run it in the
> >>experimental queue, or people could use it without having to patch the
> >>code to remove the artificial minimum version constraint put in the code.
> >>
> >>Something like:
> >>
> >>if CONF.libvirt.live_snapshot_supported:
> >>    # do your thing
> >>
> >>[1] https://bugs.launchpad.net/nova/+bug/1334398
> >
> >So, it works. If you aren't booting / shutting down guests at exactly
> >the same time as snapshotting. 

Tried this exact case independently, and cannot reproduce, as stated by
Dan (and others on the bug) in this thread.

> >I believe cburgess said in IRC yesterday
> >he was going to take another look at it next week.
> >
> >I'm happy to put this into dansmith's pattented [workarounds] config
> >group (coming soon to fix the qemu-convert bug). But I don't think this
> >should be a normal libvirt option.
> >
> >	-Sean
> >
> 
> Yeah the [workarounds] group

Is there any URL where I can read about this more?

> is what got me thinking about it too as a config option, otherwise I
> think the idea of an [experimental] config group has come up before as
> a place to put 'not tested, here be dragons' type stuff.

-- 
/kashyap



More information about the OpenStack-dev mailing list