[libvirt] Trying to debug "Received unexpected event 3" from libvirt

Sat Jan 16 17:02:56 UTC 2016

On Fri, Jan 15, 2016 at 9:37 AM, Michal Privoznik <mprivozn at redhat.com>
wrote:

> On 14.01.2016 19:51, Yaniv Kaul wrote:
> > On Fri, Jan 8, 2016 at 7:00 PM, Michal Privoznik <mprivozn at redhat.com>
> > wrote:
> >
> >> > On 30.12.2015 11:30, Yaniv Kaul wrote:
> >>> > > Hi,
> >>> > >
> >> >
> >> > Hey,
> >> >
> >> > sorry for getting to you so late.
> >> >
> >>> > > I'm trying to debug this issue, which may be affecting my
> inability to
> >>> > > perform live snapshot.
> >>> > > 1. I'm not sure what 'Waking up a tragedian" in the debug log
> means -
> >> > what
> >>> > > exactly is a tragedian?
> >> >
> >> >
> >> > It's the thread that has issued the change state API (shutdown,
> reboot,
> >> > ..) and it's waiting for confirmation on the monitor. For instance,
> mgmt
> >> > app issues virDomainPMSuspendForDuration() which in qemu driver is
> >> > implemented via some agent calls. So the flow is like this:
> >> >
> >> > 1) the control gets to qemuDomainPMSuspendForDuration()
> >> > 2) libvirt does some checks and issues 'guest-suspend-disk' command
> (or
> >> > corresponding command to the selected target).
> >> > 3) qemu-ga running inside guest tries (!) to suspend the guest (it may
> >> > not necessarily succeed)
> >> > 4) meanwhile, as guest is writing something onto disk (saving its RAM
> -
> >> > but one is unable to tell from outside), the libvirt API is blocked
> >> > 5) finally, guest kernel calls 'HALT' to which qemu responds by
> sending
> >> > libvirt 'RESET' event
> >> > 6) the libvirt event loop finds out that an event occurred on the
> domain
> >> > monitor and calls callback
> >> > 7) the callback will wake up the sleeping API if the event the API is
> >> > waiting for matches the one obtained on the monitor
> >> >
> >>> > > 2. In any case, it'd be great if the WARN would mention
> mon->await_event
> >> > -
> >>> > > is it the event libvirt is actually waiting for?
> >> >
> >> > Sure, that would be much more helpful - mind posting a patch?
> >> >
> > Attached and tested. I can't post it properly, as the git is a bit
> 'dirty'
> > with some po files that I can't clean:
> >
> > On branch workbranch
> > Your branch is up-to-date with 'origin/master'.
> > Changes not staged for commit:
> >   (use "git add <file>..." to update what will be committed)
> >   (use "git checkout -- <file>..." to discard changes in working
> directory)
> >
> >         modified:   po/ar.po
> >         modified:   po/as.po
> >         modified:   po/bg.po
> >         modified:   po/bn.po
> >         modified:   po/bn_IN.po
> >         modified:   po/bs.po
> >         modified:   po/ca.po
> >         modified:   po/cs.po
> >         modified:   po/cy.po
> >         modified:   po/da.po
> >         modified:   po/de.po
> >         modified:   po/el.po
> >         modified:   po/en_GB.po
> >         modified:   po/es.po
> >         modified:   po/et.po
> >         modified:   po/fi.po
> >         modified:   po/fr.po
> >         modified:   po/gl.po
> >         modified:   po/gu.po
> >         modified:   po/he.po
> >         modified:   po/hi.po
>
>
> I guess you were doing 'make dist' or 'make rpm'. Both of them result in
> generating of translation strings. Anyway, you can just drop them and
> have clean working tree: git checkout po/
>

Indeed, I've used 'make rpm', as I could not get 'sudo make install'
correctly install libvirt on my Fedora - failed to get the right
directories set in the './configure' script.
I wish it would have auto-detected Fedora and used the right defaults.
After failing to do so normally, I went and look again into libvirt.org -
where I failed to find 'how to contribute' document.
Only today I've found http://libvirt.org/compiling.html - which has a hint
to use ''--system' to autogen.sh - perhaps it should be on by default?

That's part of a bigger problem - no 'how to contribute' section that
easily found on the site. Using search, I've found first hit to an email
archived[1] which pointed me to [2] - which is the right document, probably
in the wrong place (under internals?). And that one didn't have the hint
above...

[1] http://www.redhat.com/archives/libvir-list/2014-April/msg00004.html
[2] http://libvirt.org/hacking.html

> >
> > ...
> >
> >  Thanks,
> > Y.
> >
> >
> >>> > >
> >>> > > (Both from qemu/qemu_agent.c)
> >>> > > 3. I reckon event 3 is QEMU_AGENT_EVENT_RESET ? (from
> qemu/qemu_agent.h)
> >>> > >
> >> >
> >> > Correct.
> >> >
> >>> > > 4. I'm also getting 'End of file while reading data: Input/output
> error'
> >>> > > messages, not sure what they mean yet.
> >> >
> >> >
> >> > Usually they mean crashing daemon. If you are able to get a stacktrace
> >> > please do share it somewhere.
> >> >
> >> > Michal
> >> >
> >
> >
> > qemu_agent.c.diff
> >
> >
> > diff --git a/src/qemu/qemu_agent.c b/src/qemu/qemu_agent.c
> > index f979f82..924c177 100644
> > --- a/src/qemu/qemu_agent.c
> > +++ b/src/qemu/qemu_agent.c
> > @@ -1241,7 +1241,7 @@ void qemuAgentNotifyEvent(qemuAgentPtr mon,
> >          }
> >      } else {
> >          /* shouldn't happen but one never knows */
> > -        VIR_WARN("Received unexpected event %d", event);
> > +        VIR_WARN("Received unexpected event %d (expected %d)", event,
> mon->await_event);
> >      }
> >  }
> >
>
>
> ACKed and pushed. I've committed the patch under your name. I hope
> that's okay with you.
>

Yes, much appreciated - tried to send the patch but perhaps it doesn't work
easily via GMail.
Y.

> Michal
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/libvir-list/attachments/20160116/dd9bc5ac/attachment-0001.htm>