Re: qemu-kvm (1.0) sqeeeze-backports request to fix libvirt (0.9.8-2) and debian linux-ha clusters
On 21/02/12 07:21, Michael Tokarev wrote:
> On 20.02.2012 18:40, Jelle de Jong wrote:
>> What is the current status on fixing the regression?
>>
>> I still have the problems on my testing clusters and still have
>> pending updates on the production servers.
>
> Jelle, can you please awake and look around already?
>
> The initial issue has been fixed in squeeze version (0.12)
> and is now included into last squeeze point release, together
> with a few additional security issues. You updated your
> systems already, including other fixes which went into 6.0.4
> squeeze, right?
>
> Meanwhile, I uploaded backports of qemu and qemu-kvm into
> bpo60, together with backports of a few other required
> packages (including ipxe) last week, they're apparently
> still pending inclusion into real bpo60 archives.
I updated everything to the following versions:
corosync 1.4.2-1~bpo60+1
libvirt-bin 0.9.8-2~bpo60+2
pacemaker 1.1.6-2~bpo60+1
qemu-kvm 0.12.5+dfsg-5+squeeze8
I restarted the nodes to be sure they are all in the correct state.
I tried a migration of the quests, but they still failed. I tried to
migrate both ways. This normally takes a few seconds, but now it takes
more then five minutes and most of the migrations fails.
These are the /var/log/libvirt/libvirtd.log outputs:
http://paste.debian.net/157025/
# qemuMonitorTextGetMigrationStatus:1532 : operation failed: cannot
query migration status
# qemuDomainObjBeginJobInternal:789 : Timed out during operation: cannot
acquire state change lock
# qemuMonitorIO:560 : internal error End of file from monitor
On the #linux-ha the recommendation is to downgrade libvirt-bin since it
is still incompatible with the qemu-kvm version.
It may be a combination of things that are going wrong, I just wanted to
let you know the regression is still there and the clusters ain't
working like they did.
Appreciating all the hard work! I will just wait until libvirt-bin or
qemu-kvm gets updated. Hope this will be soon.
Kind regards,
Jelle
Reply to: