[libvirt] Migration failure related to EOF from Monitor

Michal Privoznik mprivozn at redhat.com
Wed Feb 22 14:18:43 UTC 2012


On 22.02.2012 14:46, Shradha Shah wrote:
> Hello All,
> 
> I am trying to migrate a KVM guest while implementing the patches for PCI-Passthrough of SRIOV Vf's and
> I repeatedly see the following error:
> [root at c6100k cwd]# virsh migrate --live dibenchvm1 qemu+ssh://c6100l.uk.level5networks.com/system
> error: operation failed: migration job: unexpectedly failed
> 
> I have turned on debugging in /etc/libvirt/libvirtd.conf and I see the following debug in 
> /var/log/libvirt/libvirtd.log
> 
> 10:37:26.632: 2449: error : qemuMonitorIO:584 : internal error End of file from monitor
> 10:37:26.632: 2449: debug : qemuMonitorIO:617 : Error on monitor internal error End of file from monitor
> 10:37:26.632: 2449: debug : virEventPollUpdateHandle:145 : Update handle w=21 e=12
> 10:37:26.632: 2449: debug : virEventPollInterruptLocked:676 : Skip interrupt, 1 -1497479264
> 10:37:26.632: 2449: debug : qemuMonitorIO:640 : Triggering EOF callback
> 10:37:26.632: 2449: debug : qemuProcessHandleMonitorEOF:125 : Received EOF on 0x7f588c003540 'dibenchvm1'
> 10:37:26.632: 2449: debug : qemuProcessStop:3271 : Shutting down VM 'dibenchvm1' pid=18739 migrated=0
> 10:37:26.632: 2449: debug : qemuMonitorClose:765 : mon=0x7f588c00c7f0
> 10:37:26.632: 2449: debug : virEventPollRemoveHandle:172 : Remove handle w=21
> 10:37:26.633: 2449: debug : virEventPollRemoveHandle:185 : mark delete 9 20
> 10:37:26.633: 2449: debug : virEventPollInterruptLocked:676 : Skip interrupt, 1 -1497479264
> 10:37:26.633: 2449: debug : qemuProcessKill:3217 : vm=dibenchvm1 pid=18739 gracefully=0
> 10:37:26.633: 2449: debug : qemuProcessAutoDestroyRemove:3736 : vm=dibenchvm1 uuid=a4452511-0f97-734a-dbcc-f4c66f821956
> 10:37:26.633: 2449: debug : virSecurityDACRestoreSecurityAllLabel:504 : Restoring security label on dibenchvm1 migrated=0
> 
> May I ask for suggestions on how I can debug the monitor and find out the reason why the qemu monitor dies?

I've seen this many times esp. when using upstream qemu. I mean in
general when qemu dies during migration. Some info may be in domain log
under /var/log/libvirt/qemu/<domain>.log;

But if qemu dies there is not much libvirt can do to resurrect it.

Michal




More information about the libvir-list mailing list