[ovirt-users] migration failures

[Michael Watters]

On 02/23/2017 10:28 AM, Francesco Romani wrote:
> On 02/23/2017 04:20 PM, Michael Watters wrote:
>> I have an ovirt cluster running ovirt 4.0 and I am seeing several errors
>> when I attempt to put one of our nodes into maintenance mode.  The logs
>> on the source server show errors as follows.
>>
>> Feb 23 10:15:08 ovirt-node-production3.example.com libvirtd[18800]: operation aborted: migration job: canceled by client
>> Feb 23 10:15:08 ovirt-node-production3.example.com libvirtd[18800]: internal error: qemu unexpectedly closed the monitor: 2017-02-23T15:12:58.289459Z qemu-kvm: warning: CPU(s) not present in any NUMA nodes: 2 3 4 5 6 7 8 9 10 11 12 13 14 15                                                                               
>>                                                                                2017-02-23T15:12:58.289684Z qemu-kvm: warning: All CPU(s) up to maxcpus should be described in NUMA config                                                                                                                                                 
>>                                                                                2017-02-23T15:15:07.889891Z qemu-kvm: Unknown combination of migration flags: 0
>>                                                                                2017-02-23T15:15:07.890821Z qemu-kvm: error while loading state section id 2(ram)
>>                                                                                2017-02-23T15:15:07.892357Z qemu-kvm: load of migration failed: Invalid argument
>>
>> This cluster does *not* have NUMA enabled so I am not sure why this
>> error is happening. 
> It's one implementation detail. NUMA is enabled transparently because it
> is required for memory hotplug support.
> It should be fully transparent.
>
>
>> Some migrations did succeed after being restarted
>> on a different host however I have two VMs that appear to be stuck.  Is
>> there a way to resolve this?
> The load/save state errors are most often found when the two sides of
> the migration have different and incompatible version of QEMU.
> In turn, this is quite often a bug, because forward migrations (e.g.
> from 2.3.0 to 2.4.0) are always supported, for obvious upgrade needs.
>
> So, which version of libvirt and qemu do you have on the sides of
> failing migration paths?
>
> Bests,
>