[Openstack-operators] Grizzly: live-migration with attached volume?

Stephen Cousins steve.cousins at maine.edu
Fri Sep 12 17:44:19 UTC 2014


​Hi Abel,

Yes, it is the instance load. The console.log seems to get truncated to 0
bytes whenever I do live-migration. All machines (including the nfs server)
are using ntp. I just checked visually and all of their times match, at
least to the second. Probably it is closer.

I can't find much in the logs.

Does anyone live-migrate VM's successfully when volumes are attached?

Thanks,

Steve



On Fri, Sep 12, 2014 at 1:17 PM, Abel Lopez <alopgeek at gmail.com> wrote:

> That’s pretty interesting. I haven’t seen that before, tbh, I had been
> using gridcentric for live migrations in grizzly, but that’s just because
> we didn’t have shared storage.
> So, the instance is what is generating the load, right? anything
> interesting on the console log? I wonder if there is some time-sync issue
> that’s going haywire when it lands on the new host.
> But now I’m just speculating.
>
> On Sep 12, 2014, at 10:13 AM, Stephen Cousins <steve.cousins at maine.edu>
> wrote:
>
> Hi Abel,
>
> I just tried and it did the same thing. Load went up to 800 and the VM is
> unreachable.
>
> Any ideas?
>
> Thanks,
>
> Steve
>
> On Fri, Sep 12, 2014 at 12:59 PM, Abel Lopez <alopgeek at gmail.com> wrote:
>
>> Does it work if you first detach the volume and then migrate?
>>
>>
>> On Friday, September 12, 2014, Stephen Cousins <steve.cousins at maine.edu>
>> wrote:
>>
>>> I am fairly new to Openstack having just inherited a Grizzly system with
>>> NFS storage. "nova live-migration" is working unless the VM has an attached
>>> volume. When it doesn't work, the VM does move but it is not possible to
>>> connect to it and the CPU load on that VM goes up to the number of cores
>>> assigned to it. That is, for a VM with 8 cores in it, the load for that
>>> qemu-system-x86_64 process goes up to about 800%. Is this a known problem
>>> with Grizzly?
>>>
>>> Here is a description of our system:
>>>
>>>    three machines set up with HA for Quantum (with Open vSwitch),
>>> Cinder, RabbitMQ, and Nova. These nodes are all compute nodes too. We also
>>> have another machine that is just compute.
>>>
>>> Cinder and Glance storage is all NFS connected to each machine with the
>>> same mount points.
>>>
>>> ​The command I am using to do the migration is:
>>>
>>>     nova live-migration $UUID compute3
>>>
>>> In some cases the migration has worked when a volume has been attached.
>>> Most of the time however, it does not work.
>>>
>>> Thanks,
>>>
>>> Steve
>>>
>>> --
>>> ________________________________________________________________
>>>  Steve Cousins             Supercomputer Engineer/Administrator
>>>  Advanced Computing Group            University of Maine System
>>>  244 Neville Hall (UMS Data Center)              (207) 561-3574
>>>  Orono ME 04469                      steve.cousins at maine.edu
>>>
>>>
>
>
> --
> ________________________________________________________________
>  Steve Cousins             Supercomputer Engineer/Administrator
>  Advanced Computing Group            University of Maine System
>  244 Neville Hall (UMS Data Center)              (207) 561-3574
>  Orono ME 04469                      steve.cousins at maine.edu
>
>
>


-- 
________________________________________________________________
 Steve Cousins             Supercomputer Engineer/Administrator
 Advanced Computing Group            University of Maine System
 244 Neville Hall (UMS Data Center)              (207) 561-3574
 Orono ME 04469                      steve.cousins at maine.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20140912/21944084/attachment.html>


More information about the OpenStack-operators mailing list