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

Michael Still mikal at stillhq.com
Sat Sep 13 01:26:53 UTC 2014


I don't know why this is happening, but it sure seems like something you
should file a bug for.

https://bugs.launchpad.net/nova/+filebug

Thanks,
Michael


On Saturday, September 13, 2014, Stephen Cousins <steve.cousins at maine.edu>
wrote:

> Just confirmed that it isn't happening with 4-core VM's either. Just 8. I
> haven't tried a 16-core VM but I can guess.
>
> The flavors are matched with 1 GB per core so an 8-core VM has 8 GB of
> RAM. Does that lead to any ideas?
>
> On Fri, Sep 12, 2014 at 3:10 PM, Stephen Cousins <steve.cousins at maine.edu
> <javascript:_e(%7B%7D,'cvml','steve.cousins at maine.edu');>> wrote:
>
>> Thanks for the information. I just tried doing it with a VM with only 2
>> cores and it has been working fine. So it seems that it may be specific to
>> the number of cores that are involved. I'll try with 4 to see if that works
>> ok.  Thankfully we don't have many 8-core machines but we still have the
>> need to migrate them for maintenance.
>>
>> Does this core-specific information help to figure out what the problem
>> is?
>>
>>
>>
>> On Fri, Sep 12, 2014 at 2:31 PM, Jonathan Proulx <jon at jonproulx.com
>> <javascript:_e(%7B%7D,'cvml','jon at jonproulx.com');>> wrote:
>>
>>> On Fri, Sep 12, 2014 at 1:44 PM, Stephen Cousins
>>> <steve.cousins at maine.edu
>>> <javascript:_e(%7B%7D,'cvml','steve.cousins at maine.edu');>> wrote:
>>>
>>> > Does anyone live-migrate VM's successfully when volumes are attached?
>>>
>>> My setup is significantly different so probably not terribly helpful,
>>> but to answer the question I did do live migrations with Volumes
>>> attached on my Grizzly cloud (since upgraded to Icehouse).  We don't
>>> have shared storage for the ephemeral disks and our Volumes are iSCSI
>>> attached (from an equallogic san), so something like:
>>>
>>> nova live-migration --block-migrate <server> [<host>]
>>>
>>> I wouldn't say it was 100% but the volume piece wasn't a problem. Well
>>> once I figured out in my case I needed to flip some bits on the SAN
>>> side to allow multiple logins to the same target as source and
>>> destination are both connected during a migration, but I wouldn't
>>> expect that to be an issue with NFS
>>>
>>> -Jon
>>>
>>
>>
>>
>>
>
>
> --
> ________________________________________________________________
>  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
>
>

-- 
Rackspace Australia
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20140913/f0b4ffd0/attachment.html>


More information about the OpenStack-operators mailing list