Hi Josh,

Thanks for the info.  So if I want to do live migration with VMs that were
launched with boot-from-volume, I'll need to use virsh to do the migration,
rather than Nova.  Okay, that should be doable.  As an aside, I will
probably want to look at the OpenStack DB and figure out how to tell it
that the VM has moved to a different host.  I'd rather there not be a
disconnect between Nova and libvirt about where the VM lives.  =)

Additionally, thanks for saying that the migration is safe with the RBD
cache enabled.  I was going to ask that as well.

On Tue, Mar 12, 2013 at 4:38 PM, Josh Durgin <josh.dur...@inktank.com>wrote:

> On 03/12/2013 01:28 PM, Travis Rhoden wrote:
>
>> Thanks for the response, Trevor.
>>
>>
>>  The root disk (/var/lib/nova/instances) must be on shared storage to run
>>> the live migrate.
>>>
>>>  I would argue that it is on shared storage.  It is an RBD stored in
>> Ceph,
>> and that's available at each host via librbd.
>>
>
> Agreed.
>
>
>  You should be able to run block migration (which is a different form of
>> the
>>
>>> live-migration) that does not require shared storage.
>>>
>>>  I think block-migration would not be correct in this instance.  There
>> is no
>> file to copy (there is no disk file in /var/lib/nova/instances/<**
>> domain>).
>> Where is it going to copy it from/to?  It's already an RBD.
>>
>> I know this is supposed to work [1].  I'm just wondering if it requires
>> disabled the "true" live migration in libvirt.  I think Josh will know.
>>
>
> Yes, it works with true live migration just fine (even with caching). You
> can use "virsh migrate" or even do it through the virt-manager gui.
> Nova is just doing a check that doesn't make sense for volume-backed
> instances with live migration there.
>
> Unfortunately I haven't had the time to look at that problem in
> nova since that message, but I suspect the same issue is still
> there.
>
> Josh
>
>  [1] 
> https://lists.launchpad.net/**openstack/msg15074.html<https://lists.launchpad.net/openstack/msg15074.html>
>>
>> On Tue, Mar 12, 2013 at 4:13 PM, tra26 <tr...@cs.drexel.edu> wrote:
>>
>>  Travis,
>>>
>>> The root disk (/var/lib/nova/instances) must be on shared storage to run
>>> the live migrate. You should be able to run block migration (which is a
>>> different form of the live-migration) that does not require shared
>>> storage.
>>>
>>> Take a look at: http://www.sebastien-han.fr/****
>>> blog/2012/07/12/openstack-**<http://www.sebastien-han.fr/**blog/2012/07/12/openstack-**>
>>> block-migration/<http://www.**sebastien-han.fr/blog/2012/07/**
>>> 12/openstack-block-migration/<http://www.sebastien-han.fr/blog/2012/07/12/openstack-block-migration/>
>>> >**for information regarding the block level migration.
>>>
>>>
>>> -Trevor
>>>
>>> On 2013-03-12 15:57, Travis Rhoden wrote:
>>>
>>>  Hey folks,
>>>>
>>>>
>>>> Im wondering if the following is possible.  I have OpenStack (Folsom)
>>>> configured to boot VMs from volume using Ceph as a backend for Cinder
>>>> and Glance.  My setup pretty much follows the Ceph guides for this
>>>> verbatim.  Ive been using this setup for a while now, and its all
>>>>
>>>> been really smooth.
>>>>
>>>> However, I if I try do a live-migration, I get this:
>>>>
>>>> RemoteError: Remote error: RemoteError Remote error:
>>>> InvalidSharedStorage_Remote vmhost3 is not on shared storage: Live
>>>> migration can not be used without shared storage.
>>>>
>>>> One thing I am doing that may not be normal is that I am trying to do
>>>> the "true" live migration in KVM/libvirt, having set this in my
>>>> nova.conf:
>>>>
>>>>
>>>> live_migration_flag=VIR_****MIGRATE_UNDEFINE_SOURCE,VIR_**
>>>> MIGRATE_PEER2PEER,VIR_MIGRATE_****LIVE
>>>>
>>>>
>>>> Anyone know if this setup should work?  Or if there is something I
>>>> should tweak to make it work?  I was thinking that having the RBD
>>>> available via librbd at both the source and destination host makes
>>>> that storage shared storage.  Perhaps not if I am trying to do live
>>>> migration?  If I do OpenStacks normal "live" migration, it will pause
>>>>
>>>> the VM and move it, which is less than ideal, but workable.
>>>>
>>>> Thanks,
>>>>
>>>>   - Travis
>>>>
>>>
>
_______________________________________________
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

Reply via email to