openstack team mailing list archive

[Sébastien Han <han.sebastien@xxxxxxxxx>]

+1 John

Since CephFS is not production ready what you can do is map RBD device
to all of your compute nodes and then mount them in
/var/lib/nova/instances. The downside of this that you have way more
IOPS since you only have one RBD per compute node for EVERY VMs that
will end up into this compute. With the boot from volume it's one RBD
per instance which brings way more IOPS to your instance. Still with
boot from volume you can also enjoy the rbd cache on the client side,
cache that will also helps with buffered IO.

Cheers!
--
Regards,
Sébastien Han.


On Thu, Jan 31, 2013 at 7:53 PM, John Griffith
<john.griffith@xxxxxxxxxxxxx> wrote:
>
>
> On Thu, Jan 31, 2013 at 11:43 AM, Sébastien Han <han.sebastien@xxxxxxxxx>
> wrote:
>>
>> Try to have a look at the boot from volume feature. Basically the disk
>> base of your instance is an RBD volume from Ceph. Something will be
>> remain in /var/lib/nova/instances but it's only the kvm xml file.
>>
>> http://ceph.com/docs/master/rbd/rbd-openstack/?highlight=openstack
>>
>> Cheers!
>> --
>> Regards,
>> Sébastien Han.
>>
>>
>> On Thu, Jan 31, 2013 at 7:40 AM, Wolfgang Hennerbichler
>> <wolfgang.hennerbichler@xxxxxxxxxxxxxxxx> wrote:
>> > Hi,
>> >
>> > I'm sorry if this has been asked before. My question is: can I integrate
>> > ceph into openstack's nova & cinder in a way, that I don't need
>> > /var/lib/nova/instances anymore? I'd like to have EVERYTHING in ceph,
>> > starting from glance images to nova-disk-images and volumes (cinder). And
>> > more important: Can this be done in a way so that my horizon-users can still
>> > use horizon (including snapshotting) without the need of abusing the
>> > command-line? :)
>> > I've read that cinder and glance are supported, but I didn't find much
>> > information on nova-disk-images and ceph.
>> >
>> > thanks for a reply
>> > wolfgang
>> > _______________________________________________
>> > Mailing list: https://launchpad.net/~openstack
>> > Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>> > Unsubscribe : https://launchpad.net/~openstack
>> > More help   : https://help.launchpad.net/ListHelp
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~openstack
>> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~openstack
>> More help   : https://help.launchpad.net/ListHelp
>
>
> Boot from volume is one good option, another is to setup ceph and point
> nova's insances_path to point to it.  I haven't done it with Ceph but I've
> done this sort of thing using iSCSI mounts on the compute node.