Open Stack

Hi Marcus,

The silence you hear is because Ceph isn't an OpenStack project,
though a lot of us (myself included) do use it heavily with OpenStack.

Your questions are very cpeh specific rather than  about how to get it
to work with OpenStack and would be better answered on the ceph-users
mailing list by people with a deeper understanding of the internals:

http://ceph.com/resources/mailing-list-irc/

I may be out of date on this but my understanding (& certainly the
version I'm running) of Ceph is strongly consistent so all replicas
are written before returning success.  Also objects are served from a
single master regardless of the number of replicas, though if the
master goes away another copy takes it's place.  Both of these have
negative consequences for geo replication, but this is also (I think)
and area of active interest in ceph not sure how near or far tehy are
from adressing it though.  Swift is much better today (IMO) at
georeplication, but Ceph has RBD ontop of the object store so you can
use it for Cinder block storage, and Nova ephemeral volume backend
that uses CoW clones of RBD images in the Glance store which speeds
startup and saves space.

I don't have a wide geographic dispersion so Swift's much more
compelling story there doesn't mean much to me, but I do want a single
resilient, elastic storage cluster behind glance, nova, and cinder
which Ceph can get me.

Swift is an OpenStack project so if you have technical question about
what it can do or how it works here is a good place to ask.

I'd strongly suggest you take the Ceph questions to the Ceph list to
be sure I'm not giving you bad or out dated info on what it can do
today.

-Jon

On Sat, Oct 4, 2014 at 12:45 PM, Marcus White
<roastedseaweed.k at gmail.com> wrote:
> Hello,
> A bump!
>
> MW
>
> On Tue, Sep 30, 2014 at 2:17 PM, Marcus White
> <roastedseaweed.k at gmail.com> wrote:
>> Hello,
>> Had some basic Ceph questions.
>>
>> a. Ceph is strongly consistent and different from usual object, does
>> that mean all metadata also, container and account etc is all
>> consistent and everything is updated in the path of the client
>> operation itself, even for a single site? Is it the same for block and
>> file?
>>
>> b. If it is strongly consistent, is that the case across sites also?
>> How can it be performant across geo sites if that is the case? If its
>> choosing consistency over partitioning and availability...For block, I
>> read somewhere that it is now eventually consistent(AP vs CP) via DR.
>> Gets a bit confusing with all the literature out there. If it is DR,
>> isnt that slightly different from the Swift case?
>>
>> c. For block, is it CP on a single site and then usual DR to another
>> site using snapshotting?
>>
>> d. For block, is it just a linux block device or is it SCSI?
>>
>>
>> TIA,
>> MW
>
> _______________________________________________
> Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
> Post to     : openstack at lists.openstack.org
> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack