[openstack-dev] Quantum + ZeroMQ (with a sprinkling of Nova) in Folsom

Eric Windisch eric at cloudscaling.com
Tue Sep 25 20:55:27 UTC 2012


(I don't think reverting the format will be a good idea - but this change may become in part redundant with the ServiceGroupAPI - so TBD. I think this message format is better and will be better in the long run, even if we need a version flag to identify it) 

Regards,
Eric Windisch


On Tuesday, September 25, 2012 at 16:53 PM, Eric Windisch wrote:

> Note that before this patch, it wasn't at all compatible with Quantum. 
> 
> Even leaving Nova as it is, it could simply be "unsupported" in Quantum.
> 
> For Grizzly, we can either opt to revert this change to the message format, or
> support versioning.
> 
> Regards,
> Eric Windisch
> 
> 
> On Tuesday, September 25, 2012 at 16:45 PM, Mark McLoughlin wrote:
> 
> > On Tue, 2012-09-25 at 13:28 -0400, Eric Windisch wrote:
> > > Russell, perhaps. The issue is that if the message format changes, it
> > > would be lower cost to do it now. Also, I'm moving this to the public
> > > list, because I agree that it is a good time to do it. 
> > > 
> > > IF there is an RC3 for Nova, I'd like for consideration to get the
> > > message-format change in, because it reduces complexity for Grizzly
> > > (having to understand two formats), and because it will make the
> > > Quantum code supportable. This binary needs to move outside of Nova to
> > > avoid this unnecessary interdependency.
> > > 
> > > It might be resolvable by adding a version to the message format, but
> > > it doesn't solve the problem today - which, arguably, doesn't have to
> > > be solved. We can just release with the zeromq stuff broken in Quantum
> > > for Folsom, without any ability to backport, and I'm prepared for that
> > > to happen. I've just rallied for a best-effort to get it in, because
> > > it would be a nice to have, after finding out at F3 that it wouldn't
> > > work.
> > > 
> > > As it is, the patch that is already in Quantum should be enough to get
> > > us going for Folsom if we can also get the patch into Nova. The two
> > > patches pending review can be backported, if necessary. 
> > 
> > 
> > 
> > 
> > 
> > It took me a while to make any sense of this
> > 
> > This change which was merged into openstack-common a few days ago:
> > 
> > https://review.openstack.org/11408
> > 
> > is what changed the message format?
> > 
> > I must admit I didn't realize there was a backwards compatibility issue
> > with that patch.
> > 
> > So, the message format used by zmq in Nova is incompatible with what's
> > used in Quantum.
> > 
> > At this point I'd be inclined to document the zmq driver as unstable and
> > warn users that it's only included in Folsom as a "preview". Making
> > incompatible changes days before the release should tell us that we're
> > not ready to commit to maintaining compatibility with this driver yet.
> > 
> > If we don't document it as unstable, we have two choices before the
> > final Folsom release - either merge the new format into Nova or revert
> > it from Quantum.
> > 
> > Cheers,
> > Mark.
> 






More information about the OpenStack-dev mailing list