[openstack-dev] [devstack] [neutron] OpenDaylight devstack support questions

Kyle Mestery mestery at noironetworks.com
Fri Feb 28 21:05:53 UTC 2014


Thanks Salvatore. I spent some time talking with dtroyer on IRC a few
days ago, and I have a path forward. One thing I wanted to point out is
that eventually the goal is to manage the lifecycle of ODL inside of
devstack. This is similar to what is being done with Ryu and Trema
already in devstack. But the first cut will simply allow for configuration
and de-configuration of OVS to work with ODL.

On Feb 28, 2014, at 11:46 AM, Salvatore Orlando <sorlando at nicira.com> wrote:

> Hi Kyle,
> 
> I think conceptually your approach is fine.
> I would have had concerns if you were trying to manage ODL life cycle through devstack (like installing/uninstalling it or configuring the ODL controller).
> But looking at your code it seems you're just setting up the host so that it could work with opendaylight.
> 
> I agree however that extras.d is probably not the right place, as devstack already has hooks in places for plugin configuration.
> I think they are at least:
> - configure
> - check
> - init
> - install
> - start 
> 
> big switch, midokura, nec, ryu, and nsx already use these hooks.
> I appreciate the fact that since this is a mech driver rather than a plugin, this solution won't work out of the box, but at first glance it should not be to hard to adapt it.
> 
> Salvatore
> 
> 
> 
> On 26 February 2014 22:47, Kyle Mestery <mestery at noironetworks.com> wrote:
> So, I have this review [1] which attempts to add support for OpenDaylight
> to devstack. What this currently does, in Patch 7, is that it uses the
> "extras" functionality of devstack to program the OVS on the host so that
> OpenDaylight can control it. On teardown, it does the reverse. Simple and
> straightforward. I've received feedback this isn't the correct approach here,
> and that using a plugin approach in lib/neutron_plugin/opendaylight would
> be better. I need hooks for when devstack is finished running, and when
> unstack is called. Those don't appear in the plugin interface for Neutron
> in devstack.
> 
> Another point of inconsistency I'd like to bring up is the fact that patches
> for Neutron in devstack which propose running an Open Source controller
> are being flagged with -1. However, the Ryu plugin is already doing this. I
> suspect it was grandfathered in, but it sets an inconsistent precedent here.
> I propose we either remove Ryu from devstack, or continue to let other
> Open Source SDN controllers run inside devstack. Please see Patch 6
> of the review below for the minimal work it took me to add OpenDaylight
> there.
> 
> Feedback appreciated here, I've been sitting on this devstack patch with
> minimal changes for a month. I'm also working with the Linux Foundation
> for the 3rd party testing requirements for ODL so the ML2 MechanismDriver
> can also go in.
> 
> Thanks,
> Kyle
> 
> [1] https://review.openstack.org/#/c/69774/
> 
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> 
> 
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev




More information about the OpenStack-dev mailing list