Open Stack

On May 22, 2013, at 7:13 PM, Aaron Rosen <arosen at nicira.com> wrote:
> 
> Hi Kyle, 
> 
> Nope, Clark figured it out exactly. The gating doesn't use any setting in the ovs section so it didn't catch this. Here's a review that reverts the patch that broke it: https://review.openstack.org/#/c/30189/ 
> 
> Good find
> 
Nice, thanks for the quick turnaround Aron, +1'd your commit already.

Kyle

> Aaron
> 
> 
> On Wed, May 22, 2013 at 4:58 PM, Kyle Mestery (kmestery) <kmestery at cisco.com> wrote:
> On May 22, 2013, at 6:33 PM, Clark Boylan <clark.boylan at gmail.com> wrote:
> > On Wed, May 22, 2013 at 3:04 PM, Kyle Mestery (kmestery)
> > <kmestery at cisco.com> wrote:
> >> A recent change to Quantum to change the OVS config file to use lower-case [ovs] instead of capital [OVS] for this particular section in the config file caused me to spend a day wondering what was wrong with my setup. I've submitted a devstack change to fix this, the review is listed below for those interested. I'm unclear how this was approved into Quantum given that it pretty much broke everyone using devstack with the OVS plugin, I would have expected a matching devstack patch to make devstack work as well. Regardless, the change to devstack is relatively minor, but will thus make devstack only work with the latest version of Quantum, given this config file change. Since I don't believe devstack works with older versions of anything, I believe this to be ok.
> >>
> >> Thanks,
> >> Kyle
> >>
> >> https://review.openstack.org/#/c/30172/
> >>
> > I went digging through logs to figure out how this got past
> > gate-tempest-devstack-vm-quantum which gates all core and client
> > projects. In the gate test devstack configures quantum [1] and as part
> > of that copies the ovs plugin ini file from within the quantum tree.
> > Later on devstack configures the ovs plugin [2] but since we don't set
> > various settings like tunnels, mappings, or run this on xenserver none
> > of the `iniset OVS` calls are run in [3].
> >
> > Long story short, the gate test is not sophisticated enough to have
> > caught this as it is just using the default ovs plugin ini. I do note
> > that oslo.config 1.1.1 is being installed into the devstack
> > environment so it is possible that even if we had a more sophisticated
> > quantum setup in the gate we would have missed this for CDers like
> > Robert. I would've expected this to work with oslo.config 1.1.1 for
> > Kyle if it does actually fix the problem. Kyle, maybe that is
> > something you can check?
> >
> So, devstack is not updating oslo.config for both Fedora and Ubuntu
> for me. See here for Fedora 18:
> 
> [kmestery at km-dhcp-64-188 ~]$ sudo pip-python list|grep oslo.config
> oslo.config (1.1.0)
> [kmestery at km-dhcp-64-188 ~]$
> 
> And here for Ubuntu 12.04:
> 
> kmestery at ucs-3:~$ pip freeze | grep oslo.config
> oslo.config==1.1.0
> kmestery at ucs-3:~$
> 
> So should devstack be doing this? And even if it did, I think this
> will break the way Robert is deploying. I think this change may
> need some more thought it appears.
> 
> Thanks,
> Kyle
> 
> > [1] http://paste.openstack.org/show/37629/
> > [2] http://paste.openstack.org/show/37631/
> > [3] https://github.com/openstack-dev/devstack/blob/master/lib/quantum_plugins/openvswitch#L42
> >
> >
> > Clark
> >
> > _______________________________________________
> > 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
> 
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev