Open Stack

The second iteration of the nova conductor fleet change to devstack is 
approved:

https://review.openstack.org/#/c/477556/

The first attempt blew up a few jobs because of things like quotas and 
notifications which are now fixed on master, either in the devstack 
change itself (for notifications) or in nova (for quotas).

I went through the various non-voting job failures in the change this 
afternoon and triaged them all. The only one that looked remotely 
related was the dvr-ha multinode (3-node) job which failed to map the 
subnode-3 node to the cell, but I think that's more due to a latent race 
that's been around in our devstack-gate/devstack setup since Ocata, and 
is maybe made worse in a 3-node job.

I know it's not fun doing this so close to feature freeze but we need as 
much time as possible to burn this in before the pike-rc1 phase.

If you see anything blow up as a result, please reply to this thread or 
yell at us (mriedem/dansmith/melwitt) in the openstack-nova channel.

-- 

Thanks,

Matt