[openstack-dev] [nova] cells v2 next steps

Matt Riedemann mriedem at linux.vnet.ibm.com
Thu Sep 1 01:33:11 UTC 2016


On 8/31/2016 1:44 PM, Matt Riedemann wrote:
> Just to recap a call with Laski, Sean and Dan, the goal for the next 24
> hours with cells v2 is to get this nova change landed:
>
> https://review.openstack.org/#/c/356138/
>
> That depends on a set of grenade changes:
>
> https://review.openstack.org/#/q/topic:setup_cell0_before_migrations
>
> There are similar devstack changes to those:
>
> https://review.openstack.org/#/q/topic:cell0_db
>
> cell0 is optional in newton, so we don't want to add a required change
> in grenade that forces an upgrade to newton to require cell0.
>
> And since cell0 is optional in newton, we don't want devstack in newton
> running with cell0 in all jobs.
>
> So the plan is for Dan (or someone) to add a flag to devstack, mirrored
> in grenade, that will be used to conditionally create the cell0 database
> and run the simple_cell_setup command.
>
> Then I'm going to set that flag in devstack-gate and from select jobs in
> project-config, so one of the grenade jobs (either single node or
> multi-node grenade), and then the placement-api job which is non-voting
> in the nova check queue and is our new dumping ground for running
> optional things, like the placement service and cell0.
>

FYI, this is the change I'm using to test the grenade/devstack series:

https://review.openstack.org/#/c/363971/

That's similar to what's proposed in the job updates in project-config:

https://review.openstack.org/#/c/363937/

We have a dependency chain going on now where the top devstack change 
depends on a nova change that depends on the top grenade change, so it's 
all kind of self-testing.

-- 

Thanks,

Matt Riedemann




More information about the OpenStack-dev mailing list