search

rdo-infra / weirdo

It's not weird if it works.
Apache License 2.0
9 stars 7 forks source link

Support multinode deployments #7

Open gbraad opened 8 years ago

gbraad commented 8 years ago

Currently tested deployments are based on a single node, either using the standard all-in-one or a scenario.

dmsimard commented 8 years ago

So, this is a tricky one. The scope of WeIRDO is to run gate jobs outside of the gate to improve our CI coverage.

Right now, none of the implemented projects (packstack, puppet-openstack, kolla) have multi-node jobs upstream.

I do not want to implement custom jobs in WeIRDO. If we want to test multi node, it will need to follow the right path and be done upstream first.

I know it is possible and @EmilienM looked into it a while back for puppet openstack jobs but it gets a bit hard.

We can definitely keep this issue opened for when upstream starts testing multi node.

gbraad commented 8 years ago

I agree with your viewpoint that this is something that needs to be upstream; however for both PackStack and Kolla a multi-node deployment is possible to realize. Just not so sure how this can be achieved and implemented at the OpenStack infrastructure.

dmsimard commented 8 years ago

I believe upstream multinode will be facilitated by zuul v3 -- at which point they actually intend to get rid of Jenkins as it will no longer be necessary. There's more details available about it here: https://specs.openstack.org/openstack-infra/infra-specs/specs/zuulv3.html

We're talking Ocata timeframe for Zuul v3 unless significant effort is put towards it in the meantime.