Create an extensible all-in-one openstack manifest

Registered by Robert Starmer

Create an extensible all-in-one openstack manifest that allows for the deployment of a single node that can then be expanded. The manifest should be based on the Cisco OpenStack Installer method of installation, and should be able to be run in a "net free" mode. It should also be able to deploy against either RedHat/Fedora or Canonical/Ubuntu distributions.

Blueprint information

Status:
Complete
Approver:
Mark T. Voelker
Priority:
Medium
Drafter:
None
Direction:
Needs approval
Assignee:
Mark T. Voelker
Definition:
Approved
Series goal:
Accepted for grizzly
Implementation:
Implemented
Milestone target:
milestone icon g.3
Started by
Mark T. Voelker
Completed by
Mark T. Voelker

Related branches

Sprints

Whiteboard

Pull request here:
https://github.com/CiscoSystems/grizzly-manifests/pull/175 [MERGED]

Note that this pull request provides control+compute functionality but does not yet provision Ceph. I'm working on that as a followup pull request once an upstream patch is merged and backported to stable/grizzly:
Master: https://review.openstack.org/#/c/47310 [MERGED]
stable/grizzly: https://review.openstack.org/47605 [MERGED]

In order for Ceph-backed Glance to work in the all-in-one install, the following upstream patch will need to be merged and cherryvpicked to stable/grizzly:
Master: https://review.openstack.org/#/c/47520 [MERGED]
stable/grizzly: https://review.openstack.org/47607 [MERGED]

We'll also need to get a couple of patches to our modules merged in order for ceph on all-in-one to work:
https://github.com/CiscoSystems/puppet-ceph/pull/9 [MERGED]
https://github.com/CiscoSystems/grizzly-manifests/pull/174 [MERGED]
https://github.com/CiscoSystems/puppet-coe/pull/18 [MERGED]

Once those are merged, a few more minor updates will be needed to core to make everything work. Pull request here:
https://github.com/CiscoSystems/grizzly-manifests/pull/180 [MERGED]

There's one remaining issue in that the above patches give you a complete standalone all-in-one deployment, but it uses defaults for a couple of rabbit parameters. This means that you'll be unable to join additional compute nodes to a cloud started with an all in one node, which is something we'll want to address. Pull request to fix that:

https://github.com/CiscoSystems/grizzly-manifests/pull/182 [MERGED]

Further, we know some users will want the force_config_drive functionality we offer in other compute nodes. Enabling this will require an upstream patch:
Master: https://review.openstack.org/#/c/47905/ [MERGED]
stable/grizzly: https://review.openstack.org/#/c/48117/ [MERGED]

Also necessary for this feature were a couple of upstream patches related to enabling provider networks in the openstack::all class:
Master: https://review.openstack.org/#/c/46306/ [MERGED]
stable/grizzly: https://review.openstack.org/#/c/46345/ [MERGED]

(?)

Work Items

This blueprint contains Public information 
Everyone can see this information.

Subscribers

No subscribers.