Open Stack

From: Jesse Pretorius <jesse.pretorius at gmail.com<mailto:jesse.pretorius at gmail.com>>
Reply-To: "OpenStack Development Mailing List (not for usage questions)" <openstack-dev at lists.openstack.org<mailto:openstack-dev at lists.openstack.org>>
Date: Tuesday, March 22, 2016 at 7:40 AM
To: "OpenStack Development Mailing List (not for usage questions)" <openstack-dev at lists.openstack.org<mailto:openstack-dev at lists.openstack.org>>
Subject: Re: [openstack-dev] [tc][fuel][kolla][osa][tripleo] proposing type:deployment

On 22 March 2016 at 09:15, Thierry Carrez <thierry at openstack.org<mailto:thierry at openstack.org>> wrote:

For OSA, we don't produce the deployment tool, only a set of playbooks. I was thinking we might need a type:packaging tag to describe which things we produce are just about packaging OpenStack things for usage by outside deployment systems (Ansible, Puppet, Chef, Deb, RPM...). So I'm not sure your type:deployment tag would apply to OSA.

Yeah, I suppose it depends on how you define 'deployment tool'. OSA is an umbrella project providing Ansible roles which deploy services, and playbooks that put them together in an integrated deployment.

Fuel similarly has libraries, Puppet roles, plugins, etc which are all packaged together to provide what we call 'Fuel'.

I expect that there are other similarities - for instance 'Keystone' may be a service, but that service has libraries and all combined together we call it a daemon service.

I guess it would be nice to have some sort of designation to allow easier filtering for consumers, assuming that this actually does add value to Operators/Packagers who consume these projects.

Jessie,

The only requirement is:

  *
The repository contains software that deploys at minimum
  deliverables tagged with starter-kit:compute in the
  projects.yaml file.

I guess we could add more if needed, but I'm a big fan of less is more, so I'd  be open to adding requirements if the above is unclear that the tool (puppet/chef/osa/kolla/fuel/triploe0) needs to deploy OpenStack and it needs to be functional afterwards.

I think the "functional afterwards" is unstated and probably needs an update to the patch to differentiate between packaging efforts and deployment efforts.

I also think the project should deploy the dependencies required to operate start-kit:compute which include a database of their choosing and a message queue service supported by oslo.

Note compute-kit is not onerous - there are only a few projects which have the starter-kit:compute tag.  They include keystone, glance, neutron, and nova.  Clearly that could change in the future, but at present, it wouldn't be a burden on any deployment project to just simply apply the tag and move on.

Thanks for jogging my thought processes - I'll update the review this morning.

Regards
-steve

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20160322/e8e6a08a/attachment.html>