OpenStack Reference Architecture Support by OpenCrowbar

[Soheil Eizadi]

I wanted to discuss the reference implementation documented here:
http://docs.openstack.org/openstack-ops/content/example_architecture.html

Although it is a little dated in that the OS and Dell Server hardware have moved on, the specialized roles e.g. Controller, Compute and Storage are still valid. There is still the need for several networks to provide isolation.

I think it is good reference implementation for us to discuss how OpenCrowbar would fit in.

From what I know OpenCrowbar is today, it can discover the server configurations and build a database that can be queried. You would need a higher level orchestrator that understands how to configure this type of cluster. This orchestrator could be another Chef cookbook/recipe that knows how to implement this using nodes defined by Table 1.1 building blocks. This would drive the specific assignment of discovered nodes to specific OpenCrowbar constructs e.g. deployment and roles. Opencrowbar would perform the specific  functions of bonding NICs and setting up the networking, setting up mirroring for the OS drives and setting up the Storage Node RAID configuration.

Has something like this already done for OpenCrowbar?

-Soheil

[Rob Hirschfeld]

Soheil,

You are correct - it would be possible to do that.  That RA matches Crowbar target deployments and the skeleton for it was started in https://github.com/opencrowbar/openstack but not updated since Dell stopped contributing.  That work was done to start porting the original Crowbar cookbooks (now maintained by SUSE) towards the community cookbooks.

There are several approaches to this based on which tools you are most comfortable using.  It could be done with Chef, Ansible, Saltstack, Puppet or even Juju.  For any (or all) it's worth considering how to leverage the OpenStack community around those tools.  OpenCrowbar was designed specifically to have a clean hand-off (Ready State) to these efforts.

This would be a great discussion for the weekly community call (Wed 11 central).

Rob


PS: We can talk 1x1 if you'd like too.

--
You received this message because you are subscribed to the Google Groups "Crowbar" group.
To unsubscribe from this group and stop receiving emails from it, send an email to opencrowbar...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Soheil Eizadi]

Hi Rob,

I am familiar with the crowbar/openstack project, we have been working with the community to update the barclamp-neutron to add support for our external DDI solution: https://github.com/crowbar/barclamp-neutron/pull/186

I think MAAS/JuJu is probably the closest to the overall solution I am looking for, it could also be a good starting point to build this capability. From what I know using Chef_provisioning and OpenCrowbar would be a better starting point for what I am trying to do. I am still trying to validate this, next week I will try to build out a Dell hardware cluster and become more familiar with OpenCrowbar. Assuming I move forward with OpenCrowbar, I will writeup a description and review it with the team on the weekly conference call.

Thanks for the feedback,

-Soheil

[Rob Hirschfeld]

Soheil,

Interesting.  What is your overall objective?

Rob

Rob
____________________________
Rob Hirschfeld, 512-773-7522
RackN CEO/Founder (r...@rackn.com)

I am in CENTRAL (-6) time
http://robhirschfeld.com
twitter: @zehicle, github: cloudedge & ravolt

[Soheil Eizadi]

My  objective is to develop an automated environment to build out a cluster like the one I sent out earlier for OpenStack. Ideally I like to automate it closer to when the server is racked and wired, why I am looking at Opencrowbar, versus a generic PXE server.

-Soheil

[Rob Hirschfeld]

Soheil,

That makes a lot of sense.  There are a few ways to build OpenStack from there.  I know of efforts on Chef, Puppet, Saltstack and Ansible.  I'd love to see some effort to improve hand off the Ready State into one or more of those tools so that we could automate from start to finish.  

That's very doable yet something that will need some directed focus.

Rob

[Soheil Eizadi]

I think we are on the same page, there is no shortage of OpenStack orchestrators, I would add Fuel to the list also :)

I would not try to duplicate what they are doing, I see this layer above the basic OpenStack provisioning in understanding the underlying deployment and the range of server node hardware types and networking. That is the gap that I am trying to fill.

-Soheil