ActiveBlog

OpenStack at ActiveState: SysAdmin Kevin Woolley on Folsom
by Diane Mueller

Diane Mueller, October 15, 2012

OpenStack Folsom OpenStack train keeps rolling with "Folsom" release

In preparation for heading down to next week's OpenStack "Grizzly" Summit in San Diego thought I'd check in with ActiveState's sysadmin, Kevin Woolley, to see how the most recent OpenStack "Folsom" release was working out for him. I figured I could gain some insights from his war stories to share with the OpenStack Community as we gather together and gear up for another release in six months.

Kevin, better known as KJ, has been actively involved in the Stackato development process and in helping to ensure that Stackato runs on "any cloud". He has a unique perspective, having to deploy and maintain multiple clouds ranging from vSphere to AWS to OpenStack.

KJ recently upgraded some of our on-premises test servers for Stackato to OpenStack Folsom. It's the sixth release of OpenStack in its two-plus year history, and it was just pushed out the door 15 days ago. I figured KJ would have some good war stories to share.

Here's how the conversation went...

Why upgrade now? So soon (15 days after the release of Folsom)?
We'd been chomping at the bit since around a month before the Folsom release. I'd been using the test bits for experimentation right up until the release, and as soon as the release hit the Ubuntu Cloud Archive I started on the final implementation.
What were your first impressions of the Folsom release?
My first impressions are overwhelmingly positive. Our last OpenStack installation was based on Diablo and there are many things that have been fixed and improved since then.
Were there any new features that you were particularly looking forward to?
The killer features I was looking forward to were ease of deployment and a better installation procedure, and both of those were delivered.
What does our OpenStack installation look like?
We're using Keystone, Glance, Nova, Swift, and Horizon.
How big is our cluster? Are there any details that are particular to us that made OpenStack tricky to install?
Our cluster is four systems: one controller that runs everything and three nodes that just run nova-compute and nova-network. The controller is 12 cores/24 threads with 64GB of RAM and several TB of disk space. One of the compute nodes has similar specs, and the other two are older machines with 8 cores/16 threads and 24GB of RAM. Nothing about these made it tricky to install OpenStack. One of the things I quite like about OpenStack is that it's quite tolerant of being installed on commodity hardware.
What was your overall impression of the installation process?
Much better than Diablo. In comparison to other platforms it ranks somewhere around the middle ground of ease of installation. The installation docs are very approachable, though, and that helps a great deal.
What's your overall impression of Folsom so far?
I love it. We had some hiccups and weirdness with our Diablo installation and it felt somewhat rickety. Whether that's Diablo or the results of a first time installing a new and complex system is unclear. Folsom feels solid and we haven't seen any of the odd behavior of the old system.
Is there anything that you'd like to see improved?

Of course! Every project has room for improvement. The installation docs have come a long way and seem to be focusing on a process-oriented approach. Seeing more of that would be a good thing.

Horizon has a great set of functionality so far and seeing more there would be great, especially if it included migrating instances between servers.

Another wish-pony would be making acting as an admin in the project dashboard of Horizon safe. Currently there's a weirdness in how Nova interacts with Keystone and lets you shoot yourself in the foot (according to the "You currently have the power to damage your OpenStack cloud..." dialog). I haven't pulled the trigger to see how much of my leg would disappear!

Is there anyone in the project that you'd like to give a shout-out too?
The whole team, really. If I were to try to single out groups or individuals I'd be giving short shrift to the folks who did the work that I'm relying on without even knowing about it. My thanks go to all of them.

Alas, no war stories to take back to the troops gathering in San Diego – just a "job well done" and kudos to everyone involved!

Subscribe to ActiveState Blogs by Email

Share this post:

Category: stackato
About the Author: RSS

Diane Mueller is a Cloud Evangelist at ActiveState. She has been designing & implementing financial applications at Fortune 500 corporations for over 20 years. Diane has been actively involved in development efforts of XBRL Open Standard (http://www.xbrl.org) since 1999.