Opscamp San Francisco

We ran our third Opscamp this weekend in San Francisco.  Like the Austin Opscamp we ran back in January, this one was great.   We had over 150 people registered and wound up with about 100 actual attendees.  GWOS provided the facility and did a great job with breakfast and lunch.  I personally felt that our original Opscamp Austin was a rock show; however, I must admit the Bay-Area folk raised the “bar”.  In general this Opscamp focused more on core ops discussions.  We also followed the strict Opscamp rule of “No What is a Cloud” discussions allowed.

Some of the more interesting discussions on topics were as follows:

  • An Operations Repository
  • What is Devops?
  • Behavior Driven Operations
  • Cloud Application Impact on Operations

We started the day with sponsored lightning talks from the following vendors:

  • James Watters of VMware
  • Michael Johnson of rPath
  • John Willis (Me) of OpsCode
  • Mark Hinkle of Zenoss
  • Tara Spalding of Groundwork Open Source
  • Mike Malloy of New Relic

Mark Hinkle and Mike Malloy pulled a fast one by presenting without any slides.  That’s the essence of cool at an un-conference.  Not to be undone by my good friend @mrhinkle I decided to work without a wire and use only one of my slides (The obligatory “Must Use In All Presentations” slide as deemed by Michael Cote of Redmonk.)  

The un-panel was, as usual, great.  Five brave souls tested their expertise against an expert audience.   The following individuals volunteered for the un-panel:

  • Damon Edwards, DTO Solutions
  • Wesley Beary , Creator of Fog
  • Michael Johnson of rPath
  • Robert Bera of
  • Lee Thompson of DTO Solutions

After the un-panel we broke out into the un-sessions.  For those of you not familiar with the Cloudcamp/Opscamp style, the un-sessions begin as nothing but open time.  Participants at the conference propose sessions and they are added to a post it board and used as the conference schedule. 

I was able to attend four sessions.  Here is a brief overview of the sessions I attended.

Operations Repository

This session addresses repository state issues and how to use a repository to build out and describe infrastructures.  Different solutions were suggested ranging from Cfengine, Puppet, or Chef abstraction based repositories to artifact based solutions like rPath.  Other possible solutions included simple GIT->TAG->BUILD->PUSH solutions.  Also, filesystem based repositories were presented such as using AFS snapshots and possible repository in place solutions. 

What is Devops?

Will Devops be a “C” level discussion next year?  We discussed the “Devops” movement and how it should look in an ideal state.  Also the discussion of ITIL and service management crept into the discussion.  One of my favorite take aways from the session was the question of whether enterprises will be asking questions like what is your Devops budget or strategy for 2011?

Behavior Driven Operations

BDD for operations!  The session started with a great overview of the BDD space and history from a Ruby perspective.  We talked about using Cucumber and Cucumber-Nagios for operations.  Different perspectives on how to accomplish BDO were discussed.  One approach suggested was to let test failures drive builds.  Kind of flipping BDD on it’s back. 

Cloud Application Impact on Operations

The focus of this session was to look at how applications should be designed in the cloud.  A number of de-coupled architecture structures were discussed.   No ’cloudy” conference is complete without at least one heated “private” cloud debate.  The general consensus of this group was that “public” clouds will inevitably become ubiquitous. 

All and all this was another great Opscamp.  In the close-up session we talked about how to keep the conversation going.  Dave Neilsen has set up a LinkedIn group called Opscamp and we have created two Google groups (Opscamp and Opscampboard).  Please stay tuned for upcoming Opscamps in your town at

John Willis