OpenStack Summit LiveBlog – An Evaluation of OpenStack Deployment Frameworks

Note: This is a liveblog. I didn’t proof read / spellcheck etc.

An Evaluation of OpenStack Deployment Frameworks

Symantec has a diverse set of development needs across product lines, etc.

Consolidation of those systems

OpenStack and other OSS building blocks

Desire to submit patches back to the project, security fixes, etc.

Tried 5 different tools, a number of criteria:

  • In Symantec DC, on Symantec hardware (Private cloud)
  • Bare metal -> OpenStack
  • Resillancy in the deploy tool
  • OpenStack validated tests
  • Documentation / Replicated

Network segregation:

  • IPMI
  • Admin
  • Private
  • API
  • Public

Three clusters of 5 nodes. Single node swift, Swift needs more by default, minimum three.


Tools evaluated:

  • Fuel Web
  • MaaS/JuJu
  • Crowbar
  • Foreman
  • Rackspace Private Cloud


  • Uses Postgres internally, MySQL for OpenStack


  • Multi-site out of the box
  • 10 nodes minimum
  • Strong debugging in JuJu


  • Local Chef server
  • Integrated Nagios / Ganglia
  • Very quick out of the box


  • Smart proxies
  • Uses Puppet for OpenStack
  • Requires customization to go from baremetal => OpenStack

Rackspace Private Cloud

  • Chef recipies
  • Not a deployment framework ( baremetal => OpenStack )
  • L3 agent was single point of failure
  • Recommended install uses Nova Networking

Testing OpenStack

Three drivers:

  • Horizon smoke tests
  • CLI tool
  • Tempest

There is a diagram / slide here with an impressive list of tests on it.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.