Openstack - getting it all up magically - and when the magic fails.
Submitted by Anshu Prateek (@anshprat) on Tuesday, 12 January 2016
To discuss the experience of openstack setup and testing using CI/CD based on puppet, what all we did and when and where does it fail. At the end of the session, the audience will know about what all works and helps to setup an openstack cluster with no manual intervention. And what all can go wrong in the same openstack production cluster.
This talk will discuss my experience with Openstack, Puppet and setting up the same in a completely automated, no-manual intervention method at Reliance Jio. We used the same openstack setup to create further openstack setups! These mini openstack setups on top of the base openstack setup are used for testing and gating new code changes and further running Acceptance tests. And at the heart of it is the same code which runs itself, runs production, runs and tests itself! At the heart of this setup is a puppet module which is like nothing you would ve seen. This puppet setup is a standalone setup - i.e, no master-client setup. This puppet module is unique in the sense that it manages/uses other puppet modules as well uses itself as a module of itself! Tying all the components of openstack together for service discovery and avoiding dependency failures till a required service is up (by means of a DNS blocker) is consul.
Basic knowledge of openstack and puppet.
I have been working on large scale environments for the last seven years. Starting with Yahoo! search, where I spent the first four years - and from where I decided to move one fine day when I realised I have “touched” each and every machine in the search inventory! From there on, the next stopover was at Aerospike, which brought altogether different numbers (A million TPS in less than a millisecond!). Am presently working with Reliance Jio as Platforms Engineering Manager / Devops on efforts of setting up a public cloud.