Abstract:
What does it mean to be practising Infrastructure as Code? Can we tick that checkbox when we use puppet and have the code checked into SCM? This talk will cover the kaizen journey of a Build Engineering team that started with that, and have slowly taken traditional development practises (such as DVCS, pairing, code reviews, BDD, continuous integration, continuous delivery) and combined them with traditional DevOps tools (such as puppet, mcollective, veewee, vagrant, librarian-puppet, puppet-lint) and practises (automation, provisioning from scratch, cattle instead of pets mentality) in managing a largish build platform (20 physical machines, 260 VMs, and 300 EC2 instances, Linux & Windows) that is critical to Atlassian's ability to deliver software to their customers. With the use of traditional development practises, come with development team related challenges, such as slow local development feedback loop, slow CI, flakey tests, ineffective code reviews, large infrequent releases causing large failures. You'll discover the tools & practises that worked for the team in managing this build platform, what challenges that were faced and how they were resolved.
Speaker: Peter Leschev