Sun, 25 August 2013
Roundup ‘13 - Devops and Culture ChangeFully formatted shownotes can always be found at http://javaposse.com Extra thanks to D.J. Hagberg for the excellent and complete show notes. - Netflix "Freedom & Responsibility" - Every team can organize as they choose - "Firemen" who get paged when system goes down. Sets best practices - "Site Reliability Team" - sets best practices: deployment, monitoring, ... - "Cloud Performance Team" - sets best practices for cloud performance - Groovy - Python - Standarization: set up practices/tools such that choosing the "Standard" makes life easier - Regulatory constraints (Sarbanes-Oxley, PCI DSS) - Communication with Devops: Build, Deploy, Release - Documentation: in many cases, auditors are happy with scripts - tl;dr - Adrian Cockcroft - InfoQ, YouTube, @adrianco - Opportunities for Ops to deploy systems, devs to assist & learn from Ops - Uptime and Availabitility - metrics - Wake up the developer at 3am when they break things. Devs need to understand Operational Risk of changes - PCI Compliance mitigation ideas . Separate out the SOX/PCI regulated parts of application, pared down to most minimum possible . Put those parts under separate accounts, machines, audit trails, etc. - Metrics: what do you use to measure? Customer Happiness, Money, Loyalty? (Netflix: "movie start metric") - Twitter - measure sentiment (for consumer-facing/high-profile services) - Automation: bite-size, recoverable pieces - Eiffel design-by-contract concept: preconditions/postconditions - Redefining the job of operations - Executive sponsorship - QA does the deployment through DevOps-provided tools - Post-Mortem, military AAR, Incident Review, Agile Retrospective: have thresholds for Big Meetings vs small outages/rollbacks - Metric: time to get new feature to the customer - Self-directed improvement: Deming/Lean improvement cycles - Kaizen - Visible Ops Handbook - Gene Kim - The Phoenix Project (book) - Gene Kim - business Case for improvements? What was cost of the last outage? - Chase Manhattan - 3 day outage in 2010 - Every business in the future will be a software business - Move developers into Ops teams - Google engineers get extra "hazard" pay/leather jacks - DevOps Culture Hacks - Practices brought back from Ops to Dev . JMX: Measure/Instrument *all* the things (esp. consumer facing) . Vagrant/Opscode Chef/Puppet stack into Dev environment - exercise deployment on every Dev change. - "Foreign Exchange" (hostage exchange?) program - "Don't make customers modify XML Files" - Static Config to Files. Stays the same for release. - Dynamic Config - looked up from service - Archaius - Netflix - Feature Switches/Feature Flags - Etsy . Enable/disable features & settings at runtime . Hierarchical/tagged/scoped configurations . Use config changes for rollback Thanks
The Java Posse consists of Tor Norbye, Carl Quinn, Chet Haase and Dick Wall |
To browse other episodes, use the Archives list on the right >>