10+ Deploys Pe
summary
Flickr runs >10 deploys per day by automating infrastructure, keeping code and configs in a single SVN repo, and using a one-step build plus one-step deploy workflow (Hudson for CI, optional Capistrano/RPMs). Hundreds of feature flags enable dark launches and gradual traffic buckets; Ganglia and deploy timestamps on dashboards tie deploys to metrics so rollbacks are rare and teams roll forward. Shared runbooks, IRC event feeds, search-indexed logs, read-only access for devs, and regular fire drills reduce risk and speed recovery.
AI-written summary · about
chapters (15)
- 0:00Introduction to Flickr
- 1:25Dev and Ops relationship
- 4:51Lowering risk of change
- 7:43Automated infrastructure
- 9:48Version control and building
- 12:37Deployment strategies
- 16:02Feature flags
- 21:23Shared metrics and monitoring
- 25:00Adaptive feedback loops
- 27:56Communication and context
- 29:41Building a culture of respect
- 32:41Trust and collaboration
- 36:16Sharing tools and access
- 39:25Handling failure constructively
- 42:10Conclusion and summary
related talks
24mContinuous Deployment at Scale
01h 01mI deploy on Fridays (and maybe you should too)
44mI Deploy On Fridays
36mContinuous Deployment with Rails
47mDeploying Like a Boss: From 2 Servers to 20,000
43mTo make mistakes is human. To auto-deploy it is DevOps.
26mThe Many Ways to Deploy Continuously
32mContinuous deployment at scale