Continuous deployment at scale
summary
Etsy runs continuous deployment that actually ships product not just code: about 20–40 production deploys a day, driven by master-first commits guarded by feature flags (variants and percent rollouts), preflight tests via a local “try” runner, a web green-button deploy to a pre-prod called princess, and a Slack/IRC push-queue. Cleanup uses a weekly cron that finds stale flags. Monitoring (graphite/statsd, Supergraph/Supertop) shows push markers and enables low mean time to recovery; blameless postmortems produce tracked remediation tasks.
AI-written summary · about
chapters (28)
- 0:00Intro
- 0:53Buongiorno!
- 2:20principles
- 2:42enable innovation
- 3:09optimize for mastery
- 3:50quick iteration
- 4:15continuous improvement
- 6:09tooling
- 7:03Feature
- 8:53Experimentation
- 10:47continuous integration
- 11:08ready to release
- 13:41anybody can push
- 15:38dark changes
- 16:37config pushes
- 19:16push train
- 21:52dashboards
- 22:33StatsD + Graphite
- 24:08poka-yoke
- 24:35culture
- 25:08assume best intentions
- 25:19cultivate empathy
- 26:06failure is an option
- 26:19failure is an opportunity
- 27:02blameless post-mortems
- 27:19+ remediation
- 29:05mixer
- 31:2030+ config deploys








