Patterns for Docker Success
summary
Shopify runs about 4,000 containers on bare‑metal across two datacenters and spends six months to get Docker into production, then operates it for a year; deploys drop from ~15 to 3 minutes and CI from ~15 to 5. Secrets use ejson encrypted secrets and a custom init to decrypt at boot; logging is still file‑based and brittle; orchestration stays with Chef rather than a new scheduler; image builds rely on a custom image build system (~1–2k LOC) and finally use overlayfs in kernel to avoid AUFS pain. Kernel lockups and logging design mistakes cause early failures.
AI-written summary · about
chapters (22)
- 0:00Introduction
- 0:57About Shopify
- 1:55Why Docker
- 3:28Infrastructure as a Service
- 5:56Shopify
- 6:51Docker adoption
- 15:52Greenfield approach
- 18:06CI benefits
- 19:22Docker adoption matrix
- 19:37Use cases
- 21:17Security
- 22:37Roadmap
- 24:09Exploration
- 27:17Adaptation
- 30:06Secrets
- 33:05Logging
- 36:51Getting ready for Docker
- 39:36Production
- 40:39Zombie processes
- 41:54Building samples
- 42:58Union file systems
- 45:22Problems with Docker
related talks
26mContainerization War Stories
01h 04mPreparing your Dockerised Application for Production
43mServer Infrastructure for Rails in 2016
01h 09mDocker in Production
40mGoes Beyond CI to Production Scale PaaS with Docker
36mDocker Inside/Out
51mProduction level Docker
24mFrom Zero to a Full Docker Based IT Teams