Continuous Delivery with Kubernetes at Box [I]
summary
Box runs Kubernetes across four clusters (three production) hosting about 80 services for ~80,000 organizations and builds a CI/CD pipeline using Jenkins, Docker, Artifactory and declarative jsonnet templates. Commits trigger Jenkins pipelines that inject parameters, generate per-environment manifests (dev canary → staging → prod), and a continuously running kube-applier service applies and re-syncs the repo state to clusters, overriding manual drift and enabling fast rollbacks and cluster recovery. Continuous delivery requires manual approval to push to production.
AI-written summary · about
chapters (19)
- 0:00Intro
- 0:33Boxs tooling
- 1:34Changes to services
- 2:51Incremental rollouts
- 3:50Automation
- 5:06Declarative Model
- 6:29Introducing Kevin
- 9:23App repo
- 10:53Jenkins pipeline
- 13:43Deployment mechanism
- 16:04Deployment config
- 17:54Cubaplier
- 22:22Challenge
- 24:17Benefits
- 25:23Wrapup
- 26:41What about manual changes
- 28:32Secrets
- 31:16Alternative model
- 32:05Access model
related talks
46mHow Box Runs Containers in Production with Kubernetes
47mContinuous Delivery Meets OpenShift
01h 01mContinuous Delivery with GitLab CI and Kubernetes
56mContinuous Deployment Pipeline in Kubernetes
50mKubernetes First Continuous Delivery for Developers
28mAutomate Everything
48mA Journey to Safe, Mission Critical Continuous Delivery at Descartes Labs
50mContinuous Deployment With Jenkins X And Kubernetes
