Cluster as Code
summary
A pipeline builds and verifies three images built in ~20 minutes by converting Dockerfile-like definitions into bootable OS images, exercising the kernel and running KVM-based integration tests (Simian) that spin up Mesos + Marathon plus Zookeeper. Clusters live as declarative class definitions in source control and change via pull requests and CI, which orchestrates Ironic, Calico networking, Vault/secrets and image rollout so ephemeral Mesos/Marathon nodes replace mutable hosts.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 4:43HighLevel Approach
- 5:45Curation Tools
- 7:30Chemic
- 9:12Master Image
- 10:35Isolation Source Control
- 12:12Setup Agent
- 12:37Operating System
- 14:47Verification
- 16:16Isolation
- 17:26Pain Points
- 23:10Class Definition
- 24:52Changing Master Image
- 25:17Changing Enemy Image
related talks
18mCluster Management: Heterogeneous, Lightweight, Safe. Pick Three
23mOrchestration and CI/CD
09mBehind the Curtain: Multi-Cluster Backstage Deployments with GitO...
28mAutomate Everything
06mBuilding Continuous Delivery Pipelines
58mOptions Galore to Get from Source Code to Container
37mShaping Applications for Docker, CoreOS, Kubernetes & Co.
23mBuilding a Secure By-Design Pipeline with an Open-Source Stack