Release Engineering as a Service - The Art of Integrating with K8s
summary
Story of building Release Engineering as a Service that integrates with Concourse and Kubernetes using Terraform, Ruby and Go. Replaces brittle bash and huge YAML with templated pipelines, YAML anchors and object-like classes; moves cluster creation into a shared job so pipelines run in parallel (one job goes from 2 hours to 15 minutes; overall pipelines speed up 4x). Measured flakiness with 112 runs and 55% success, prioritized fixes, cuts release cadence from six months to ~10 days and shortens some bug-patch turnarounds to ~48 hours.
AI-written summary · about
chapters (18)
- 0:00What is release engineering in our context
- 1:06What is Concourse?
- 1:48Terminologies
- 3:42No Testing
- 6:58Why is duplicate yaml
- 8:14DRY YAML: Decouple
- 8:38ERB Embedded
- 9:12Cons: ERB (Embedded Ruby)
- 10:07Cons: anchors & aliases
- 10:44Attempt 3
- 11:10templates +
- 11:54Environment contention
- 14:36Pipelines Take
- 15:01Pipelines Are Slow
- 16:32Pipelines Are Flaky
- 17:50No Integration Process
- 19:02Releng as a Service!
- 20:42Open Source Licensing
related talks
31mRelease Engineering as a Service - The Art of Integrating with K8s
17mA Brief History of Release Engineering
54mFrom Code to Customer
38mSIG Release - Deep Dive: Release Engineering Subproject
20mStreamlining Release Management for DevOps Success
30mTestOps: Transforming Chaos into Streamlined Story-Based Releases
19mKeynote: Releasing a More Mature and Stable Kubernetes
28mBuilding K8s on AWS & Enabling Developers to Deliver. What to Do Better - Greg Sansom & Fahad Tariq