Operating Tekton Efficiently and Securely for multi-team use cases
summary
Runs a single Tekton installation with per-team node pools and two Tekton Chains controllers per team to write provenance to team registries; uses GKE autoscaler (10s detect + ~40s node boot) and introduces balloon pods to make task scheduling near-instant instead of ~50s; provisions an image pool job (cronjob) with two persistent volumes that flip active/inactive to cache Docker sidecar images, improving cache hit rate but not eliminating registry pulls or the need for a pipeline-level sidecar proposal.
AI-written summary · about
chapters (8)
- 0:00Introduction and goals
- 1:03Multi-team Tekton setup
- 2:50Control plane and isolation
- 4:36Resource balancing and node pools
- 6:02Compute cost optimization
- 7:51Docker caching challenges
- 10:21Summary and developer feedback
- 11:07Q&A session
related talks
16mOperating Tekton with Secure Defaults
47mTekton, a new approach to pipelines
22mLessons Learned Evolving K8s Extensions: The Future Of Tekton Pipelines
48mTekton: From Source to Production Inside Kubernetes
24mWhat a Life! - Understanding the Lifecycle of a Tekton Pipeline
21mBuild a Large Scale Cloud Native CI/CD System Based on Tekton
29mCloud Native CD Pipelines with Tekton
39mPower-Up With Clouds And Pipelines