Another Cluster Bites the Dust... and That’s Just Fine!
summary
Proposes ephemeral infra that provisions per PR using Crossplane XRD/compositions as the control plane and kube‑green sleepinfo patching to scale workloads to zero, then shrink node pools to one. Example builds an EKS cluster (3 t3.small nodes) in ~20–50 minutes; nightly/weekend scaling cuts compute cost about 50% (example $13.82 → ≈$7/week) and halves carbon from ~1 kg to ~0.5 kg/week. Management cluster runs Crossplane; runtime cluster runs kube‑green; reclaiming deletes the whole stack.
AI-written summary · about
chapters (17)
- 0:00Infrastructure shutdown strategies
- 2:08Resource usage patterns
- 3:33Cloud challenges and costs
- 7:56Environmental impact of idle resources
- 11:14Ephemeral environments for efficiency
- 14:05Crossplane automation framework
- 15:28Crossplane resource management
- 18:20Composite resources in Crossplane
- 20:30Self-service infrastructure claims
- 21:13KubeGreen optimization
- 22:58Ephemeral environment workflow
- 26:18Multi-layer optimization with KubeGreen
- 29:57Cost and carbon comparison
- 33:51Monthly savings projection
- 36:30Hands-on workshop
- 37:37Tracking carbon intensity
- 41:36Q&A
related talks
47mBuilding sustainable architectures with Kubernetes
15mDemo: Resource(Cost) optimization on planet scale Kubernetes clusters.
27mScaling K8s Clusters Without Losing Your Mind or Money
47mPostmortem: A Modern Approach to Managing Kubernetes Clusters
39mRunning Kubernetes on Ephemeral Infrastructure
36mOptimizing Resource Usage in Kubernetes
45mkube-green
29mThe Frugal k8s Engineer: Strategies for Amazon EKS