Scalable CI/CD Infrastructure with EKS and Karpenter
summary
Centralized EKS cluster in a single CI AWS account runs GitLab CI runners with Karpenter provisioner for just‑in‑time nodes, warm pools to cut startup latency, and OIDC+service accounts for cross‑account IAM role assumption; pods get scheduling directives via Kyverno mutating policies (spot vs on‑demand, AZ spread, GPU hints), and ArgoCD with ApplicationSet automates dynamic service account and app creation from self‑service requests. Infrastructure replaces per‑account runners, scales horizontally, and cuts cost by about five times while keeping secure, auditable access.
AI-written summary · about
chapters (9)
- 0:00Start
- 1:51AWS Environment
- 2:21Why: Code Management
- 3:12Why: Old Infra Management
- 4:02CI/CD Runners EKS Cluster
- 6:55Karpenter Provisioner Example
- 7:17Dynamic CI Runners Provisioning
- 11:16Approaches: App of Apps
- 12:09Approaches: Application Set
related talks
31mKubernetes workloads with Karpenter & Amazon EKS
01h 01mJust-in-time Nodes for Any AWS EKS Cluster: Karpenter
25mHow to take your Kubernetes cluster to next level with Karpenter, Ray and Amazon EKS?
40mOptimize EKS Cost and Scaling with Karpenter
22mKarpenter Efficient scaling of Kubernetes clusters
29mDevOps Practisting Platform on EKS with Karpenter autoscaling
22mUse Karpenter to scale, optimize & upgrade K8s
42mOptimize Amazon EKS workloads for performance and cost
