Kubernetes on AWS: Practices & Opinions
summary
EKS offers a managed control plane (EKS) where masters run in AWS and worker nodes stay in customer accounts; AWS promises an upstream‑compatible API and deep AWS integrations (CNI plugin, CloudWatch, CloudTrail). Community tools include kube‑aws/cops for HA clusters, and kube2iam/kube‑iam as a kube2iam daemonset pattern to give pods IAM roles. Zalando runs about 50 clusters (one per AWS account), provisions everything with CloudFormation-driven provisioning, uses immutable node replacement, flannel, daemonsets for Fluentd and Prometheus metrics, and a cluster registry + watcher to automate updates.
AI-written summary · about
chapters (12)
- 0:00Start
- 1:57Kubernetes cluster setup
- 4:48Manage a Kubernetes cluster
- 6:43Elastic Container Service for Kubernetes
- 15:05Setup with CloudFormation
- 15:31Base setup
- 17:44AWS Identity and Access Management
- 19:51IAM role for Pods
- 22:54AWS IAM
- 31:13Monitoring setup
- 31:46Application monitoring
- 33:10Closing remarks
related talks
17mKeynote: Kubernetes: Lessons from the Trenches
17mKeynote: Kubernetes: Lessons from the Trenches
29mAmazon Elastic Container Service for Kubernetes (EKS): A case study - M. Re Ferre' and G. Carabelli
11mAmazon EKS Roadmap
35mLiving with the Pathology of the Cloud: How AWS Runs Lots of Clusters
37mkubernetes on AWS gone wild
29mThe Frugal k8s Engineer: Strategies for Amazon EKS
21mDeploying Kubernetes at scale
