When SysAdmins Quit: Protecting Kubernetes Clusters When the Owner of Multiple...
summary
Shows that admin KUBECONFIG cannot be revoked directly and proves a working manual fix: create a new root and intermediate CAs, use cfssl to generate new certificates for API server, kubelet, controller-manager and admin, replace static manifests and restart kubelets so the old admin kubeconfig stops working. Demo uses kind, copies certs into /etc/kubernetes, and yields one-node downtime (multi-node avoids full outage). Notes lack of automation and need to also rotate application secrets, quotas and infra ownership.
AI-written summary · about
chapters (11)
- 0:00Cluster ownership transfer
- 1:04Why transfer is challenging
- 3:11Transferring cluster resources
- 7:03Admin kubeconfig problem
- 10:13Demo setup and cert rotation
- 13:27New certificates and configs
- 17:55Updating manifests and restarting
- 21:07Restarting kubelet and access
- 23:31Final verification and limits
- 24:13Best practices and future
- 27:08Q&A
related talks
32mStop giving root access and start securing your Kubernetes clusters instead
49mCluster Wrangling: How to make Kubernetes clusters secure and usable
32mOh No Our Kubernetes Cluster Has Been Compromised!Will YOU Save the Day?
30mManaging Kubernetes Without Losing Your Cool
53mFlying Above the Clouds: Securing Kubernetes
29mEverything Wrong with K8s Authentication and How We Worked Around It
45mBeyond cluster-admin: Getting Started with Kubernetes Users and Permissions
36mCertifik8s: All You Need to Know About Certificates in Kubernetes