The state of on-prem Kubernetes

on YouTube ↗49m

" summary

Built on eight years of on‑prem work and three months of backporting, the claim is that Kubernetes itself is easier than integrating it into a 15,000‑line Perl provisioning system. Tools compared: kubeadm over SSH (kubespray/ansible) for simple installs, cluster API requires another Kubernetes to bootstrap so your laptop or an admin cluster becomes critical, EKS‑Anywhere + Tinkerbell, Rancher/k3s, OpenShift, Talos/Omni. Trade‑offs: API vs SSH provisioning, lack of SSH on Talos, operational buy‑in, and utilization arguments like “only 40% CPU used” versus real packing gains.

AI-written summary · about

" chapters (8)

1. 0:00My first Kubernetes journey
2. 5:21Defining pets vs. cattle
3. 6:47Why run on-premises?
4. 7:46Addressing environment constraints
5. 26:01The Cluster API debate
6. 26:59Tooling and operational reality
7. 36:08Q&A session
8. 41:08Final thoughts on backups

" related talks

• 36mThe Road to More Usable KubernetesJoe Beda · KubeCon + CloudNativeCon North America 2017
• 45mKubernetes Operability ToolingBridget Kromhout, Brendan Burns · DevOpsDays Seattle 2019
• 43mKubernetes Operability ToolingBridget Kromhout · GOTO Chicago 2019
• 35mNavigating the Kubernetes Odyssey: Lessons from Early Adoption and Sustained...Raul Benencia · SREcon24 Americas
• 42mKubernetes on-premisesManuel Carlo Ranieri, Claudio Pozzoli · Voxxed Days Ticino 2017
• 21mDeploying Kubernetes at scaleEric Wright · AWS re:Invent 2019
• 35mKubernetes journeySébastien Lecocq, Geoffrey Thery · DevOps D-DAY 2019
• 24mWe Broke Free From Cloud-Managed Kubernetes...Kārlis Akots Gribulis, Per Hedegaard Christiansen · Platform Engineering Day Europe 2026