How NOT to Start with Kubernetes
Talkat KubeCon + CloudNativeCon North America 2021 · on Wednesday, 2021-10-13· at 11:00· Petree Hall C + Online·
summary
Argument: don’t default to Kubernetes; pick it only when the tradeoffs match. Details: a manual kubeadm POC leads to upgrade broke cluster, Rancher can lock in a CNI so Flannel no network policies, wrong service/Pod CIDRs cause CIDR overlap blocked access, and using image tags like don’t use :latest causes nondeterministic failures. Practical fixes: automate cluster provisioning, use private registries and base images, enforce RBAC and policies (Gatekeeper/OPA), plan ingress and certs, choose reliable storage/CSI, require health checks, resource requests, and local testing tools.
AI-written summary · about
chapters (20)
- 0:00Intro
- 0:442018 - How everything started...
- 2:46Do not install K8s from scratch
- 4:31Automation everywhere
- 5:34Network CNI and Configuration
- 8:28How we could reach our service?
- 10:44We need a way to store some files...
- 13:01RBAC
- 14:00Logging and Monitoring
- 15:40All ingresses are down!?
- 18:12Wrong motivation
- 19:19Local Development
- 20:35latest - why isn't it working anymore? E
- 22:08Private Registry & Base images
- 23:54Not using the power of Kubernetes
- 24:26Health checks
- 25:18Resource Limits and Requests
- 26:44Env Variables for Config and Secrets ?!
- 27:56helm
- 29:46Summary
related talks
32mLaying the Foundation: Real World Kubernetes Deployment Patterns
36mKubernetes Confessions: Real Mistakes We’ve All Made (and What They Teach Us)
52mKubernetes isn't the answer
43mKubernetes: Use it, Contribute to it, and Enjoy it!
17mAnti-patterns for Kubernetes Operators
36mHow to make Kubernetes rhyme with prod readiness - Tiffany Jernigan and Matthias Haeussler
41mKubernetes the Very Hard Way
29mHow to make Kubernetes rhyme with prod readiness