State of the Kubernetes Provider
summary
Kubernetes support moves from a VM-like v1 model to a manifest-first v2 provider with concrete results: added manifest deployment stages, Helm and Customize renderers plus a git repo artifact type, namespace override and Helm expression escapes, and first‑class blue/green rollouts by managing labels on ReplicaSets (not Deployments). Caching and startup work cuts JVM allocations roughly 133M→37M, defers permission checks to speed boot, and enables dynamic accounts so Spinnaker scales to hundreds of clusters; SIG triage and beginner-friendly issues make contributing and migration planning practical.
AI-written summary · about
chapters (17)
- 0:00Spinnaker Summit
- 0:35Overview
- 2:00A little history...
- 3:49Creating a Service
- 3:59Deploying a ReplicaSet
- 4:54And now we're here.
- 6:31Manifest Stage Improvements
- 7:39Rollout Strategies
- 8:35Bake Manifest
- 9:52Dynamic Accounts
- 11:51Caching architecture
- 14:56What if caching is inefficient? • High CPU/memory usage of clouddriver
- 16:06Improvements . -3-4x reduction in memory allocations per caching cycle
- 17:16Startup time and resiliency
- 21:26Where are we going?
- 24:12What to bring to the SIG
- 26:45Issue Triage • Explicit process of triaging/resolving Github issues
related talks
30mManaged Delivery for Kubernetes: A Collaboration Story
35mNavigating Kubernetes Cloud Provider
35mKubernetes journey
30mManaged Kubernetes Service Day Zero Survival Pack
22mDeep Dive: SIG Cloud Provider
36mHow to make Kubernetes rhyme with prod readiness - Tiffany Jernigan and Matthias Haeussler
27mKeynote: State of Kubernetes in 2022: 1.24 and Beyond
33mManaged Kubernetes Service: Day Zero Survival Pack