How We Progressively Deliver Changes To Kubernetes Using Canary Deployments and Feature Flags
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 16:00· Level 0 | ICC Capital Hall | Room J·
summary
1,600 customer instances run across eight Kubernetes reefs and updates move through staff → insiders → early adopters → stable, with a rotating 5% canary cohort updated within 1–3 days. Rollouts can be halted instantly via an Andon cord and new behavior is gated with OpenFeature flags so features ship merged into main but stay off in production. Result: lead time to change falls to ~4 days, SLOs near 99.9%, and high-severity defects drop below 1%.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 3:07Company architecture
- 5:52Maintenance window challenges
- 9:06Canary deployments and Andon cord
- 13:17Initial canary limitations
- 14:43Faster upgrades and feature flags
- 16:50Trunk-based development
- 21:40Combined approach advantages
- 23:03Detecting and resolving issues
- 25:55Feature toggle management
- 26:37Feature toggle strategies
- 27:38Key takeaways
related talks
20mHow We Progressively Deliver Changes To Kubernetes Using Canary Deployments and Featur...
32mMaster your software delivery with feature flags
31mFlag-tastic and Fearless: Feature Flags Meet K8s for Deployments That Sing!
59mFaster & less risky releases with feature flags
20mProgressive Delivery with the Kubernetes Gateway API
35mLeveling Up Your CD: Unlocking Progressive Delivery on Kubernetes
32mReducing Deployment Risks with Progressive Automated Canary and Spinnaker in a Reg...
38mScaling Kubernetes Deployments Across +15'000 Namespaces