Lessons Learned From Maintaining Continuous Delivery While Migrating From a Static Infrastructure to Kubernetes [B]
summary
FINN runs about 350 microservices with roughly 1,000 deploys/week and a median 6 minutes git->prod. Migration moves workloads from static servers/OpenStack to on‑prem Kubernetes (bare metal) and GKE for dev, layered with FIAAS, Helm, a Kafka-triggered deploy operator (eos-deploy-daemon), a HAProxy ingress using Unleash feature toggles for percentage traffic shifts, Prometheus+Fluentd for metrics/logs, and a canary app. Expect weeks per legacy app, DB/timeouts and DNS surprises, and measurable rollback thresholds.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:05Evolution of infrastructure
- 2:58Moving to containers
- 8:14Kubernetes adoption
- 10:34Finn infrastructure service
- 13:57Defining infrastructure
- 16:52Deployment pipeline
- 18:27Observability
- 22:10Application migration
- 25:20Traffic and safety
- 27:42Conclusion and Q&A
related talks
39mStrangling the Legacy with Kubernetes
35mKubernetes journey
35mPerforming Infrastructure Migrations at Scale
38mBuilding the Perfect Infrastructure with Kubernetes
44mMigrating a Monolith to Kubernetes
45mLearning How to Pronounce Kubernetes to Production in 3 Months!
24m"Misadventures in Continuous Delivery"
22m10 Kubernetes is hard!