Taking the Helm - Ancestry's Journey to Kubernetes
summary
Ancestry migrates from thousands of Windows VMs to Docker on CoreOS and Kubernetes, cutting deploys from 20–50 minutes to seconds/minutes; production runs live for ~1 year with hundreds of namespaces and pods across dev, stage, prod, PCI and support clusters. Uses Quay Enterprise for registry, Prometheus for cluster metrics, a custom log forwarder (based on logspout) piping logs to Kafka, a QuickStart called Barbossa, one-namespace-per-service and one-container-per-pod conventions, hourly cluster backups and resource quotas.
AI-written summary · about
chapters (14)
- 0:00Intro
- 2:03Data drives our business
- 5:07Deployment Processes
- 6:32Change is Hard
- 7:58Adopting New Technologies
- 10:25How It Started
- 13:07Early Discoveries
- 15:28Linux Built for Containers
- 17:36Kubernetes to the Rescue
- 19:59Conventions/Standards
- 23:54Production Environment
- 30:56Quick Start Tools
- 35:30Our Journey So Far
- 37:05The Power of Kubernetes
related talks
46mAround & After Kubernetes: The Principles and Ideas that Guide Us
40mFrom dev to prod. One Year of Using Kubernetes in Production
26mHow Ancestry Got Kubernetes to Run 2x Better Per Dollar Using AI
35mKubernetes journey
36mCyscale's Journey on Kubernetes
35mNavigating the Kubernetes Odyssey: Lessons from Early Adoption and Sustained...
22mCreating a Micro Open-Source Community with Helm
53mManaging Containers at Scale with CoreOS and Kubernetes