Migrating Legacy Infrastructure to Kubernetes: Olark’s Lessons From the Trenches
summary
Company migrates ~200 Puppet‑managed Ubuntu VMs to Google Container Engine, keeping problematic Redis with Sentinel on bare metal because containerizing it proves too expensive; billing service moves first as a low‑risk pilot, chat frontend represents ~40% of traffic and forces sticky XMPP connections so service‑load‑balancer hacks and long pod termination grace periods handle persistence. Uses Linkerd (a Linkerd daemonset per node) for consistent metrics and gradual cutover, measures baselines before changes, and relies on service endpoints that point to legacy hosts during the staged migration.
AI-written summary · about
chapters (6)
- 0:00Start
- 0:34Why We Decided To Embark on this Kubernetes Mission
- 7:56Dev Environment in Kubernetes
- 14:42The Most Risky Service
- 25:53Pick Your Battles Wisely
- 27:17Service Load Balancer
related talks
35mNavigating the Kubernetes Odyssey: Lessons from Early Adoption and Sustained...
35mThe Oregon Trail to Kubernetes [I]
46mMigrating from VMs to K8s
35mPerforming Infrastructure Migrations at Scale
38mThe Good, the Bad & the Ugly of Migrating Hundreds of Legacy Applications to Kubernetes
36mFrom Mesos to Kubernetes
18mRe-Architecting Monoliths Into Kubernetes Microservices at Million-User Scale: Hard...
01h 02mLearnings from the Enterprise-level Kubernetes transformation at Elkjøp Nordic