Migrating a live site across the country without downtime
summary
Claims a live datacenter migration to EC2 with zero downtime by rebuilding the platform in parallel and running rigorous process-level tests. Uses Chef to reproduce old hosts, a site-to-site VPN plus VPC routing and HAProxy to forward traffic during DNS switchover, and Continuent Tungsten to coordinate MySQL replication and hold queries during failover. Solves Redis/Resque by writing to dual Resque queues and draining the old queue, lowers DNS TTLs, and rehearses the six-step cutover until deterministic.
AI-written summary · about
chapters (13)
- 0:00Migration challenge
- 1:58Zero downtime for billing
- 4:28EC2 versus data centers
- 6:15Automating with Chef
- 9:03Testing end-to-end reliability
- 14:00Share-nothing architecture
- 15:45VPN and VPC setup
- 18:32Top three pain points
- 20:42Redis failover challenges
- 23:13MySQL replication
- 28:31Query processing during failover
- 30:35Migration plan and rollback
- 33:26Benefits and future plans
related talks
34mZero Downtime Database Migrations and Deployments
24mMigrations under Production Load: How to Switch...
19mZero-Downtime by Design: High-Availability Migration Patterns
36mHow we finally migrated an eCommerce-Platform to GCP
22mEffective migration from Newton to Queens with minimal downtime
32mSchroedingers Backup
24mZero Downtime Mastering Cassandra Live Migrations at Scale
40mCloud Databases: Pros and Cons