Shopify in Multiple Datacenters
summary
Shopify claims it now runs shops concurrently out of multiple datacenters and demonstrates how that works: hundreds of thousands of shops, tens of thousands of requests per second, database sharding plus an isolation pattern called pods, routing via a Lua/Nginx module dubbed sorting hat, and a scripted one-minute failover that has been exercised in production. The team inventories cross-shard routes, enforces them with a “shit list” deprecation gate, moves shop ownership between pods, and reports the approach shrinking blast radius and enabling multi-DC deployments.
AI-written summary · about
chapters (12)
- 0:00Shopify’s infrastructure journey
- 3:30Flash sale scale challenges
- 7:14Single vs multi-tenant platforms
- 9:05Vertical scaling and sharding 2013
- 11:13Resiliency and failure points 2014
- 13:42Moving to multiple data centers 2015
- 16:09Pod architecture introduced 2016
- 22:14Multi-data center rules
- 24:04Refactoring with shit list
- 27:48Resiliency and scalability in pods
- 28:31Future goals: zero downtime failover
- 29:32Q&A: transactions, sharding, pods
related talks
40mHow Shopify Scales Rails
23mStanding On the Shoulders of Giants...
17mHow We Deploy Shopify
33mHigh availability by offloading work
04mKeynote: One Deploy Plane, Every Runtime: How Shopify Generalized Cloud Native Beyon...
29mBlack Friday - Lessons in Resiliency and Incident Response at Shopify
23mHow Flash Sales Improved Our Capacity to Handle Scale
37m5 Years of Rails Scaling to 80k RPS