How Devsisters launches blockbuster games in new countries with ease
summary
Devsisters migrated to CockroachDB in 2019 and uses it to scale Cookie Run into new regions including China on Tencent Cloud; production runs CockroachDB on EKS with a dedicated node pattern, Datadog monitoring, and EBS volumes. Operational rules: roll three nodes at a time after a failed decommission of 7–8 nodes, set rebalancing chunk to 128 MB, and watch IOPS/q‑length and unavailable range metrics. Analytics admission control cuts P95 query latency from ~2s to ~100ms, and a Checkpoint webhook prevents accidental StatefulSet/PVC deletes so recovery succeeds.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 0:34About Devsisters and the Cookie Run franchise
- 1:21About ChangWon Lee, Lead DevOps Engineer at Devsisters
- 2:06What does it mean to be on a Central DevOps team?
- 2:45How CockroachDB powers Cookie Run
- 4:38Why did Devsisters choose CockroachDB?
- 6:33CockroachDB vs DynamoDB and RDS
- 6:54CockroachDB on EKS
- 7:35How to monitor CockroachDB
- 8:46Rolling CockroachDB nodes without downtime
- 10:07How to address EBS hiccups
- 11:43Admission control for analytics workloads
- 12:47How to introduce a fail-preventing Webhook
- 16:55Characteristics of mobile game development
- 19:00Serverfarm: Deploying isolated server applications
- 21:11Serverfarms and CockroachDB
- 22:16Automation: Using Kubernetes jobs
- 22:45How Devsisters uses Terraform
- 24:30Why you shouldn't reinvent the wheel
- 25:40Key points recap
related talks
48mBuilding mission-critical applications with CockroachDB
31mFuture-proofing operational data at scale
15mScaling to support America’s #1 sportsbook
14mHow CockroachDB can enable Multi-Region Deployments and Data Locality 
33mWhat We’ve Learned Building a Multi-Region DBaaS on Kubernetes
15mScaling to support America’s #1 sportsbook
28mGlobal Payments: Subsecond Performance Worldwide
23mWhy Squarespace adopted CockroachDB for better global performance at scale