Scaling One of the Largest On-prem Managed MySQL Fleets in India
summary
Flipkart describes building Alter, an on‑prem MySQL platform that provisions production clusters in less than two minutes, manages 600 clusters over 1600 VMs with ~1000TB footprint, and runs on a control plane (MongoDB, Go + gRPC) plus a Zookeeper monitoring orchestrator and agents. Features include automated failover/auto‑promote, backup/restore to object store, binlog streaming, prebuilt Grafana dashboards, password rotation and no SSH access, built and run by a small central team.
AI-written summary · about
chapters (21)
- 0:00Start
- 2:43Agenda
- 4:58Overview of Databases at Flipkart Mysql
- 7:37Storage Flavors
- 10:55Developer Productivity
- 12:53The Core versus Context Problem
- 14:19How Do We Reduce the Overall Time To Create a Mysql Cluster
- 19:07Storage Choices
- 20:26Bring Your Own Hardware
- 21:47How To Ensure There Are Tight Security Controls
- 23:06Differentiate between Human and Machine Access
- 31:22High Availability
- 32:04What Do Failures Mean
- 37:43Db Tuning
- 42:09Seven Challenges How To Ensure Good Observability despite the Lack of Root Access
- 43:59Event Based Alerting
- 44:16Auditing and Event Logging on the Cluster
- 49:21Mysql Asynchronous Replication
- 51:15Automatic Disk Divergence between Primary and Replica
- 59:22Architecture
- 1:00:12Performance Monitoring Dashboards
related talks
50mRunning Databases Efficiently at Scale - Case Study of Flipka...
26mOperationalizing 19 years of infrastructure via MySQL-as-a-Service
42mBuilding Elastic Infrastructures
41mDemocratizing MySQL: From Cloud Managed to Kubernetes Managed
29mFlipkart: Scaling Aerospike at Flipkart with Kubernetes
32mData @Scale Seattle- Eric Hwang introduced by Ganapathy Krishnamoorthy
23mRunning MySQL on K8s. Deployment, Scaling, Backups and advanced configuration
06mScaling your applications with Cloud SQL for MySQL