When Three Kubernetes Pods Try to Run One Database Migration

This article addresses a common Kubernetes issue where multiple replicas of an application attempt to run the same database migration, causing lock contention and pod failures. It explains why a startup command like 'prisma migrate deploy && node server.js' fails when scaled to three pods, as all pods race to execute the migration. The proposed solution is a migration coordinator using PostgreSQL advisory locks, allowing only one pod to run the migration while others wait and retry. The article also recommends using a Kubernetes Job when possible, as it provides better isolation and failure handling. It includes code examples for implementing the coordinator with Prisma and the 'pg' library, making it a practical guide for developers facing similar deployment challenges.