How One Engineer Rewrote a Legacy Database Without Downtime
from The Software Engineering Podcast with Fexingo: Code, Architecture, and Engineering Best Practices Hosted by Fexingo
Episode 75 of The Software Engineering Podcast tells the true story of a senior engineer at a mid-sized logistics firm who migrated a 15-year-old PostgreSQL monolith to a sharded, horizontally-scalable architecture without a single minute of planned downtime. The project touched 4.3 million lines of trigger code and 2,800 stored procedures. By combining logical replication, a write-ahead log change data capture pipeline, and a phased cutover with canary reads, the team moved 12 terabytes of data incrementally over six weeks. The episode breaks down the exact strategy: how they avoided dual-write complexity, handled schema drift, and rolled back within 90 seconds when a hot-spot partition caused latency spikes. Lucas and Luna discuss the tradeoffs between trigger-based replication versus streaming replication, why they chose NOT to use an ORM abstraction layer, and what happened when a foreign key constraint broke the CDC pipeline at 2 AM. This is a deep, practical look at legacy database modernisation for engineers facing similar migrations. #LegacyDatabaseMigration #PostgreSQL #DatabaseSharding #ChangeDataCapture #ZeroDowntimeMigration #LogicalReplication #SoftwareEngineering #TechPodcast #DatabaseArchitecture #ProductionEngineering #LucasAndLuna #FexingoBusiness #BusinessPodcast #EngineeringBestPractices #DataMigration #Postgres #WriteAheadLog #CanaryDeployments Keep every episode free: buymeacoffee.com/fexingo
Embed this episode
No transcript yet.