Why we run two PostgreSQL clusters for one booking system: separating writes from reads at scale

A booking system has a split personality. Every search query wants to read availability across thousands of properties for a given date range. Every booking confirmation wants to write a row, lock inv