You already know how to write a join. You've run EXPLAIN. You've been on the wrong end of a deadlock at least once. And still, occasionally, Postgres wakes you up at three in the morning. Same code, same hardware, something changed, and you don't know what.
PostgreSQL: From MVCC to Production is the book that closes that gap. Not a feature tour. Not a reference. A focused, opinionated build-up of the mental model that makes everything else legible: how MVCC works and what it costs, why vacuum matters and when it falls behind, how the query planner makes decisions and how to disagree with it, what every connection actually owns, and how replication, failover, backups, and tuning fit together into a system you can reason about.
The book covers twenty-eight chapters across four parts. Part I lays the foundation: MVCC, storage, WAL, processes and memory, indexes from first principles. Part II is querying like an expert: plan shapes, index types, statistics, GIN, partial and expression indexes, concurrent operations. Part III is concurrency: row locking, MVCC contention, deadlocks, autovacuum, connection pooling with PgBouncer. Part IV operates Postgres at scale: streaming replication, logical replication, Patroni-based HA, pgBackRest backups, configuration tuning, capacity planning, and incident runbooks.
Every example runs against the cinetrack dataset, a movie-and-reviews schema with MovieLens 25M data, reproducible in a single docker-compose up. Pure SQL throughout. No framework code.
Targets Postgres 17. Written for senior engineers who want to stop discovering things the hard way.
