An overview of PostgreSQL’s recovery mechanism, covering WAL replay and how the same core logic powers crash recovery, PITR, and replication.

CYBERTEC PostgreSQL

PostgreSQL's recovery mechanism uses Write-Ahead Logging (WAL) to restore databases to consistent states after failures. The recovery process involves replaying WAL records through a lifecycle managed by StartupProcessMain, which determines recovery type (crash, archive, or standby) by examining control files and signal files. The core PerformWalRecovery loop reads WAL records, applies them via resource managers, and continues until reaching a consistent state defined by recovery targets or available WAL. The same underlying mechanism powers crash recovery, point-in-time recovery (PITR), replication, and hot standby, with consistency points varying based on recovery type and configuration parameters like checkpoint settings, recovery targets, and prefetch buffers.

PostgreSQL Recovery Internals