Glossary — in plain language

Checkpoint


In plain English

A checkpoint is the moment PostgreSQL flushes all the changes that are still only in memory (and in the WAL) down to the real data files, then records a “we are safe up to here” mark. After a crash, recovery only needs to replay WAL written since the last checkpoint.

Why it matters

Checkpoints cause bursts of disk writes. If they happen too often you see I/O spikes and warnings in the log; spacing them out (bigger max_wal_size, longer timeout) smooths I/O but lengthens recovery.

Was this helpful?

← Browse the full glossary