Back to Glossary

PostgreSQL Performance & Recovery

Checkpoint

Learn what PostgreSQL checkpoints do, how they affect crash recovery time and write performance, and which signals to monitor.

Definition

A point in the WAL where all dirty data pages have been written to disk, providing a consistent recovery point.

What a Checkpoint Is

A checkpoint is when PostgreSQL flushes dirty buffers to data files and records a recovery boundary in WAL.

After a crash, PostgreSQL replays WAL from the last checkpoint, so checkpoint behavior influences restart recovery time.

Why Checkpoints Matter in Production

Aggressive checkpoints can increase write pressure and latency variance. Sparse checkpoints can lengthen recovery replay windows.

Checkpoint tuning is therefore a tradeoff between steady-state performance and incident-time recovery speed.

  • Too frequent: higher IO and write amplification
  • Too infrequent: longer replay after crash
  • Balanced: predictable latency with acceptable recovery time

Operational Signals to Watch

  • Checkpoint frequency and duration
  • Background writer vs backend write pressure
  • WAL generation rate around checkpoint cycles
  • Recovery replay time observed in drills

How Checkpoints Connect to PITR

Checkpoints do not replace WAL archiving, but they shape how much WAL must be replayed for crash and restore workflows.

For full DR readiness, checkpoint strategy should be validated together with base backups and PITR drills.

Related Glossary Topics

Copy-on-Write (COW) Connection Pooling Clone (Database Clone) ACID Atomicity Archiving (WAL Archiving) archive_command archive_mode

Frequently Asked Questions

Do checkpoints improve PostgreSQL performance?
Properly tuned checkpoints improve stability; poorly tuned checkpoints can create latency spikes.
Do checkpoints reduce recovery time?
Yes, generally. More recent checkpoints reduce WAL replay distance after crashes.
Can checkpoints replace backups?
No. Checkpoints are runtime consistency events, not backup artifacts.
What causes checkpoint-related latency spikes?
High dirty-buffer flush pressure and IO contention during checkpoint execution are common causes.
How should teams tune checkpoint strategy?
Tune for workload characteristics, monitor latency + WAL behavior, and validate outcomes in realistic load tests.