PostgreSQL Checkpointer and Buffer Management Metrics Guide

Maxwritten_clean

• What it measures: Counts how often the background writer stops flushing dirty buffers because it reached the bgwriter_lru_maxpages limit per cycle.
• Why it matters: High values indicate that the background writer is frequently hitting its operational limits, forcing checkpoints or backend processes to handle remaining dirty buffers. This can lead to I/O spikes and uneven performance. Monitoring this metric helps identify when bgwriter parameters need adjustment to ensure smoother buffer management.
• Ideal value & Best Practice: Low values are ideal. If this metric is consistently high, increase bgwriter_lru_maxpages or adjust bgwriter_lru_multiplier to allow more aggressive background writing.

Percentage of Timed Checkpoints

• What it measures: The ratio of scheduled checkpoints (checkpoints_timed) to total checkpoints, showing whether checkpoints are occurring predictably or due to workload pressure.
• Why it matters: A high percentage of timed checkpoints indicates stable, predictable I/O patterns, while a high percentage of requested checkpoints suggests unexpected workload pressure that can cause I/O spikes and latency issues.
• Ideal value & Best Practice: Target >70% timed checkpoints. If requested checkpoints dominate, consider increasing max_wal_size or optimizing write patterns to reduce sudden WAL growth.

Bgwriter Buffer Cleaning Efficiency

• What it measures: How effectively the background writer handles buffer flushing compared to backend processes.
• Why it matters: High efficiency means the bgwriter is proactively managing buffers, reducing I/O pressure on backends. Low efficiency forces backends to handle flushing, which can cause contention and performance degradation.
• Ideal value & Best Practice: Aim for high bgwriter efficiency. If efficiency is low, tune bgwriter_delay, bgwriter_lru_maxpages, and bgwriter_lru_multiplier to make the background writer more aggressive.

Checkpoint Contribution to Buffer Writes

• What it measures: The percentage of buffer writes triggered by checkpoints versus other processes.
• Why it matters: Low checkpoint contribution indicates balanced I/O distribution, while high contribution suggests checkpoints are handling excessive write loads, potentially causing performance spikes during checkpoint operations.
• Ideal value & Best Practice: Lower percentages are better. If checkpoints dominate writes, optimize checkpoint_timeout and max_wal_size to distribute I/O more evenly.

Buffers_backend_fsync

• What it measures: How frequently backend processes must manually invoke fsync() to flush buffers.
• Why it matters: Frequent backend fsyncs indicate that the background writer and checkpoints aren't keeping up with buffer management, forcing backends to handle disk synchronization and potentially causing I/O contention.
• Ideal value & Best Practice: Low values are optimal. High values suggest need for better bgwriter tuning or increased shared_buffers to reduce backend flushing pressure.

Buffers_alloc

• What it measures: The total number of new memory buffers allocated for data management.
• Why it matters: High allocation rates may indicate memory pressure, frequent cache evictions, or inefficient query patterns that aren't leveraging caching effectively.
• Ideal value & Best Practice: Monitor for sudden spikes. Consistently high values may indicate need for increased shared_buffers or query optimization to improve cache utilization.

Checkpoint Sync Overhead

• What it measures: The percentage of checkpoint time spent on data synchronization versus actual writing.
• Why it matters: High sync overhead indicates potential storage latency issues or too-frequent checkpoints, while low overhead suggests efficient synchronization typical of fast storage systems.
• Ideal value & Best Practice: Lower percentages indicate better performance. High values may require storage optimization or checkpoint interval adjustments.

---

Try pghealth Free Today 🚀

Start your journey toward a healthier PostgreSQL with pghealth.
You can explore all features immediately with a free trial — no installation required.

👉 Start Free Trial