PostgreSQL Field Notes

Notes for the problems that show up after launch: bad plans, awkward migrations, index debt, vacuum pressure, replica lag, and the small decisions that make PostgreSQL easier to operate.

PostgreSQL Topic Hubs

PostgreSQL Guide (Pillar)

pgvector Monitoring Query Optimizer Sizing Slow Query Monitoring Query Monitoring

PostgreSQL Tools

HNSW Tuning Lab Plan Autopsy Index Rollout

Tool Comparisons

Compare PostgreSQL monitoring tools

Browse by topic

pgvector and RAG Indexes Slow Queries Locks and Transactions Backups and Durability Replication and WAL Connections and Pooling CPU, Memory, and I/O Vacuum and Bloat Cloud Providers

May 8, 202610 min read

PostgreSQL WAL and Checkpoints: Tuning Write Spikes Without Guesswork

Write-heavy PostgreSQL systems usually fail through WAL pressure, checkpoint I/O, replication lag, or storage stalls. The fix starts with measuring the write path, not raising random knobs.

May 8, 202610 min read

pg_stat_statements: Turning Slow Query Noise Into a Tuning Queue

pg_stat_statements is where PostgreSQL performance work becomes concrete. The trick is ranking by impact, variance, I/O, and calls instead of staring at one dramatic slow query.

May 8, 202611 min read

EXPLAIN ANALYZE BUFFERS: Reading PostgreSQL Plans Like a Production Trace

EXPLAIN ANALYZE is not just a plan tree. With BUFFERS and real row counts, it becomes a production trace for estimates, I/O, loops, spills, and wasted work.

May 8, 202611 min read

Autovacuum and Query Performance: When Dead Tuples Become User Latency

Autovacuum is not just cleanup. When it falls behind, dead tuples, bloated indexes, stale statistics, and transaction age turn into slow queries and operational risk.

May 8, 202610 min read

PostgreSQL Index Performance: Finding Unused, Duplicate, and Expensive Indexes

Indexes speed reads and tax writes. Production tuning means finding indexes that help real queries, indexes that nobody uses, and indexes that make every insert slower.

May 8, 202611 min read

PostgreSQL Join Performance: Fixing Slow Joins Without Guessing

Slow joins are rarely solved by memorizing join types. The real workflow is finding row estimate errors, missing indexes, bad join order, and parameter-sensitive plans.

May 4, 20265 min read

Rerankers in RAG: The Expensive Fix That Only Works After Retrieval

Rerankers can improve answer quality, but they cannot recover evidence that retrieval never found. Use them after recall, cost, and latency are understood.

May 4, 20266 min read

pgvector HNSW Tuning: Why Default Settings Quietly Kill Recall

HNSW defaults can look fast while missing useful results. Production tuning needs recall@k, p99, index build time, memory, and filtered result count together.

May 4, 20266 min read

RAG Quality Metrics: Stop Measuring Only Latency

A fast RAG system can still be wrong. Production teams need recall@k, MRR, answerability, citation coverage, freshness, no-hit rate, and drift signals.

May 4, 20265 min read

Vector Deletes, Freshness, and Permissions: The Hidden RAG Incident

A RAG system is unsafe if deleted documents, permission changes, and stale embeddings can still appear in answers. Freshness is part of correctness.

May 4, 202610 min read

PostgreSQL 18 New Features: What Actually Matters in Production

PostgreSQL 18 is already here. The useful question is not whether the release is exciting, but which features reduce real production pain and which upgrade risks need rehearsal.

May 4, 20266 min read

Re-Embedding Without Breaking Production Search

Embedding model upgrades are migrations. Versioned embeddings, dual indexes, shadow queries, backfills, and rollback decide whether search quality survives.