3 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.
Database 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 keep PostgreSQL, MySQL, MariaDB, and SQL Server easier to operate.
Featured Guides
Slow query optimization
Find the queries dragging your Postgres down, read their plans, and fix them in order of impact.
Read the guide →Index optimization
Which indexes to add, which to drop, and how to spot redundant index debt before it costs you writes.
Read the guide →Vacuum guide
How autovacuum works, when it falls behind, and how to keep bloat from eating your storage and your buffer cache.
Read the guide →Connection pooling
Sizing pools, PgBouncer modes, and how to stop idle-in-transaction sessions from exhausting max_connections.
Read the guide →Replication monitoring
Track lag, WAL throughput, and replication slots before they turn into stale reads or a full disk.
Read the guide →Browse by Topic
Backups and Durability
9 articlesBackups, restores, PITR, archive gaps, and recovery drills for PostgreSQL.
Cloud Providers
5 articlesAWS RDS, Azure Database for PostgreSQL, Google Cloud SQL, AlloyDB, and managed PostgreSQL monitoring notes.
Connections and Pooling
10 articlesConnection storms, pool saturation, PgBouncer behavior, and max_connections pressure.
CPU, Memory, and I/O
17 articlesCPU, memory, disk, temp files, checkpoints, buffers, and workload pressure.
Indexes
37 articlesIndex design, index debt, reindexing, constraint indexes, and evidence-driven DDL.
Locks and Transactions
22 articlesLock waits, deadlocks, blocking chains, idle transactions, and transaction hygiene.
MariaDB
56 articlesGalera clustering, InnoDB and Aria internals, replication, optimizer behavior, and MariaDB operations notes.
MySQL
107 articlesInnoDB internals, replication, locking, query performance, and MySQL operations notes.
pgvector and RAG
15 articlesHNSW, IVFFlat, recall, embedding search, and production RAG performance notes.
Replication and WAL
28 articlesReplica lag, WAL growth, failover readiness, hot standby behavior, and replication slots.
Slow Queries
38 articlesQuery plans, EXPLAIN ANALYZE, planner regressions, pagination, joins, and statistics.
SQL Server
47 articlesWait statistics, tempdb, Query Store, Always On availability groups, and SQL Server operations notes.
Vacuum and Bloat
16 articlesAutovacuum, dead tuples, bloat, wraparound, MVCC, freeze age, and visibility.
Latest Field Notes
3 min read
Chunking Is a Database Design Problem, Not a Prompting Trick
Chunk size and overlap decide retrieval quality, citation accuracy, freshness, permissions, and cost. Treat chunks as serving data with ownership.
9 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.
6 min read
PostGIS in Production: Why Geo Queries Get Slow After the Demo
Slow PostGIS systems usually do not fail because geography is hard. They fail because the query shape stops the spatial index from pruning work early.
3 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.
3 min read
Filtered Vector Search: The Real Vector Database Problem
Vector search looks easy until tenant filters, permissions, freshness, and deleted content arrive. The hard part is not nearest neighbors; it is filtered recall under production rules.
3 min read
Multi-Tenant Vector Search: Namespaces, Metadata Filters, or Partitions?
Multi-tenant vector search is a correctness and isolation problem. Namespaces, filters, and partitions each fail differently under tenant skew and ACL rules.
3 min read
Hybrid Search in Production: BM25 Plus Vector Is Not a Magic Button
Hybrid search helps when exact terms and semantic meaning both matter. It fails when teams blend rankings without query intent, calibration, or evaluation.
7 min read
Vector Databases: How I Decide Between pgvector and a Dedicated Vector Store
The hard vector database decision is not pgvector versus Pinecone on a checklist. It is whether your filters, recall target, update rate, and incident budget still fit inside Postgres.
7 min read
Reading EXPLAIN ANALYZE Output Without Drowning
EXPLAIN ANALYZE is readable once you know the order: find the slow node, compare estimates to actuals, check loops, then confirm buffers and waits.