Mar 26 · 4 min read · PostgreSQL will not tell you when an index is missing. It will dutifully execute a sequential scan across 50 million rows, return the correct result in 3.2 seconds, and move on. No error. No warning. Just a query that could run in 5 milliseconds taki...
Join discussion
Mar 24 · 3 min read · There is a class of PostgreSQL incident that is entirely self-inflicted: dropping an index from the primary that turns out to be critical on a replica. The index shows zero scans on the primary because transactional workloads never use it. But the an...
Join discussion
Mar 14 · 11 min read · Backend systems revolve around data. User accounts, payments, orders, logs, analytics, permissions — almost everything a backend service does eventually becomes data stored somewhere. Because of that,
Join discussionMar 12 · 4 min read · PostGIS Spatial Indexing: Why Your Queries Are Doing Sequential Scans Last year I helped a delivery platform debug why their "restaurants near me" feature was timing out during lunch rush. The restaurants table had 4.2 million rows. Every proximity q...
Join discussion
Mar 7 · 4 min read · Scaling pgvector: Memory, Quantization, and Index Build Strategies pgvector is the easiest way to add vector search to PostgreSQL. At small scale -- a few hundred thousand embeddings -- it just works. HNSW indexes deliver sub-millisecond similarity s...
Join discussion
Mar 5 · 4 min read · Your pgvector Queries Are Doing Sequential Scans -- Here's Why pgvector is the fastest way to add vector similarity search to an existing PostgreSQL database. It's also the fastest way to create a performance problem that hides in plain sight: uninde...
Join discussion
Mar 3 · 4 min read · pgvector Index Selection: IVFFlat vs HNSW for PostgreSQL Vector Search A team I worked with spent three weeks convinced their OpenAI embeddings were low quality. Their RAG pipeline returned vaguely related documents instead of the precise matches the...
Join discussion
Feb 27 · 5 min read · How to Find and Fix Missing Indexes in PostgreSQL A sequential scan on a 100-row table takes microseconds. A sequential scan on a 10-million-row table takes hundreds of milliseconds. The SQL is identical. The schema is identical. The only difference ...
Join discussion
