The JusDB Database Engineering Blog

Add vector similarity search to PostgreSQL with pgvector. Covers HNSW and IVFFlat indexes, cosine similarity queries, OpenAI embedding integration, and hybrid search.

PostgreSQL RLS enforces tenant isolation at the database level, bypassing application filtering bugs entirely. Learn to implement RLS policies, index for performance, and test policies before production.

PostgreSQL's logging system is powerful but requires careful configuration to be useful in production. Learn how to capture slow queries, connection events, and DDL changes without drowning in noise.

Upgrade PostgreSQL major versions with sub-second downtime using logical replication. Step-by-step guide covering schema copy, subscription setup, lag monitoring, and sequence resync.

Implement PostgreSQL declarative range partitioning with partition pruning, propagated indexes, instant DETACH for archival, and pg_partman for automated partition management.

Build a complete database observability stack with Prometheus metrics, slow query logs, and OpenTelemetry distributed traces. Includes alert rules for replication lag and cache hit rate.

Master pg_stat_statements for PostgreSQL query analysis. Find top queries by total time, detect plan instability via stddev, identify I/O-heavy queries, and track WAL generation.

Right-size your PostgreSQL connection pool using the (cores*2)+1 formula. PgBouncer configuration, per-user limits, pool saturation monitoring, and transaction mode pitfalls.

PostgreSQL 17 ships eager freezing and improved autovacuum prioritization, the most significant vacuum improvements in a decade. Learn how to tune autovacuum for high-write tables and prevent transaction ID wraparound.

Guide the PostgreSQL query planner with enable_xxx settings, pg_hint_plan extension, and extended statistics for correlated columns. Fix bad plans without rewriting queries.

Use pgbench to benchmark PostgreSQL before and after configuration changes. Covers scale factor selection, TPC-B and read-only workloads, custom SQL scripts, and latency percentiles.

Use pg_repack to reclaim PostgreSQL table bloat without the exclusive lock of VACUUM FULL. Covers installation, table vs index repack, disk space requirements, and throttling.