baekenough/postgres-best-practices

PostgreSQL Best Practices

Query Optimization

EXPLAIN ANALYZE (CRITICAL)

• Use EXPLAIN ANALYZE to understand query plans
• Identify slow operations: Seq Scan, Nested Loop
• Check row estimates vs actual rows
• Monitor buffers: shared hit vs read

Indexing (CRITICAL)

• B-tree: default, most use cases
• GIN: JSONB, arrays, full-text search
• GiST: geometry, range types
• BRIN: large sequential tables (time-series)
• Partial indexes: filtered queries
• Covering indexes (INCLUDE): avoid heap fetches

Index Maintenance

• Create indexes concurrently: CREATE INDEX CONCURRENTLY
• Monitor usage: pg_stat_user_indexes
• Remove unused indexes
• Reindex bloated indexes

Table Design

Partitioning (HIGH)

• Range partitioning: time-series data
• List partitioning: categorical data
• Hash partitioning: even distribution
• Declarative partitioning (PG 10+)

Data Types

• Use appropriate types (int vs bigint, varchar vs text)
• JSONB for semi-structured data
• Arrays for multi-value columns
• UUIDs for distributed IDs

Performance Tuning

Vacuum and Autovacuum

• Autovacuum: default enabled
• Monitor bloat: pg_stat_user_tables
• Tune autovacuum thresholds
• Manual VACUUM for large updates

Connection Pooling

• Use pgBouncer or PgPool
• Transaction pooling for short transactions
• Session pooling for long transactions
• Max connections: tune based on workload

Configuration

• shared_buffers: 25% of RAM
• work_mem: per operation, tune carefully
• effective_cache_size: 50-75% of RAM
• random_page_cost: 1.1 for SSD

References

• PostgreSQL Performance Optimization