How to Check ParadeDB Memory Usage in PostgreSQL

How do I check ParadeDB memory usage in PostgreSQL?

paradedb.memory_usage() returns detailed shared-buffer and local memory statistics for all ParadeDB indexes in the current PostgreSQL database.

Welcome to the Galaxy, Guardian!

Oops! Something went wrong while submitting the form.

Description

Example H2

Example H3

Why monitor ParadeDB memory usage?

Memory-heavy vector indexes can slow queries and evict useful data from shared buffers. Frequent checks help you tune work_mem, shared_buffers, and autovacuum settings before problems appear.

What does paradedb.memory_usage() show?

The function lists each ParadeDB index, its size on disk, the portion cached in shared buffers, local work_mem consumed by the last operation, and the percent of total shared_buffers in use.

How do I call paradedb.memory_usage()?

Run the function in any database where the ParadeDB extension is installed. Superuser rights are not required because the view is SECURITY DEFINER.

Step-by-step example

1. Connect to your analytics database.
2. Execute SELECT * FROM paradedb.memory_usage();
3. Review the shared_mb and work_mem_mb columns for spikes.

How to interpret the output?

shared_mb > 75% of shared_buffers signals pressure; increase shared_buffers or drop rarely used indexes. work_mem_mb > current work_mem means sorts spilled to disk; raise work_mem or batch your queries.

Best practices for ParadeDB memory

Create indexes only on frequently searched embeddings. Use VACUUM (INDEX_CLEANUP) after large deletions to reclaim space. Schedule REINDEX during low-traffic windows to shrink bloat.

Common mistakes

Creating huge HNSW indexes without raising shared_buffers leads to cache eviction. Forgetting to set maintenance_work_mem high enough slows bulk index builds.

Why How to Check ParadeDB Memory Usage in PostgreSQL is important

How to Check ParadeDB Memory Usage in PostgreSQL Example Usage


-- Example: Find ParadeDB indexes consuming over 500 MB shared memory in an ecommerce DB
SELECT table_name, index_name, shared_mb
FROM paradedb.memory_usage()
WHERE shared_mb > 500
ORDER BY shared_mb DESC;

How to Check ParadeDB Memory Usage in PostgreSQL Syntax


-- Enable extension (once per database)
CREATE EXTENSION IF NOT EXISTS paradedb;

-- Check memory usage for all ParadeDB indexes
SELECT *
FROM paradedb.memory_usage();

-- Filter for a single table (e.g., Products embeddings index)
SELECT *
FROM paradedb.memory_usage()
WHERE table_name = 'products';

Common Mistakes

Mistake: Running paradedb.memory_usage() on the wrong database. ParadeDB tracks indexes per-database, so switching \\c to another DB returns empty results. Fix: Connect to the database that hosts your ParadeDB indexes before executing the function.
Mistake: Assuming shared_mb equals on-disk size. The column only shows data cached in shared buffers. Fix: Compare shared_mb with index_size to understand total versus cached memory.