Skip to main content
SMKRV MCP Studio uses a Redis-backed queue to control concurrent query execution per database connection. This prevents overwhelming databases with too many simultaneous queries.

​
Overview

When an MCP client calls a tool that executes a SQL query:
  1. The tool requests a query slot for its bound connection
  2. If a slot is available (under max_concurrent_queries), the query executes immediately
  3. If all slots are taken, the query is placed in a FIFO queue
  4. When a running query completes, the next queued query is released
  5. If a queued query waits longer than queue_timeout_seconds, it is rejected

​
Per-Connection Settings

Configure queue settings in the connection editor: go to Connections, open a connection, and scroll to the Queue section. Each connection has independent settings:

​
Choosing Values

  • Small databases or shared hosts: Keep max_concurrent_queries low (2-5)
  • Dedicated database servers: Increase to 20-50 depending on capacity
  • Long-running analytical queries: Increase queue_timeout_seconds to 120-300
  • Quick OLTP queries: Keep timeout at 10-30 seconds

​
Queue Monitor

The Queue Monitor tab on the Server page displays real-time metrics:

​
Fallback Behavior

When Redis is unavailable, MCP Studio falls back to in-memory concurrency control:

​
Troubleshooting

​
Queries Timing Out in Queue

  • Increase queue_timeout_seconds
  • Increase max_concurrent_queries if the database can handle more load
  • Check for long-running queries holding slots

​
Redis Connection Failed

5-container docker-compose setup:
All-in-one Docker Hub image:

​
Queue Not Limiting Queries

  • Verify queue_enabled is true on the connection
  • Confirm Redis is connected (check the indicator on the Server page)

​
See Also

Connections

Configure queue settings per connection

Deployment

Deploy the MCP server that uses the query queue

Configuration

Redis and queue environment variables

Troubleshooting

Redis connection issues and debugging tips