In this project we're going to add:

• A couple Query knobs (sever parameters)
• New metrics (either serverStatus or queryStats)

  Description of Linked Ticket

  Epic Summary

  Summary 

The performance tiger team has identified a recurring issue called "multi-plan storms." This occurs when many concurrent operations engage in multi-planning for the same problematic query shape. These storms can lead to out-of-memory issues, causing the mongod process to slow down significantly. The work involved in multi-planning prevents the query subsystem from completing any plans, caching the winning plan, and improving future planning attempts.

Motivation

Addressing multi-plan storms is crucial for improving the performance of the query system. The current situation, where multiple operations simultaneously plan the same query shape, results in high memory and CPU usage. This not only slows down the mongod process but also leads to instability. By introducing a serialization point, we aim to reduce redundant work and resource consumption, thereby improving the overall performance and stability of the system.

Documentation

Investigation Report: A Trio of OOM-inducing Workload Patterns
Product Description
Scope
Technical Design
Docs Update