Reducing CPU Usage by Utilizing readPreference: 'secondaryPreferred' in MongoDB

Hello MongoDB community!

Environment settings:
5 shards M60
3 nodes for each shard
Mongo 5.0

I’m working on a project where we’ve observed significant traffic in our database operations. Currently, we handle a total of 48,842 operations per second, with approximately 52.5% of these operations being reads and 47.5% being write operations.

Here are some key metrics from the latest operations of the week:

  • Read operations per second: 25,685
    • query: 2,194
    • getmore: 23,491
  • Write operations per second: 23,157
    • update: 7,004
    • insert: 1,858
    • cmd: 14,295

We are exploring the possibility of optimizing performance and load distribution in our MongoDB environment, and we have considered changing the read preference to ‘secondaryPreferred.’


We want to calculate how much we can reduce CPU usage by moving a portion of read operations to secondary nodes using the secondaryPreferred read preference.


  1. Has anyone implemented this configuration before and could share their experience?
  2. How can I calculate the estimated reduction in CPU usage by switching to readPreference: 'secondaryPreferred'?
  3. Are there specific factors I should consider when making this change in a production environment?

We appreciate any advice, past experiences, or information that can help us make an informed decision. Thank you in advance for your collaboration and expertise!

We use it for our clusters.

Just do a quick benchmark.