High CPU Usage in MongoDB with Debezium CDC Source Connector

Despite implementing various CPU optimization strategies in our connector configurations, we continue to see elevated CPU usage on MongoDB. We have already:

  1. Increased polling intervals to 6 minutes (mongodb.poll.interval.ms: 360000)
  2. Optimized batch sizes for bulk processing (max.batch.size: 500)
  3. Limited the number of concurrent tasks (1 per collection)
  4. Implemented connection pooling and timeout optimizations
  5. Set snapshot mode to “never” to avoid initial snapshots
  6. Added custom operation filtering to reduce unnecessary events

What MongoDB Atlas-side optimizations can reduce CPU impact of CDC operations across multiple collections ?
2. What are Best practices for MongoDB configuration when using Debezium with multiple collections ?
3. What Recommendations for tuning Atlas cluster settings for high-volume CDC workloads ?