MongoDB has introduced several changes in its storage engine and database as a service offering, making it essential to set up alerts to monitor the performance of MongoDB deployments. The old Lock % metric is no longer relevant due to the use of Document-level locking, which improves performance for most workloads. Instead, metrics like "Tickets Available" and "Queues" can be used to detect queries that took longer than expected due to load. Disk Utilization should also be monitored, with alerts set at 90% or lower, depending on the specific needs of the application. CPU Steal is another critical metric to monitor, with an alert set at 10% to prevent slow queries and difficulty connecting. Finally, indexes can impact performance, especially if too many are created, so monitoring metrics like "Query Targeting Scanned/Returned" can help identify inefficient queries. Each MongoDB Atlas instance size has a connection limit, which should be monitored with an alert set between 80-90% of the limit to prevent connection issues.