We had recently fielded questions about high availability of the MMS App servers (as opposed to just the MMS agent) and here is a blurb we came up with help from the MMS team. Please consider adding this information to

https://mms.mongodb.com/help-hosted/current/mms-manual.pdf:

"The On-Prem MMS app server is designed to be horizontally scalable. The key is to configure each of the MMS servers conf-mms.properties file to point their mms.centralUrl to a load balancer. Then configure the load balancer to balance between the pool of MMS app servers. The MMS app servers are stateless between requests and any of them can handle requests as long as they are reading from the same backing MongoDB replica set. In such an on-prem configuration, where multiple MMS app servers are present for high availability, for the backing MongoDB a replica set (rather than a stand-alone) is recommended. The following steps are required:

1. Configure a load balancer with the pool of MMS app server locations (Since load balancers are very specific to the environment and can have various possible configurations, we don't have a recommendation for load balancers).
2. Configure the 'mms.centralUrl' property of each MMS app server's conf-mms.properties file to point to the load balanced URL
3. Edit the conf-mms.properties file on each MMS app server to define the replication hosts for the backing MongoDB"