This document describes the issue when the Deployment Manager (DM) resource mapping is deleted after some time and provides a solution.
There are a few sites defined in DM Release 7.0.2 and a few data sources have custom resource mapping from UAT to Production, which is a typical setup. These are deleted after a short time without any specific action from the user. It is to be noted that the UAT is a cluster and it's load balancer is used in the mapping in DM.
DM does not support the configuration where the "site" is a load balancer. DM has not been designed to support this configuration. A DM "site" is defined in the documentation as a CIS instance (not a cluster). The impact is that resource IDs are not shared across cluster nodes, which leads to inconsistent behavior.
Even when the mapping is done to a single node/instance, if the resource IDs are changed due to any activity such as CAR import and so on, the behavior is inconsistent. Resource IDs are used by endpoint mappings in order to keep track of resources when they get relocated or renamed. There are also CAR import scenarios where the import might replace a preexisting datasource at a specific location with a new data source of a different data source kind. In this case it probably does not make sense to preserve the endpoint mappings. In this case it would make sense to drop the mappings even if the data source path and resource type remain the same.
The solution for this issue is to use a Cisco Information Server (CIS) instance instead of a load balancer in the mapping.