To minimize Model Repository Service downtime, the Service Manager restarts the Model Repository Service on the same node or on a backup node if the Model Repository Service is unavailable.
The Model Repository Service fails over to a backup node in the following situations:
The Model Repository Service fails and the primary node is not available.
The Model Repository Service is running on a node that fails.
The Service Manager restarts the Model Repository Service based on domain property values set for the amount of time spent trying to restart the service and the maximum number of attempts to try within the restart period.
Model Repository Service clients are resilient to temporary connection failures during failover and restart of the service.