Scenario: Having a load balanced HEAT system that points to two HEAT application\web servers. When both machines are up, the log in time is as expected, however in tests when either machine is off, the log in time extends to 40s.
This behavior occurs because we need to ensure that all the Application Servers have an updated version cached as it is possible to modify the configuration from any of the servers, so they need to be able to communicate among them.
However, when a server is down, it is possible to turn it into inactive, so the available server will not try to contact to the other server in the load balancing. For this case, login into the ConfigDB tenant, open the Web Servers tenant, and uncheck the option "Server is Active" for the server that is down. Once saved the logon time will be normal.