Hi Neha,
Approach 1:
UD1 and UD2 operating in failover mode is defined in the Siteminder user directory setup.
If UD1 is not contactable, PS will failover to the UD2.
With Policy Server log, the error returned from the database will be logged there e.g: bind error or server not contactable. Error message is associated with specific database (UD1/ UD2).
Administrator can find out if certain database was unavailable at a certain timeframe and when did the Policy Server fail back to the primary database.
Policy Server trace can clearly identify the database (UD1/ UD2) of which the authentication/ validation request was made against.
Approach 2:
UD is defined in Siteminder user directory setup while there are 2 database operating in failover mode defined in ODBC driver setup.
If the primary server is not contactable, the ODBC driver will failover to the next server in line.
However, the Policy Server will always reference the connection to user store database as UD, as if it's a single database server. It will not be able to differentiate if it's going to primary or secondary server.
With this approach you will need ODBC tracing to be enabled to track any error returned from the database server or track which database server was the request made against.
During the failover process, your application may experience a short pause while the driver establishes a connection on an alternate server. If your application is time-sensitive (a real-time customer order application, for example) and cannot absorb this wait, you can set the ODBC Driver Failover Preconnect connection option to true. Setting the Failover Preconnect option to true instructs the driver to establish connections to the primary server and an alternate server at the same time. Your application uses the first connection that is successfully established. If this connection to the database is lost at a later time, the driver saves time in reestablishing the connection on the server to which it fails over because it can use the spare connection in its failover process.
Best regards,
Kelly