A JDBC provider and data source defined at cluster scope are configuration objects that apply to every member of the cluster, and a test of the data source is a genuine attempt to open a connection to the database, not a validation of the stored settings. The administrative console is served by the deployment manager, but the deployment manager does not hold the JDBC driver classes or open database connections itself. For resources at node, cluster, or cell scope, the test-connection request is forwarded to a node agent, which runs the test in its own JVM using the provider's class path and the native path settings configured for that node. Option C, the node agent, is therefore the process that performs the operation.
The other processes do not carry it out. The deployment manager, option B, coordinates the request and displays the result, and it owns the master configuration, yet it is not where the connection is made. The administrative agent, option A, performs the equivalent role for unfederated nodes, and it is not part of a cell managed by a deployment manager. The application server, option D, hosts the data source for applications at run time, but the console test at this scope does not run inside a cluster member.
An administrator should therefore remember that a successful test reflects the environment of the node agent's host. A cluster whose members are on different hosts may succeed in the test on one node and still fail on another, for example because of a firewall rule or a missing driver path, so each member's SystemOut.log should also be checked after the application is started.
[Reference: JDBC providers and data sources; Testing a data source connection; Resource scope in a cell; Node agent.]
Submit