Debugging in a Cluster Configuration 

A typical cluster configuration consists of several instances (represented by 00, 03, and 07 in the following figure). Each of these instances, in turn, consists of one dispatcher and one or more server processes (Server0, Server1, ... ServerN in the figure). The Software Deployment Manager (SDM) node is present in only one of the Java instances.

As you can see, the situation here is much more complicated. We must first select the server processes that can be involved in the debugging (the debuggable nodes), and the nodes that will never be used for debugging (the productive nodes). This does not mean that all debuggable nodes will be used for debugging all the time. This only means that if a problem with an application occurs, you can perform debugging on them.

Then, before starting the debugging, first you need to make sure the debuggable nodes do not participate in the cluster’s load balancing mechanism, and no client requests are sent to them.

See Preparing the Debugging Process for the detailed steps to configuring the J2EE Engine for debugging.

For productive systems with a small number of processes (one or two) in each Java instance, we recommend creating a separate process used only for debugging, and deleting it after the debugging. You can also create a whole new Java instance with all processes used for debugging only. This will not disturb the overall load-balancing among the different instances. If you are using the SAP Web Dispatcher as a load balancer, it will automatically skip sending any requests to that Java instance. If you are using a third-party load balancer, you must skip adding the instance for debugging to it.