Registering Server Processes for HTTP Load Balancing 

When you scale your cluster by adding a new Java instance, you expect that at a certain point the new cluster elements will be registered with the system to participate in client requests load balancing. Concerning HTTP load balancing, new server processes for load balancing are not registered immediately after they go to started mode. Started mode means that all Enterprise Runtime managers and service components of the server process are successfully started. However, there is a lag before the applications are synchronized and started by the Deploy Service (this concerns only applications that must be automatically started at server process startup!). For the time being of this lag, the corresponding server process is not available for HTTP requests load balancing. After the applications startup process is completed, the server process is registered for load balancing on the Java dispatcher, and SAP Web Dispatcher (if running in an ABAP + Java Web AS scenario).

This feature addresses the possible pitfall situations of getting a 503 “Application stopped” error response if a client request is dispatched to a server process where the application has not been started yet. It concerns the load balancing on two different levels:

·        Homogeneous HTTP load balancing between server processes within a single Java instance

Heterogeneous load balancing within a single Java instance is not an issue since it uses session information to dispatch requests to the appropriate server processes.

·        Load balancing between many Java instances (SAP Web Dispatcher level).