Route requests across clusters on Linux
The HTTP Server plug-is that comes with IBM WebSphere Application Server is typically used to balance requests for applications across members of the cluster. We can also use the On Demand Router (ODR) in dynamic, multicluster environments for routing. While an application can be mapped to more than one cluster, automatic plug-in generation does not provide routing or balancing traffic for the same application across multiple clusters. Multiple cluster environments with shared applications therefore cannot rely solely on WAS automatic plug-in generation to be able to route requests using a web server. One option in this scenario is developing a customized method for defining and maintaining the plug-in configuration file used by the web server to provide for the required routing of user requests.An important consideration in a multiple cluster environment is ensuring that all subsequent HTTP requests for a user are routed to the same cluster that processed the first HTTP request. The WebSphere Portal login processing depends upon preserving this cluster affinity during this initial time until the user has successfully logged in and session cookies maintain affinity. To guarantee that affinity is preserved during login, set the Navigator Service public.session parameter to a value of true; you should also set this parameter to true if we are using an ODR. Refer to "Portal Configuration Services" for information on how to configure this parameter.
Review the following considerations before configuring the On Demand Router (ODR) to route traffic to Portal clusters:
- Internal users can send requests directly to the ODR instead of through a front-end web server. When sending direct requests, configure the ODR to append a via header to the HTTP requests. Set the value of the ODR custom property http.compliance.via to true; see the "On demand router settings" link in the Related section for information.
This step is not required when sending user traffic through the web server to the ODR because the web server appends the via header to the HTTP request.
- The ODR can selectively route traffic to clusters based on the incoming URL. Configure IP alias values for the ODR and then define routing rules to associate user traffic for each IP alias to the appropriate WebSphere Portal cluster.
- We can use the ODR to load balance traffic among identical portal clusters. Configure a Multicluster Routing Policy (MCRP) for the ODR to identify the destination clusters and the type of load balancing; see the "Configuring the on demand router for multi-cluster failover and load balancing routing" link in the Related section for information.
If we are configuring the ODR to route traffic to remote portal static clusters using Generic Server Cluster definitions, the cell_name value used by the MCRP policy needs to be the local cellname where the ODR resides and not the remote cell where the portal cluster resides.
- We can also use the ODR to route traffic to remote portal clusters, both static and dynamic, by defining a generic server cluster for each target portal cluster; see the "Defining generic server clusters for remote ODR cells" link in the Related section for information.
If we are routing to remote static clusterm that use vertical cluster members, perform the optional step at the end to define a server custom property for each port in the generic server cluster.
Parent: Set up multiple clusters on Linux
Related:
WebSphere Virtual Enterprise Dynamic Clusters
Related reference:
Portal configuration services