Load balancing and failover with IBM HTTP Server

Edit online

In a network deployment environment, you can configure a routing server, such as IBM HTTP Server, WebSphere Application Server proxy server, or others, as a proxy server for workload balancing and failover purposes. Instead of incoming HTTP requests going directly to an application server, they go to the proxy server, which then distributes the requests across multiple application servers that perform the work.

Tip: You do not have to use IBM® HTTP Server for load balancing, you can use any other type of routing server, for example,WebSphere® Application Server proxy server, any other web server, reverse proxy server, or network layer IP sprayer.
Although the specific configuration steps differ depending on your environment, the following general tasks can help you plan for and implement workload balancing and failover with IBM HTTP Server.
  1. Install IBM HTTP Server.
  2. Install and configure the appropriate web server plug-in.
  3. Configure Secure Sockets Layer (SSL) between the deployment manager for WebSphere Application Server and the IBM HTTP Server administration server.. You must configure the Application Server to accept a self-signed certificate from IBM HTTP Server so that SSL connections are accepted and transactions are completed.
  4. Customize the Workflow Server or Workflow Center cluster so that the configuration file points to the web server and communication is enabled for HTTP over SSL or HTTP Secure (HTTPs). See Customizing Business Automation Workflow to work with a web server.
Note:
When you set up a cluster installation with a load balancer, set up sticky sessions by following the instructions for your load balancer. Otherwise, you might see one or more of the following results:
  • Unusual or unexpected User is not authenticated messages.
  • A node goes offline and the portal becomes unavailable while other nodes are still running.
  • Inconsistent data is returned when you are using the common URL set up by the load balancer. However, you see correct data if you do the tasks directly on a node.
  • Between tasks, users are asked to log in again before the session times out.