All: We have Apache 2 load balancing two tomcat 4.1.x workers. There are two primary applications running under tomcat. One app has long session times - eg the user logs-in in the morning and doesn't end their session until evening. The other app has short sessions - a connection is made, some work is performed, and they get out in under 3 minutes.
Mod_jk is providing the load balancing via weighted round-robin. Both workers are equally weighted. In theory, each worker should have the same load distribution. (Truly in theory, each worker gets an equivalent number of sessions. Right? Our apps have sessions with a disproportionate number of transactions per session.) We see worker1 averaging higher CPU usage. I can justify this behavior because of session affinity being enabled -- and the huge difference in transactions per session in the applications. Two questions. (1) Am I missing something or are our assumptions wrong? (2) How could we distribute load to the tomcat workers via tomcat CPU usage instead of round-robin? Is there is a different Apache module that can load balance via tomcat CPU usage? Any help, guidance, direction you can provide would be greatly appreciated. Scott
