users@glassfish.java.net

Re: AJP/mod_jk load balancing in a cluster - sticky sessions not so sticky

From: <glassfish_at_javadesktop.org>
Date: Wed, 07 May 2008 13:50:03 PDT

Thanks for pointing this out!

I'm highlighting this information in my blog "How to Loadbalance GlassFish Cluster with Apache Loadbalancer" at

  http://blogs.sun.com/jluehe/entry/supporting_apache_loadbalancer_with_glassfish

Under Configuration step 3, I now mention the following:

  Make sure that the name of each worker equals the value of the
  jvmRoute system property of the GlassFish instance to which the worker
  connects. This convention makes it possible for an HTTP session to
  remain sticky to the GlassFish instance on which the session was
  created, or on which the session was last resumed.

and at the bottom of the blog, I've added this information:

  As soon as the cluster instance to which an HTTP session has been
  sticky has failed, the loadbalancer will route any subsequent requests
  for the same HTTP session to a different instance. This instance will
  be able to load and resume the requested session using the in-memory
  session replication feature that has been available since GlassFish
  V2. The in-memory session replication feature is enabled only for
  those web applications that have been marked as distributable in their
  web.xml deployment descriptor, and that have been deployed to the
  cluster with the --availabilityenabled option of the asadmin deploy
  command set to true (default is false).

I will also add this information to the GlassFish FAQ.

Jan
[Message sent by forum member 'jluehe' (jluehe)]

http://forums.java.net/jive/thread.jspa?messageID=273215