ZK Developer's Reference/Clustering/Server Configuration

The configuration of a Web server really depends on the server itself. There is no standard approach.

Load Balancer (Required)

Be sure to configure sticky session on your load balancer because ZK only works correctly on this setting. <ref> Check ZK DesktopImpl.java, you will see there are lots of synchronized used in many methods including:

• generate desktop ID
• addPage(), removePage()
• enableServerPush()
• processing server push update

The keyword synchronized only works in a single JVM, so it’s impossible to keep the same state between 2 different nodes with session replication if 2 nodes receive requests at roughly the same time.

Each AU request changes a Desktop's status of a session, if 2 AU requests are processed by 2 different clustering nodes and each one has its own session, then one session state might override another session.

Apache + Tomcat

For configuring Apache + Tomcat, please refer to

• How to Run ZK on Apache + Tomcat clustering, Part I
• How to Run ZK on Apache + Tomcat clustering, Part II

More detail settings

• Tomcat Cluster

Google App Engine

For configuring Google App Engine, please refer to

• ZK Installation Guide: Google App Engine

Apache + JBoss

For configuring JBoss, please refer to

• JBoss Cluster

Glassfish

For configuring Glassfish, please refer to

• Glassfish Cluster

WebLogic

For configuring WebLogic, please refer to

• WebLogic Cluster

Version History

Version	Date	Content