Accessibility

Table of Contents

High availability clustering ColdFusion: Using ColdFusion Enterprise Manager

Clustering and ColdFusion

From a ColdFusion perspective, there have been two iterations of software clustering. The first version of clustering for ColdFusion was called ClusterCATS. This was a proprietary software level clustering and it embraced the concept of a central management piece, which controlled and monitored the cluster. This basically enabled fail-over at the web server level if a cluster member failed. ClusterCATS was a full software based clustering paradigm in its own right with no need for any additional software or hardware device.

The current version of software level clustering in ColdFusion via Enterprise Manager is based on J2EE clustering and is not a full clustering system; it is peer-to-peer at the ColdFusion server-instance level. In its basic form there is no need for a management role, however you will find the use of a management role will make building out clusters at the server-instance level easier and more intuitive. This is an opinion I formed after creating many ColdFusion clusters and it is also based on how clusters are created in JRun. The admin instance in JRun is used purely to create instances and to cluster them, it is not a member of any cluster. The key point is that the current clustering paradigm, which is based on the J2EE-JavaEE specifications is not a full clustering paradigm and will still require the use of a software or hardware clustering device to enable failover at the web server level. This implies a need for an external web server in production environments. JRun web server (JWS) is a built-in web server that comes with ColdFusion Enterprise; however JWS is not recommended for production use. Figure 4 illustrates such a configuration, which uses round robin with sticky sessions. In order to cluster the ColdFusion servers there must be an external clustering mechanism either hardware or software.

Round robin load balancing with an external
clustering mechanism.

Figure 4. Round robin load balancing with an external clustering mechanism.