Enterprise Manager - Cluster Manager
Manage ColdFusion clusters and configure load balancing across instances
Overview
The Cluster Manager in ColdFusion Enterprise Manager provides centralized control over ColdFusion clusters, enabling you to group multiple server instances for load balancing, high availability, and simplified management. Clusters distribute incoming requests across member instances to improve performance, reliability, and scalability.
This page is available in ColdFusion Enterprise Edition (and in Trial and Developer editions) with the Server configuration. ColdFusion Standard does not support multiple instances or clusters.
Cluster Configuration
Essential settings for creating and configuring ColdFusion clusters.
Cluster Name
production-web, api-cluster)prod-web-cluster, dev-api-cluster).Multicast Port
Sticky Sessions
- With Sticky Sessions: Better performance, but users lose session if their instance fails
- Without Sticky Sessions: Full failover capability, but higher overhead from session replication
Adding Cluster Members
Configure server instances to join the cluster.
Instance Selection
- Add at least 2 instances for high availability
- Use 3+ instances for production clusters to handle member failures
- Ensure all instances have identical application code and configurations
- Test cluster failover before deploying to production
Load Balancing Configuration
Cluster Manager does not offer a choice of load-balancing algorithm. Load balancing is done by the web server connector: run the Web Server Configuration Tool (wsconfig) and point it at the cluster instead of a single server (the -cluster option on the command line). The connector alternates requests among the instances and switches requests to another instance when one stops running.
Relative weighting is set per instance, not in Cluster Manager: each instance has a load balancing factor that you edit in Instance Manager (or set when registering a remote instance). It applies only when the instance is in a cluster. Adobe's example: an instance with factor 2 receives twice as many requests as one with factor 1.
Session Management
Session Replication
- Keep session data small and simple (avoid storing large objects)
- Use external session storage (Redis, database) for critical session data
- Enable sticky sessions to reduce replication frequency
- Monitor network bandwidth usage during peak traffic
Health Monitoring
Cluster Manager has no health check or auto-remove settings. Members find each other through Tomcat cluster membership on the multicast port. For a remote member, the Membership element in itsserver.xml cluster block sets this up; Adobe's sample block uses frequency="500"and dropTime="3000". The web server connector switches requests to another instance when one stops running.
Common Issues & Solutions
Cluster Members Not Communicating
- Verify firewall allows multicast traffic on configured port
- Ensure all cluster members are on same network segment
- Check that multicast port is not in use by other applications
- Verify network switches support multicast (IGMP snooping configured correctly)
- Review ColdFusion logs for cluster communication errors
Unbalanced Load Distribution
- Verify the connector was configured with wsconfig against the cluster, not a single instance
- Ensure sticky sessions are configured correctly (check cookie/URL parameters)
- Review web server connector configuration for proper worker definitions
- Check that all cluster members are running
- Check each instance's load balancing factor in Instance Manager
Session Loss After Instance Failure
- Verify session replication is enabled and functioning
- Check that multicast communication is working between members
- Enable Use J2EE session variables (Server Settings > Memory Variables) on every member instance; sticky sessions and session replication depend on it
- Review session scope usage (Adobe notes that replication does not support arrays in session-scope CFCs or variables)
- Consider external session storage (Redis, database) for critical applications
- Test failover scenario in non-production environment