Distributing Forum Traffic on Clustered Servers 

Use

In a clustered community, several application servers work together to distribute heavy traffic evenly. This increases scalability and provides fault tolerance if one or more servers fail. Clusters configure themselves, have no single point of failure, and provide near-linear scalability when additional servers are added. New forum content is automatically distributed to all servers in the cluster. Servers can join and leave the cluster at will. Cluster node failures are handled transparently.

An optimal clustering configuration depends on network traffic, existing infrastructure, and budget. An example clustering configuration is depicted below:

Load Balancer
A load balancer sits between the users and the application servers. Every user is issued a unique session cookie that maintains their current state as they use the forums. A session is created on a single application server, and is optionally distributed to the other servers for fault tolerance. If the load balancer is 'session-aware', it inspects the cookie value and always sends the user to the same application server for each request. This is the most efficient option. Otherwise, the load balancer sends the requests round-robin style to a random application server, requiring that all session data is constantly replicated between servers. Consult your application server documentation for recommended session replication and failover services.

Application Servers
In the middle tier, cluster-enabled instances of Jive Forums are installed on multiple application servers. Caches between the instances are synchronized automatically. Any application server failure is detected by the load balancer and removed from the cluster.

Database
All clustered instances of Jive Forums share the same database. This could be either a single database server or a cluster of database servers.

Clustering FAQ

●      Must clustered servers reside in the same local network?

Yes. The clustering module uses IP multicast to detect the cluster, which requires all servers to reside in the same local network.

●      Does searching work in a cluster?

Yes, searching works without making any special configuration changes. Each cluster server maintains its own search index and updates it with the latest data (when auto-indexing is enabled).

●      How do gateways work with clustering?

Gateway support with the clustering module is limited in the current release of Forums. Gateways must be identically configured with exporting enabled on every server in the cluster. If the import server fails, importing is delayed until the server is brought back online.

Procedure

Enabling or Disabling Clustering

...

       1.      From the Admin Console, choose the System tab.

       2.      In the left-hand navigation pane, choose Cache Settings.

       3.      Scroll down to Clustering under Cache Features.

       4.      Choose either Enabled or Disabled.

       5.      Save the settings.

You should also set the system time on each cluster user to match other cluster user settings as closely as possible. If you use the short-term query cache, it is important that the system time across cluster users is matched closely (within one second) – otherwise, the cache does not work correctly.