Failover Mechanism in an Active/Passive Cluster
Universal Messaging runs as a service in a cluster. Within an active/passive cluster, there only be a single instance of Universal Messaging server running at any given time. The other Universal Messaging servers are inactive.
In a clustered environment, when a client makes a request to a server, the server handles the request much the same as in an unclustered environment, except that the server writes the client information to a shared disk instead of a private data store.
The following diagram illustrates the flow of documents through a typical clustered environment.
| Steps | Description |
|---|---|
| 1 | Universal Messaging clients use the virtual IP address of the cluster to connect to the active/passive Universal Messaging cluster. |
| 2 | Cluster control software forwards the client request to the active server in the cluster. |
| 3 | The active server reads data from or writes data to the shared storage. |
| 4 | Universal Messaging returns the results to the client application. |
The following diagram illustrates the failover in a clustered environment. If a server fails, subsequent requests for the session are redirected to a spare server in the cluster that is currently active and running.
| Steps | Description |
|---|---|
| 1 | Universal Messaging clients use the virtual IP address of the cluster to connect to the active/passive Universal Messaging cluster. |
| 2 | The active server experiences failure and shuts down. |
| 3 | The cluster software returns the error code to the client. |
| 4 | Cluster control software marks the spare server as active. |
| 5 | Cluster control software forwards the client request to the active server in the cluster. |
| 6 | The active server reads data from or writes data to the shared storage. |
| 7 | Universal Messaging returns the results to the client application. |