Setting up Server Cluster

Automatic failover, a native Upland AccuRoute capability, ensures seamless and immediate recovery from a localized system failure on the server. It requires two redundant servers that are joined to form a server cluster, and a third system where critical server resources are stored.

The redundant servers are referred to as:

  • Active Server - the current server
  • Passive Server - the standby server

The third system that stores critical resources is called the “database server” because it hosts the server databases along with server configuration files. If the server is enabled for faxing, the database server also hosts the Telco share which is an intermediary repository for inbound and outbound faxes. (The database server can also function as a remote server such as Remote Administrator, Remote Composer, Remote Embedded Directive Manager, or Remote Modem Server.)

In some environments, where multi-function devices are used, a fourth system - Remote IIS - can be added to host the Intelligent Device client and specific device client installs.

Expand to view more.

Failover is driven by the Cluster Manager service which is installed on the server. (This service does not appear in the Services applet because it is not registered by default.) When an server cluster is created, this service is immediately registered and runs simultaneously on both servers in the cluster. During regular operation, the active server in the server cluster continuously writes to the server databases while Cluster Manager service on the passive server polls the database server in 45-second intervals to verify that the active server remains active.

When a localized system failure occurs on the active server, it loses communication with the database server and cannot write to the server databases. As soon as the Cluster Manager service on the passive server polls the database server again and finds no recent activity from the active server, the Cluster Manager service on the passive server starts all its local services and becomes the active server in the cluster. The transition is complete within 1 to 3 minutes of the failure.

Tip: The Server Monitor is a monitoring application that detects failures on servers. Upland AccuRoute recommends that the Server Monitor or another monitoring application be used to alert administrators that a failure has occurred so that failures can be corrected in a timely manner.

A server in the cluster can be designated as the preferred server. If a server is preferred, it is always the active server whenever it is available. Therefore, if a system failure occurs and the preferred server goes offline, the non-preferred server becomes active, but when the preferred server is online again, it becomes the active server and the non-preferred server returns to a standby state.