As shown in Figure 1, an MPE or MRA cluster can contain an additional georedundant server, called a spare server. The active server will replicate its database to the standby server as well as the spare server. In this configuration, the standby server is first in line to take over from the active server and the spare is second in line.
Clusters with Active, Standby, and Spare Servers
Active, standby, and spare servers interoperate as follows:
- The servers communicate using WAN TCP streams to perform replication, monitor heartbeats, and merge events.
- The active and standby servers share a common virtual IP (VIP) cluster address to support automatic failover.
- The spare server has a unique VIP cluster address.
- The HA database runtime process constantly monitors the status of all servers.
- If the standby server does not receive three consecutive heartbeats, it attempts to communicate with the active server. If the standby server receives no response from the active server, it assumes the active role.
- When HA misses the three heartbeats to the spare server, it instructs the spare server to assume the standby role.
The terms active, standby, and spare denote roles, or states, that the servers assume, and these roles can change automatically and at any time based on decisions made by the underlying HA database. If both the active and standby servers become unavailable, the spare server automatically assumes the active role and continues to provide service.