In cluster systems, two or more servers are operated as a single virtual server in order to enable high availability.
If a system is run with only one server, and the server or an application operating on it fails, all operations would stop until the server is rebooted.
In a cluster system where two or more servers are linked together, if one of the servers becomes unusable due to trouble with the server or an application being run, by restarting the applications on the other server it is possible to resume operations, shortening the length of time operations are stopped.
Switching from a failed server to another, operational server in this kind of situation is called failover.
In cluster systems, groups of two or more servers are called clusters, and the servers comprising a cluster are called nodes.
Clusters are classified into the following types:
Standby clusters
This type of cluster involves standby nodes that stand ready to take over from operating nodes. The mode can be one of the following modes:
1:1 hot standby
A cluster consisting of one operating node and one standby node. The operating node is operational and the standby node stands ready to take over if needed.
n:1 hot standby
A cluster consisting of n operating nodes and one standby node. The n operating nodes run different operations and the standby node stands ready to take over from all of the operating nodes.
n:i hot standby
A cluster consisting of n operating nodes and i standby nodes. The style is similar to n:1 hot standby, only there are i standby nodes standing ready to take over from all of the operating nodes.
Mutual standby
A cluster consisting of two nodes with both operating and standby applications. The two nodes each run different operations and stand ready to take over from each other. If one node fails, the other node runs both of the operations.
Cascade
A cluster consisting of three or more nodes. One of the nodes is the operating node and the others are the standby nodes.
Scalable clusters
This is a cluster that allows multiple server machines to operate concurrently for performance improvement and reduced degrading during trouble. It differs from standby clusters as the nodes are not divided into operating and standby types. If the system fails on one of the nodes in the cluster, the other servers take over the operations.
Resource Orchestrator managers support failover clustering of Microsoft(R) Windows Server(R) 2008 Enterprise (x86, x64) and 1:1 hot standby of PRIMECLUSTER.
When operating managers in cluster systems, the HBA address rename setup service can be started on the standby node.
Using this function enables starting of managed servers without preparing a dedicated server for the HBA address rename setup service, even when managers and managed servers cannot communicate due to problems with the manager or the failure of the NIC used for connection to the admin LAN.
Information
For details of failover clustering, refer to the Microsoft web site.
For PRIMECLUSTER, refer to the PRIMECLUSTER manual.
