In a OES Cluster Services cluster, the master node monitors the health of the cluster nodes. It also synchronizes updates about the cluster to eDirectory. The first server that comes up in a cluster is automatically assigned the cluster IP address and becomes the master node. If the master node fails, Cluster Services migrates the cluster IP address to another server in the cluster, and that server becomes the master node.
If a slave node fails to receive heartbeat packages from a master node within a predefined tolerance, the slave assumes that the master has left the cluster or is experiencing problems. The slave node, together with other slave nodes, tries to elect a new master node. The goal is to identify a node with a higher IP address that can be seen by the most member nodes. Each slave node follows the same process to identify a master candidate. If multiple masters are elected, the SBD (split-brain detector) guarantees that only one master survives and fences the other masters and their members.
OES Cluster Services for Open Enterprise Server (OES) 11 SP1 introduces some intelligence in the master election process when the master leaves a cluster (voluntarily or involuntarily). The same master is elected under both the old and new algorithms, but the conclusion is reached sooner, especially for larger clusters. The new algorithm substantially reduces the time needed for master election in some cases. The algorithm is available for OES 11 SP1 and later as well as in the September 2012 Scheduled Maintenance for OES 11 and OES 2 SP3.
This section briefly describes the old and new master-election algorithms.