There sould be a quorum disk or quorum partition in the
shared storage, this disk should be used in a way so
the nodes could see the cluster status even if the
cluster interconnect fails. This way split brain situations
can be avoided. A quorum to form the cluster should
exists, half nodes + 1. The quorum disk is claimed by
the nodes that can communicate, and nodes that
cannot, sould know that the quorum disk is in use and
avoid join the cluster. Like the RH Cluster
Manager/TruCluster quorum disk.
Log in to post a comment.