You must configure a fencing device for each node in the cluster. When configuring a fencing device, you should ensure that your fencing device does not share power with the node that it controls. For information on fence device configuration, see Section 4.6, “Configuring Fence Devices”. For information on configuring fencing for cluster nodes, see Section 4.7, “Configuring Fencing for Cluster Members”.

After configuring a fence device for a node, it is important to test the fence device, to ensure that the cluster will cut off access to a resource when the cluster loses communication with that node. How you break communication with the node will depend on your system setup and the type of fencing you have configured. You may need to physically disconnect network cables, or force a kernel panic on the node. You can then check whether the node has been fenced as expected.

When creating a two-node cluster, you may need to configure a tie-breaking mechanism for the cluster to avoid split brains and fence races for the cluster, which can occur when the cluster interconnect experiences issues that prevent the nodes from communicating. For information on avoiding fence races, see the Red Hat Knowledgebase solution "What are my options for avoiding fence races in RHEL 5, 6, and 7 High Availability clusters with an even number of nodes?" on Red Hat Customer Portal at https://access.redhat.com/solutions/91653. For information on avoiding fencing loops, see the Red Hat Knowledgebase solution "How can I avoid fencing loops with 2 node clusters and Red Hat High Availability clusters?" on Red Hat Customer Portal at https://access.redhat.com/solutions/272913.