Oracle Clusterware 10g will always handle node fencing (aka eviction) by rebooting one or more nodes.  Oracle didn’t have its own cluster manager software for most platforms on 9i.  So, if your customer was on a platform where Oracle didn’t have a cluster manager of its own (it only had a Linux and Windows), then building the RAC cluster would have required some 3rd party software.  Some of the 3rd party cluster managers used different schemes to fence nodes from the cluster.  Many of them used I/O fencing which is less disruptive in that it didn’t require a node reboot, but required more proprietary interfaces to be used to access and manage storage. I/O Fencing, or simply fencing,   gives write access to members of the active cluster and blocks access to non-members.  The physical components of I/O fencing in Storage Foundation for Oracle RAC are coordinator disks and data disks.  Each component has a unique purpose and uses different physical disk devices. Data disks are standard disk devices for data storage and are either physical disks or RAID Logical Units (LUNs).  These disks must support scsi-3 PR and are disks or RAID Logical Units (LUNs).  These disks must support SCSI-3 PR and are part of standard VxVM/CVM disk groups. CVM is responsible for fencing data disks on a disk group basis.  VxVM enables I/O fencing and provides additional  features.  Disk added to a disk group are 14 VERITAS Storage Foundation for Oracle RAC Installation and Configuration Guide. Coordinator disks are three (or an odd number greater than three) standard disks or LUNs set aside for I/O fencing during cluster reconfiguration.  These disks provide a lock mechanism to determine which nodes get to fence off data drives from other nodes.  A node must eject a peer from the coordinator disks before it can fence the peer from the data drives.  This concept of racing for control of the coordinator disk to gain the ability to fence data disks is the key to understanding prevention of split brain through fencing. Coordinator disks do not serve any other purpose in the Storage Foundation for Oracle RAC configuration.  Users cannot store data on these disks or include the disks in a disk group for user data.  The coordinator disks can be any three disks that support SCSI-3 PR.  VERITAS recommends using the smallest possible LUNs for coordinator disks.  Since coordinator disks do not store any data, cluster nodes need only register with them and do not need to reserve them. I/O fencing provided by the kernel-based fencing module (VXFEN), performs identically on node failures and communications failures.  When the fencing module on a node is informed of a change in cluster membership by GAB module, it immediately begins the fencing operation.  The node attempts to eject the key for departed node(s) from the coordinator disks using the preempt and abort command.  When the node successfully ejects the departed disks, in a split –brain scenario, both sides of the split would race for control of the coordinator disks.  The side winning the majority of the coordinator disk wins the race and fences the loser.  The loser then panics and reboots the system.