Call for Oracle support & training (800) 766-1884
Free Oracle Tips

Corporate Oracle Training
Custom Oracle Training
Oracle New Features Training
Advanced Oracle DBA Classes
Oracle Tuning Courses
Oracle Tips & Tricks
Oracle Training Links
Oracle Training Links
Oracle Training Links

We are top for USA Oracle Training Clients


Free Oracle Tips


Free Oracle App Server Tips


Oracle support

Oracle training

Oracle tuning


Remote Oracle

Custom Oracle Training

Donald K. Burleson

Oracle RAC Tips

Veritas I/O Fencing and SCSI-3 Persistency

Veritas accomplishes the important task of managing access to storage with SCSI-PR. The I/O fencing method is used to avoid the split-brain condition in the cluster.

I/O fencing (or disk fencing) is a method to guarantee the data integrity in the face of a split-brain condition. When the mechanism used to detect the failure of a node breaks down, the cluster splits into cluster partitions. Each node tries to perform the corrective action. The method of I/O fencing is employed to remove the risk associated with the split-brain condition. I/O fencing blocks access to storage from the specific nodes.

DBEAC uses the SCSI-3 PR method to do fencing. SCSI-3 PR supports multiple nodes accessing a device, while simultaneously blocking access to other nodes. PR reservations are persistent across SCSI bus resets and also support multiple paths from a host to a disk. SCSI-3 PR uses the concept of registration and reservation. Each participating system or host registers its own key with the disk. Registered systems establish reservations, a method which insures that only the registered systems can write.

SCSI-3 PR provides a good method of fencing. Removing the registration from a device disables the write access for the host. One of the registered systems can eject another. A system wishing to eject another system issues 'pre-empt' and 'abort' commands. Once a node is ejected from membership it cannot eject others. The Veritas I/O fencing mechanism uses this principle.

I/O fencing allows write access to systems active in the cluster and blocks access for ejected systems. VxVM implements the fencing and all disks added to VM control are fenced automatically. Three small disks or LUNS are specified as ‘coordinator disks,’ which are used for the purpose of fencing. These disks act as a global lock during the cluster reconfiguration. Under the split-brain scenario, the nodes race for control of the coordination disks. The node that gains control begins fencing the other node by ejecting and removing the registration key.

Veritas provides a kernel-based fencing module to do the I/O fencing. This module gets the cluster reconfiguration information from the GAB module and begins fencing as required. The vxfen driver, which is installed on each system, uses coordinator disks to resolve potential split-brain conditions.

For more information, see the book Oracle 11g Grid and Real Application Clusters 30% off if you buy it directly from Rampant TechPress . 

Written by top Oracle experts, this RAC book has a complete online code depot with ready to use RAC scripts.





Oracle performance tuning book



Oracle performance tuning software

Oracle performance tuning software
Oracle performance Tuning 10g reference poster
Oracle training in Linux commands
Oracle training Excel
email BC:

Copyright © 1996 -  2014 by Burleson Inc. All rights reserved.

Oracle® is the registered trademark of Oracle Corporation.