HP Serviceguard Storage Management Suite A.04.00 Release Notes

SCSI-3 Persistent Reservations Support Not Required on Storage Devices
The Veritas Cluster Server (VCS) relies on support of SCSI-3 Persistent Reservations on storage
devices - this is part of the Veritas split brain prevention mechanism. HP Serviceguard uses a
different and more robust method to prevent split brain - storage devices do not need to support
SCSI-3 Persistent Reservations in an HP Serviceguard environment.
Oracle Disk Manager (ODM) with VxFS and LVM
Oracle Disk Manager (ODM) for VxFS file systems on LVM volumes is not supported. Oracle Files
that need ODM capability must reside on a VxFS file system created on a VxVM or CVM volume.
Oracle Disk Manager (ODM) Limitation for CFS Oracle Bundles
Single instance Oracle databases are not permitted outside of Serviceguard packages in
environments using HP Serviceguard Storage Management Suite T2774EB, T2777EB, T8684EB,
T8687EB, T8694EB, and T8697EB bundles. If there are any Oracle instances not controlled by
Serviceguard packages, a halt of the system multi-node package SG-CFS-pkg could bring down
the node or cluster. Place all Oracle single instances in Serviceguard packages when deploying
these bundles. In addition, single instance Oracle packages using VxFS and VxVM, or CVM/CFS,
must have a direct dependency on the system multi-node package SG-CFS-pkg.
ODM Must be Used for Oracle RAC Data on CFS File Systems
ODM - Oracle Disk Manager is included in all SG SMS bundles. ODM must be used for Oracle
RAC data on CFS file systems. ODM can also be used for Oracle standalone databases with data
on VxVM volumes with VxFS file systems, or with data on CVM/CFS. ODM is not supported for
use on file systems on LVM logical volumes. ODM improves the I/O performance for Oracle
databases with data on VxFS and CFS file systems. The performance approaches that of Oracle
data on raw disks.
Dirty Region Logging (DRL) for Mirrored Volumes
If a system failure occurs while mirrored volumes are in use, the mirrors will recover and clean up
any writes that were in progress when the system failed. The Dirty Region Logging (DRL) feature
tracks writes that are in progress to reduce the recovery effort after a system failure. Without DRL,
the entire volume must be recovered, because there is no way to determine which blocks had writes
in progress at the time of the system failure.
The recovery operation is I/O intensive and can be time consuming, depending on the number of
volumes and how large they are. In a clustered environment the IO performance of multiple nodes
using the same volume will be impacted during recovery.
HP strongly recommends that you use DRL with mirrored data volumes in a clustered environment
to enable faster volume recovery. However, Oracle database volumes should be configured without
DRL, because the Oracle database application keeps its own logs and does not require VxVM to
maintain logs. For information on DRL and the Veritas™ SmartSync Recovery Accelerator, see the
Veritas™ Volume Manager Administrator’s Guide, HP-UX 11i v3, 5.0.1.
HP-UX Operating Environment Upgrade Limitations
When you purchase an HP Serviceguard Storage Management Suite (SG SMS) Operating
Environment (OE) bundle, the SG SMS software is provided on one DVD and the HP-UX OE (Data
Center - DCOE, or High Availability - HAOE) is provided on a separate DVD.
The software provided on the SG SMS DVD relies on products and filesets that are provided on
the OE DVD, including HP Serviceguard and the Veritas Storage Foundation products. Each SG
SMS OE version relies on specific versions of HP Serviceguard and Enterprise Cluster Master Toolkit
(ECMT) that can change with new HP-UX OE releases.
Product Issues and Limitations 43