Architecture considerations and best practices for architecting an Oracle RAC solution with Serviceguard and SGeRAC

ManualsBrandsHP ManualsSoftwareHP Serviceguard Extension for RAC (SGeRAC)

Storage management choices

Beginning with Serviceguard A.11.17, Serviceguard/SGeRAC supports the following storage

management options for Oracle RAC:

• Serviceguard Cluster File System (CFS)

• Shared Logical Volume Manager (SLVM)

• VERITAS Cluster Volume Manager (CVM)

• ASM with SLVM

• ASM with raw disks (on HP-UX 11i v3 or later only)

Both SLVM and CVM have been available for quite some time. The Serviceguard CFS storage option

and ASM with SLVM option are available starting with Serviceguard A.11.17.

Table 1 shows the storage management choices for various data used in an RAC cluster. The Oracle

binaries are for both Oracle Clusterware software and RAC software.

Table 1. Storage management options for RAC cluster

Storage option Oracle binaries Oracle Cluster

Files

Oracle data

files

Archive log

files

Shared raw LVs–SLVM or CVM (requires

SGeRAC)

No Yes Yes No

Shared raw disk devices as presented to

host (no SGeRAC)

No Yes Yes No

Oracle ASM Yes* Yes* Yes Yes

Cluster File System (requires SGeRAC) Yes Yes Yes Yes

Local File System (on each node) Yes No No Yes

* Storage for Oracle Binaries and Oracle Cluster Files available on ASM starting with Oracle RAC 11gR2

As shown in Table 1, CFS is the only storage option that can be used for all kinds of data used in an

RAC cluster. CFS also can be used for non-RAC applications running in the same cluster.

The following section provides further details on each of the storage management options.

Serviceguard CFS

Prior to SGeRAC A.11.17, when deploying RAC on HP-UX, the user had no choice but to use shared

raw volumes using SLVM or CVM for RAC database data. Beginning with Serviceguard/SGeRAC

11.17, Serviceguard SMS CFS is a supported storage option for RAC database data.

While Oracle Clusterware OCR and Voting Disk may be placed on a raw device, Serviceguard CFS

is preferred for better manageability and reliability. If a raw device is used, it must be accessible by

all nodes in the cluster simultaneously. This can be accomplished by directly accessing a physical disk

connected to all systems; however, this device has no protection from failure, unless mirrored at the

storage array or volume manager level. Devices under the CFS have inherited mirroring and

multi-pathing features from CVM to increase device resiliency should any individual disk fail or if the

need arises to relocate or resize either component. Additionally, CFS greatly simplifies management

of file growth by directly reporting currently-allocated space in use for each file, while also reducing

the number of devices to manage, since the OCR file and Voting Disk may both reside in the same

CFS mount.

Oracle Cluster files = Voting Disk and Oracle Cluster Repository (OCR)

Only if deployed on a separate VG/DG which is failed over.