Best Practices for SGeRAC and Oracle RAC on HP-UX 11i, March 2009

Note 2:

SLVM timeout would be documented in the LVM Whitepaper entitled LVM

Link and Node Failure Recovery Time found on the HP Technical

Documentation site at docs.hp.com.

When Cluster Interconnect Subnet monitoring is not used to monitor the CSS-HB subnet and if the

default values of any of the following parameters has to be changed:

• Oracle Clusterware parameter CSS MISSCOUNT

• Serviceguard cluster configuration parameter MEMBER_TIMEOUT

Then the CSS MISSCOUNT parameter should be greater than:

• For SLVM: F + SLVM timeout + 15 seconds

• For CVM/CFS: 3 times F + 15 seconds

• When both SLVM and CVM/CFS are used, then take the max of the above two calculations

Storage

SGeRAC supports a variety of storage options for Oracle RAC.

• CFS

• CVM

• SLVM

• ASM over Raw Devices

• ASM over SLVM

HP recommends you organize storage as follows, with each item using separate storage:

1. Oracle Clusterware binaries

2. RAC database binaries

3. Oracle Clusterware registry (OCR) and quorum device (Voting Disk)

4. Database data files

5. Database recovery data (archive, flash recovery)

Keeping Oracle Clusterware home and Oracle database home on separate mount points allows you

flexibility to unmount the database home for maintenance without having to shutdown the Oracle

Clusterware cluster. Placing the registry and quorum device on separate storage ensures that

deactivation of storage for RAC database does not affect Oracle Clusterware. Finally, each RAC

database uses two separate storage areas for database data files and recovery data. This is to

ensure recovery data is available in case there are problems with storage for the database data files.

Oracle Clusterware may need timely access to its home directory and binaries. The default I/O

timeout for LVM logical volumes is forever; therefore, for file systems on top of LVM, it may be

necessary to configure a logical volume timeout, for example 30 seconds per link, so that Oracle

Clusterware gets get an I/O error when the LVM physical volume is inaccessible. There are similar

considerations for RAC binaries. For VxVM, there is already a default I/O timeout of 30 seconds per

link. There are similar considerations for RAC binaries as well as Oracle Clusterware binaries with

regards to timely access.