HP Enterprise Cluster Master Toolkit User Guide (5900-2131, December 2011)

Serviceguard support for ASM on HP-UX 11i v3 onwards
This document describes how to configure Oracle ASM database with Serviceguard for high
availability using the ECMT Oracle toolkit. Look at the ECMT support matrix available at http://
www.hp.com/go/hpux-serviceguard-docs -> HP Enterprise Cluster Master Toolkit for the supported
versions of ECMT, Oracle, and Serviceguard.
Note that for a database failover, each database should store its data in its own disk group. Refer
to Oracle documentation for limitation imposed by ASM on the number of disk groups. The disk
group members must be logical volumes managed by HP Logical Volume Manager (LVM). Initial
release of HP-UX 11i v3 did not support LVM version 2. With HP-UX 11i v3 0803, LVM version
2 logical volumes are also supported as members of the ASM disk groups.
A new I/O infrastructure that enables the native built-in multipathing functionality is introduced in
HPUX 11i v3. This feature offers users a continuous I/O access to a LUN or disk if any of the paths
fails. This feature is enabled in the operating system by default. In addition, new DSF (device
special file) format is introduced in this operating system known as persistent DSF. An example of
persistent DSF is /dev/disk/disk1, compared to the legacy DSF, /dev/rdsk/cxtydz.
A new cluster wide DSF format is introduced in HP-UX 11i v3 known as cDSF. Serviceguard
A.11.20 requires the Serviceguard patch PHSS_41225 to support this feature. An example of
cDSF is /dev/cdisk/disk1, compared to the persistent DSF, /dev/disk/disk1. ASM cannot
detect cDSF format and hence this format cannot be used to create the ASM disk groups.
It is the responsibility of Serviceguard to provide the necessary I/O fencing when failover is provided
for a single-instance Oracle database instance. Serviceguard ensures I/O fencing for a failover
package via a feature in the volume manager - exclusive activation of volume groups. Hence it is
required that the ASM disk group members be LVM logical volumes.
Framework for ASM support with Serviceguard
The framework for ASM integration with Serviceguard makes use of a Multi-Node Package (MNP)
to encapsulate the per-node ASM instances, with one or more Oracle single-instance failover
packages dependent on this MNP. This configuration enables the database instance to start up in
the right order in relation to the ASM instance, and in the event of failover, to relocate to a node
where an ASM instance is available. The benefits of this package framework for single-instance
Oracle database using ASM-based storage are:
• Avoid failing over ASM instance. This reduces the failover time due to ASM instance restart.
• Many database instances can share one ASM instance per node.
• Sequencing between Oracle database instance and ASM instance during
startup/shutdown/failover.
• Flexibility of the database instances to fail over independently.
HP provides a set of scripts known as the Enterprise Cluster Master Toolkit (ECMT) scripts which
provides for the integration of ASM with Serviceguard. The operation of the Toolkit is described
below.
The start function switches to the ASM software owner user id by executing the su command. It
starts the ASM instance specified by the user using the sqlplus commands. The mounting of the
ASM disk groups associated with a database instance will be done before database instance
startup after volume group activation during the database package startup.
The stop function executes su to the ASM software owner user id. It stops the ASM instance using
sqlplus commands.
The monitor function contains a continuous loop to check if the ASM instance processes specified
are healthy. If the monitor function finds any process to have died, it means that the ASM instance
has either failed or been inappropriately shut down, that is, without using cmhaltpkg. The service
that invokes the function fails at this point and the Serviceguard package manager fails the
corresponding ASM MNP instance.
32 Using the Oracle Toolkit in an HP Serviceguard Cluster