Managing Serviceguard Extension for SAP on Linux (IA64 Integrity and x86_64), February 2008

Step-by-Step Cluster Conversion
Cluster Configuration
Chapter 3 137
NOTE The manual startup using common startsap/stopsap scripts may not
work correctly with Netweaver 2004s based Web Application Servers. In
this case the instances need to started by directly using the sapstart
binary passing the start profile as a parameter.
For example:
sapstart pf=/sapmnt/<SID>/profile/START_\
<INAME><INR>_<RELOC_IP/hostname>
The service monitor will not be able to detect a stopsap command and
will consider a manual instance shutdown to be a failure. The package
will then fail over the instance to another node. The existence of the
debug file can prevent that. If the debug file gets removed, the service
monitor continues to run. Make sure that the packaged SAP instance is
already running when removing the debug file or an immediate failover
will occur. As long as the debug file exists locally, a package shutdown
would not try to stop the SAP instance or its database. A local restart of
the package would only take care of the Serviceguard NFS portion. The
instance would not be started with the package and the monitor would be
started in paused mode.
First, the monitoring scripts need to be copied from the staging directory
${SAPSTAGE} to the current SGeSAP/LX package ${SGCONF} directory.
cp ${SAPSTAGE}/SID/sapms.mon ${SGCONF}/<SID>/sapms.mon
Example entries in <pkg_name>.control.script:
SERVICE_NAME[0]=”ciC11ms”
SERVICE_CMD[0]=”${SGCONF}/C11/sapms.mon”
OS460 Optional Step:
It is recommended to set AUTOSTART_CMCLD=1 in
${SGCONF}/cmcluster.rc. This variable controls the automatic cluster
start. If a cluster already exists, the node attempts to join it. If no cluster
is running, the node attempts to form a cluster consisting of all
configured nodes. Distribute the file to all cluster nodes.