Managing Serviceguard Extension for SAP on Linux (IA64 Integrity and x86_64), April 2009

rpm -q serviceguard
rpm -q nfs-toolkit
rpm -q sgesap-toolkit
rpm -q sgcmon
rpm -q pidentd
If the rpm packages are missing install them with the following command
rpm -Uhv serviceguard-<version>.product.<linux-os>.<platform>.rpm
rpm -Uhv nfs-toolkit-<version>.product.<linux-os>.<platform>.rpm
rpm -Uhv sgesap-toolkit-<version>.product.<linux-os>.<platform>.rpm
rpm -Uhv sgcmon-<version>.product.<linux-os>.<platform>.rpm
rpm -Uhv pidentd-<version>.product.<linux-os>.<platform>.rpm
Installation Step: IS280
Create all central instance directories below /export as specified in Chapter 2.
For example:
su - <sid>adm
mkdir -p /export/sapmnt/<SID>
mkdir -p /export/usr/sap/trans
MaxDB Database Step: MD290
Create all MaxDB directories below /export as specified in Chapter 2.
For example:
su - sqd<sid>
mkdir -p /export/sapdb/programs
mkdir -p /export/sapdb/data
mkdir -p /export/var/spool/sql/ini
Installation Step: IS300
Each package needs at least one relocatable IP address to be specified. In case there is a distinction between
front-end, server, and LANs. It is likely that at least two addresses are needed per package.
Dedicated heartbeat LANs would require additional addresses. Add all relocatable IP address information
to/etc/hosts. Copy /etc/hosts to the other cluster nodes.
Installation Step: IS380
In /etc/fstab - comment out any references to file systems that are identified as EXCLUSIVE in section
"Planning the LVM layout for Clustered SAP Environments".
Installation Step: IS390
Make sure that the content of /etc/fstab is not recreated during system reboot.
The file /etc/rc.d/boot.vginit is executed during reboot by default. It should not perform any action
for shared volume groups. If this file exists remove it from /etc/rc.d and save it in another backup directory
Installation Step: IS391
By default Linux performs an LVM vgscan whenever the system is rebooted. LVM vgscan attempts to detect
and allocated and non-allocated volumes and create /dev/xxx entries for volume groups it finds.
This could "cause problems" for volumes used in a Serviceguard environment where a Serviceguard package
has been stopped in a cluster. In this case the stopped volume would not be allocated by any cluster node.
And therefore would be claimed by the booting node. When that Serviceguard package tries to start it will
fail because the storage volume is already in use.
For SUSE - the vgscan can be prevented by removing /etc/rc.d/boot.lvm and saving it in another
backup directory.
For Redhat - use the following procedure to prevent a vgscan command at boot time:
Linux Configuration 71