HP-UX Configuration

Cluster Node Synchronization

Repeat the steps in this section for each node of the cluster that is different than the primary.

Look at the groupfile file, /etc/group, on the primary side.

If any of the groups listed in Table 2-4 “Groupfile File Groups” exist on the primary node and they do not exist on the backup node, copy them from the primary node to the backup node. If any group exists, verify that it has the same GID on both the primary and backup nodes. Merge the group members lists.

Look at the password file, /etc/passwd, on the primary side.

If any of the users listed in Table 2-5 “Password File Users” exist on the primary node, recreate them on the backup node. Assign the users on the backup nodes the same user and group ID as the primary nodes.

INFORMIX users must have the same passwords, as well, on both the backup and primary nodes. Beware of copying over into /etc/passwd if your HP-UX is running in Trusted System mode.

Look at the service file, /etc/services, on the primary side.

Replicate all services listed in Table 2-6 “Service File Services” that exist on the primary node onto the backup node.

Change the HP-UX kernel on the backup node to meet the SAP R/3 requirements.

Compare the Tunable Parameters section of /stand/system on both hosts. All values on the backup host must reach or exceed the values of the primary host. A tool, kinst, configures the kernel automatically. It is bundled with SAP R/3 releases prior to 3.0F.

Install all HP-UX patches that are recommended for MC/ServiceGuard and patches recommended for SAP R/3.

If you are planning to configure multiple Application Servers that are configured to be started parallel (see IS510) make sure the kernel parameter "tcp_conn_request_max" of your HP-UX systems are configured appropriately to 1024. This will prevent closing tcp ports (for example the LISTENER port configured for Oracle DB).

Build a new kernel with mk_kernel(1m) on the backup host if /stand/system was changed in either Step Title not available or .

If the primary node has the Central Instance installed:

If the primary node has the INFORMIX database installed:

If the primary node has the ORACLE database installed:

Create additional links in /oracle/<SID> on the primary node. For example:

su - ora<sid>
ln .dbenv_<primary>.csh .dbenv_<secondary>.csh 
ln .dbenv_<primary>.sh .dbenv_<secondary>.sh 
exit

	NOTE: If you are implementing an APP server package make sure that you install the Oracle Client libraries on all nodes you run the package on. Refer to OSS Note 180430 for more details.

If you are using ORACLE:

Create a mountpoint for the Oracle files on the backup node if it is not already there. For example:

su - ora<sid>
mkdir -p /oracle/<SID>
exit

	NOTE: Complete this step also, if you are using 4.0A, the two package concept, and the backup node will be configured to run the Central Instance package only.

If you need different configurations of the Central Instance depending on the node it runs on:Be careful if you create different instance profiles for the different machines. You have to make sure by yourself that the Central Instance is capable of doing the intended work with any of the possible configurations.

Perform the following steps to setup different configurations:

If the primary node has the Central Instance installed and the other node has no internal application server installed:

Distribute the local directory tree /usr/sap/<SID>/SYS. Do not use rcp(1), it will follow all links and copy a lot of files from the shared disks that are not needed.

For example:

On the primary node:

cd /usr/sap/<SID>/SYS
find . -depth -print | cpio -o >/tmp/SYS.cpio
use ftp(1) to copy the file over to the secondary node

On the secondary node:

su - <sid>adm
mkdir -p /usr/sap/<SID>/SYS
cd /usr/sap/<SID>/SYS
cpio -id </tmp/SYS.cpio
exit

Import the shared volume groups using the minor numbers specified in Step Title not available.

The whole volume group distribution should be done using the command line interface. Do not use SAM. SAM will scramble the minor numbers. Specify the device minor numbers explicitly by creating the groupfile manually. For example:

mkdir /dev/vgsap<SID>
mknod /dev/vgsap<SID>/group c 64 0x080000

Now you can use vgimport(1m) with the mapfile created on the primary host during Step Title not available or Step Title not available.

Create a mountpoint for the Central Instance directory so the node can run the Central Instance. For example:

mkdir -p /usr/sap/<SID>/DVEBMGS<INSTNR>

Cluster Node Configuration

Repeat the steps in this section for each node of the cluster.

Logon as root.

Check that /etc/logingroup is a link to /etc/group.

Create an .rhosts file in the home directories of the HP-UX users root, <sid>adm and (if applicable) informix. Allow login for root as root from all nodes including the node you are logged into. Allow login for root and <sid>adm as <sid>adm from all nodes including the node you are logged into. Be careful with this step, many problems result from an incorrect setup of remote access.

Check the setup with remsh commands. If you have to provide a password, the .rhosts does not work.

Create all directories below /export as specified in Step Title not available. For example:

su - <sid>adm
mkdir -p /export/sapmnt/<SID> 
mkdir -p /export/usr/sap/trans 
exit
su - informix
mkdir -p /export/informix (Informix only) 
exit

Add all relocatable IP address information to /etc/hosts. Do not forget heartbeat IP addresses. Use Table 2-7 “Relocatable IP Address Information” to record the addresses.

If you use DNS:

Configure /etc/nsswitch.conf to avoid problems. For example:

hosts: files[NOTFOUND=continue UNAVAIL=continue \
TRYAGAIN=continue]dns

If you establish frontend and server LANs to separate network traffic:

Add routing entries to the internet routing configurations of /etc/rc.config.d/netconf. This is the only phase of the whole installation in which you will need to specify addresses of the server LAN. Route all relocatable client LAN addresses to the local server LAN addresses. For example, a two package concept:

ROUTE_DESTINATION[n]=<relocdb>
ROUTE_MASK[n]=""
ROUTE_GATEWAY[n]=<relocdb_s>
ROUTE_COUNT[n]=1
ROUTE_ARGS[n]=""
ROUTE_DESTINATION[n+1]=<relocci>
ROUTE_MASK[n+1]=""
ROUTE_GATEWAY[n+1]=<relocci_s>
ROUTE_COUNT[n+1]=1
ROUTE_ARGS[n+1]=""

If you have SAP R/3 4.0A or higher:

Beginning with SAP R/3 4.0A, during installation SAP R/3 appends some entries to the standard .profile files in the user home directories instead of using a new file defined by SAP R/3. On HP-UX, by default, there is the following in the given profiles:

set -u 

This confuses the .dbenv*.sh and .sapenv*.sh files of SAP R/3. They fail during execution if the environment is not setup properly. Using SGeSAP the package startup fails trying to bring up the database.

Search the .profile of <sid>adm and remove the set -u, if found.

External Application Server Host Configuration

Repeat the steps in this section for each host that has an external application server installed.

Logon as root.

Create an .rhosts file in the home directories of the HP-UX users root, <sid>adm and (if applicable) informix. Allow login for users root and <sid>adm from all cluster nodes.

For reasons described in Step Title not available, search .profile in the home directory of <sid>adm and remove the set -u, if found.

If you have SAP R/3 4.0A or higher:

For reasons described in Step Title not available, search .profile in the home directory of <sid>adm and remove the set -u, if found.

Add all relocatable IP address information to /etc/hosts.

If you establish frontend and server LANs to separate network traffic:

Add routing entries to the internet routing configurations of /etc/rc.config.d/netconf. This is the only phase of the whole installation in which you will need to specify addresses of the server LAN. Route all relocatable client LAN addresses to the local server LAN addresses. For example, a one package concept:

ROUTE_DESTINATION[n]=<relocdbci>
ROUTE_MASK[n]=""
ROUTE_GATEWAY[n]=<relocdbci_s>
ROUTE_COUNT[n]=1
ROUTE_ARGS[n]=""