When you design a Sun Cluster HA for SAP configuration, consider the following guidelines.
Use an SAP software version that is qualified with Sun Cluster 3.0 - The Solaris 8 operating environment offers support for the Sun Cluster software.
Use an SAP software version with automatic enqueue-reconnect-mechanism capability - The Sun Cluster HA for SAP data service relies on this capability. SAP 4.0 software with patch information and later releases should have automatic enqueue-reconnect-mechanism capability.
Read all related SAP online service system notes for the SAP software release and database you are installing on your Solaris platform - Identify any known installation problems and fixes.
Consult SAP software documentation for memory and swap recommendations - SAP software uses a large amount of memory and swap space.
Generously estimate the total possible load on nodes that might host the central instance, the database instance, and the application server, if you have an internal application server - This guideline is especially important if you configure the cluster so that the central instance, database instance, and application server will all exist on one node if failover occurs.
Install application servers on the same cluster that hosts the central instance or on a separate cluster - If you install and configure any application server outside of the cluster environment, the Sun Cluster HA for SAP data service does not fault-monitor and does not automatically restart or fail over those application servers. You must manually start and shut down application servers that you installed and configured outside of the cluster environment.
Limit node names and logical hostnames to eight characters or less - This limitation is an SAP software requirement.
See your Enterprise Services representative for the most current information about supported SAP versions. The following figures illustrate sample configurations for the Sun Cluster HA for SAP data service.
This figure was a common configuration under previous Sun Cluster releases. To use the Sun Cluster 3.0 software to the full extent, follow Figure 9-1 or Figure 9-3.
Before installing the SAP software, see "Installing and Configuring SAP and the Database", and consider the following cluster-related issues.
Install SAP binaries and SAP users' home directories - Install SAP binaries and users' home directories on the cluster file system. Installation on the cluster file system, however, has some drawbacks with SAP software release upgrades. See "Determining the Location of the Application Binaries" for information about drawbacks.
After you create all the file systems for the database and for SAP, create the mount points, and put the mount points in the /etc/vfstab file on all cluster nodes - See the SAP installation guides, Installation of the SAP R/3 on UNIX and R/3 Installation on UNIX-OS Dependencies, for details on how to set up the database and SAP file systems.
Create the required groups and users on all cluster nodes - Create the required groups and users for the SAP software on all cluster nodes according to the SAP installation guides, Installation of the SAP R/3 on UNIX and R/3 Installation on UNIX-OS Dependencies.
Configure the Sun Cluster HA for NFS data service on the cluster that hosts the central instance if you plan to install some external SAP application servers - See "Installing and Configuring Sun Cluster HA for NFS" for details on how to configure the Sun Cluster HA for NFS data service.
Set up the /etc/nsswitch.conf file so that the data service starts and stops correctly during switchovers or failovers - On each node that can master the logical host that runs the Sun Cluster HA for SAP data service, the /etc/nsswitch.conf file must have one of the following entries for group.
group: group: files group: files [NOTFOUND=return] nis group: files [NOTFOUND=return] nisplus |
The Sun Cluster HA for SAP data service uses the su user command to start and stop the database node. The network information name service might become unavailable when a cluster node's public network fails. Adding the preceding entries for group ensures that the su(1M) command does not refer to the NIS/NIS+ name services if this unavailability occurs.