Planning and Installing HACMP for Linux

This chapter describes how to plan and install HACMP for Linux. It contains the following sections:
•Cluster Hardware
•Cluster Software
•Planning the HACMP Configuration
•Installing HACMP for Linux
•Contents of the Installation Media
•Installation Process Overview
•Security Considerations
•Where You Go from Here.
Cluster Hardware
This section lists examples of IBM hardware that you can use for cluster nodes, cluster networks and cluster storage disks. For complete information, see the IBM Portal on Linux website:
http://www.ibm.com/linux/
Hardware for Cluster Nodes
HACMP for Linux lets you configure up to eight HACMP nodes. You can use:
•Selected models of IBM System p™ servers
For more information, see: http://www.ibm.com/systems/p/linux/
Also, for descriptions of IBM hardware that you can use as HACMP cluster nodes in AIX, see the HACMP for AIX Planning Guide.
Hardware for Cluster Networks
HACMP for Linux supports the following interconnection networks for clusters:
•Selected modes of 10/100 Mbps Ethernet
•Selected models of Gigabit Ethernet
•Token Ring.
An Ethernet or a Token Ring network can be used as an HACMP cluster IP-based network.
Planning and Installing HACMP for Linux
Cluster Hardware

Hardware for Cluster Storage
HACMP for Linux does not provide high availability for storage resources in your cluster configuration. However you can use NFS or IBM TotalStorage disk subsystems as the storage options in your cluster.
No Automatic NFS and Volume Management
Although you can have disks and file systems configured in the same environment in which your HACMP for Linux cluster is configured, HACMP for Linux does not support NFS file systems. You cannot include file systems associated with the application into the resource groups.
This means that the file systems are not kept highly available by HACMP for Linux. In particular, during fallovers, when applications are moved to other nodes, HACMP for Linux does not automatically unmount the associated file systems on one node and mount them on the takeover node. Similarly, HACMP for Linux does not automatically perform any volume management or volume group operations for volume groups that a particular application needs to access.
However, if you want to manage storage in the cluster, you can still use NFS or GPFS to control it. To ensure that your NFS file systems work within the cluster, you must manage NFS manually, that is, completely outside of your HACMP for Linux cluster.
For example, for a two-node cluster, you can have an NFS server configured somewhere at your site, and make it to export the file system to your cluster nodes. Both nodes will need to mount the file system at boot time. The file system will be also mounted on another cluster node, the one to which the resource group may potentially fall over in cases of failures. Your application and service IP label will be running on one node. On fallover, the application and service IP label will move to the takeover node where the NFS file system has also been mounted since boot time. This way, your application has access to the file system regardless of which node is currently hosting the application. However, the NFS file systems service provided to your application is not kept highly available by HACMP.
As an alternative, here is a cluster configuration that lets you have high availability of your NFS file system in the HACMP for Linux cluster. You can configure an NFS server on a separate two-node cluster, with both nodes running HACMP for AIX, specifically, the nodes should run HACMP’s NFS component (it is part of HACMP for AIX). You can then export the file system from this highly available NFS server to the nodes of your separate HACMP for Linux cluster