Hardware configuration

---

The hardware design and configuration for this type of work must be carefully planned and thought out before purchasing any devices for the configuration. If this is not done properly, the deployment of your design may be stalled until all component issues are resolved.

There are several groups of people who would be involved in this design, and various team members may be able to assist in the configuration.

Disk storage design

The disk storage design and configuration is a critical component to a successful HACMP failover design. This disk configuration must be able to be seen by all nodes within the cluster.

Our selection for this centralized disk storage is based on IBM 7133 SSA storage arrays.

Note 

The redundant SSA controllers must be of the same version and revision. Different levels of controllers provide different raid levels, speeds, or other functions, thereby introducing incompatibility problems into the HACMP design.

Heartbeat interface

The HACMP heartbeat design is a critical component to a stable HACMP deployment.

Our design uses the Non-IP Network Serial Cable method, because of:

• Simplicity; once the cable is installed and tested, the configuration will probably never be touched again.

• There are no electrical or power issues associated with this configuration.

• The design is portable in the event you migrate from one disk technology to another (for example, SCSI to SSA).

• There are no moving parts to this configuration, so there is virtually no mean time between failure (MTBF) issues on a serial cable.

Ethernet connectivity

Proper network connectivity is critical to a successful HACMP deployment. There is little purpose to continuing without it, as HACMP will not validate or accept the configuration if the network is not properly configured.

Currently we have three Ethernet adapters per machine (en0, en1, en2), totaling 6 adapters. This configuration has six IP addresses, plus one more that is actually used for the IBM Tivoli Workload Scheduler service that all IBM Tivoli Workload Scheduler FTAs connect to (the service address).

We will use IP aliasing in the final production environment; this aliasing process promotes a very fast HACMP failover.

Notes 

• Understanding the network configuration is probably one of the most critical components to the HACMP configuration. Find assistance with this step if you do not have a good understanding of the HACMP and networking relationship.

• All adapters to be utilized within the HACMP solution must reside within different network subnets (but the netmask must be the same).

---