Guidelines for Disaster Tolerant Architectures with FibreChannel

Two data centers using FibreChannel hubs. This architecture can be implemented using a variety of disk arrays. Consult the HP 9000 Servers Configuration Guide (available through your HP representative) for a list of supported HA disks.

Two data centers using FibreChannel point-to-point and HP SureStore E Disk Array XP Series or EMC Symmetrix disk arrays.

Data centers must be self-contained such that the loss of one data center does not cause the entire cluster to fail. All Single Points of Failure (SPOFs) must be eliminated so that surviving systems continue to run if one or more systems fail.

The number of nodes in each data center must be equal.

Heartbeat networks between the data centers must be redundant and routed along physically different paths so that the loss of any one data center does not cause the network between surviving data centers to fail. All nodes in the cluster must be connected to the redundant heartbeat network links.

If you are using FibreChannel technology for the network as well, you must use completely different hardware from that used to connect nodes to mass storage devices. This includes cables, repeaters, hubs, and anything else related to the FibreChannel fabric.

Data must be mirrored at the remote site using a product such as MirrorDisk/UX. Data connections must be redundant and routed differently, like heartbeat network cables are. It is recommended to route FibreChannel cables and heartbeat network cables in such a way that no two failures can disable all hearbeat networks and both FibreChannel connections.

Data components, e.g. FibreChannel hubs, must be redundant and powered using alternative power sources as described in Chapter 1. All failover nodes must be connected to both mirror copies of the data.

Dual cluster lock disks are required, one in each data center. All nodes must be connected to both cluster lock disks. Each cluster lock disk must be on a different FibreChannel arbitrated loop.