HP P6000 Cluster Extension Software Installation Guide (5697-2275, November 2012)

1 Preparing to install P6000 Cluster Extension
This chapter provides information about general environment considerations and pre-installation
requirements.
Ideal environment
The ideal configuration environment consists of at least five servers (two at each data center site
and one as an arbitrator at a third site) with separate, redundant communications links for cluster
heartbeats, client access, and HP P6000 Continuous Access. All communications interfaces must
be installed in pairs to serve as failover components, in order to prevent single points of failure.
Network communication requirements
When using P6000 Cluster Extension, note the following:
• Network communications links and components between the dispersed data centers must be
redundant and physically routed differently to prevent the backhoe issue that is created when
all links between data centers are cut together. This redundancy and routing is especially
important to prevent the split-brain syndrome that occurs when systems in both data centers
form new clusters and allow access to both copies of the data. Using redundant and separately
routed communications and components, you can configure the failover behavior so that the
application service startup procedure stops if none of the remote cluster members can be
reached.
• Supported multipath software must be used for host-to-storage connections.
Data backup requirements
To ensure reliable data backup between the local (primary) and remote (secondary) storage systems,
at least two P6000 Continuous Access links must be available when the storage systems are
connected to provide redundancy and protection against single points of failure. Although
communications links can cover considerable distances, each network segment must be extended
to the dispersed data center to maintain a heartbeat among all servers. For extended distances,
you must purchase extender components to bundle the P6000 Continuous Access links.
Recommendation
Use two systems at each site to give local application service failover procedures among local
cluster systems priority over remote, more time-consuming failover procedures. When failing over,
the software must reconfigure the storage systems to change the mirroring direction. At the remote
site, two systems should be available in case the primary failover system experiences a hardware
or power failure.
Windows: Using an MNS quorum requires an odd number of cluster nodes, with at least one
system in a third location.
CAUTION: P6000 Cluster Extension works with one server at each location, with a single I/O
path between the server and storage system, and a single link between storage systems. However,
any configuration with a single point of failure is not considered highly available or disaster tolerant.
HP does not support configurations with single points of failure; configurations with one system at
each location are supported only if there are no single points of failure.
How P6000 Cluster Extension Software interacts with its environment
This section describes the ways in which P6000 Cluster Extension interacts with its environment.
Ideal environment 5