HP Serviceguard Cluster Configuration for HP-UX 11i or Linux Partitioned Systems, April 2009

10
up exclusively of nPartitions since an nPartition is isolated from hardware and software failures of
other nPartitions.
Figure 5. Cluster-in-a-box configuration
I/O considerations
Serviceguard does not treat OS instances running in a nPar or vPar any differently than those running
on an independent node. Thus, partitions do not provide any exemptions from the normal
Serviceguard connectivity rules (such as redundant paths for heartbeat networks, and to storage) nor
do they impose any new requirements. With Integrity VM, Serviceguard automatically waits
additional time to allow I/O on the failed node to complete before resuming cluster operations.
There are a couple of interesting aspects related to partitioned systems that should be noted:
• While not strictly a “partitioning” issue, the Superdome platform that supports nPartitions contains
its interface cards in an I/O chassis, and there can be more than one I/O chassis per partition. To
avoid making the I/O chassis a single point of failure, configure redundant I/O paths in separate
I/O chassis. Generally speaking, Superdome provides enough I/O capacity that the Serviceguard
redundant path requirement should not constrain the use of partitioning within the cluster.
• vPars on the other hand must share essentially one node’s worth of I/O slots. In this case, the
redundant path requirement can be a limiting factor in determining the number of partitions that can
be configured on a single hardware platform.
• Virtual machines can share the same I/O devices (storage and networking). This allows more
virtual machines to be created since they aren’t limited by having independent I/O devices (as
compared to vPars). However, the failure of a single physical I/O interface impacts all virtual
machines that share it. For performance, implementing network and storage redundancy at the VM
host level rather than at the VM guest level is recommended.
For example, assume we would like to create a cluster-in-a-box configuration using a Fibre Channel–
based storage device. The redundant path requirement means that each partition would need two
Fibre Channel interface cards for storage. Each partition would also need a minimum of two network