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

12
Latency considerations for Integrity Virtual Machines
With Integrity Virtual Machines there is a risk that a heavy processing load will cause Serviceguard to
lose cluster heartbeats and trigger a cluster-reformation or node failure. If you encounter this situation,
adjust the Serviceguard NODE_TIMEOUT or MEMBER_TIMEOUT upwards. This can result in longer
failover times.
Other Linux differences
There are some restrictions listed in this document that are considered strong recommendations for
Linux configurations. If these restrictions are violated then some failures will cause the failure of a
node when only an interface or network card has failed. For example, Serviceguard for Linux will
allow the use of just one dual channel Fibre Channel interface cards for storage connectivity as long
as the customer is willing to accept that the failure of this card will cause the entire server to fail.
Serviceguard for Linux does not require that redundant I/O paths be configured in separate I/O
chassis although it is strongly recommended. If redundant paths are configured in a single I/O
chassis, then failure of that chassis will result in the failure of the server.
Summary and conclusion
With careful consideration of hardware redundancy, elimination of single points of failure, use of
arbitration (as needed), and appropriate I/O and networking configuration, using Serviceguard with
partitioning or virtual machine technologies can provide you with great protection against unavailable
software and hardware.