Best Practices for HP BladeSystem Deployments using HP Serviceguard Solutions for HP-UX 11i (May 2010)

21
Figure 13: Cluster Example Spanning Multiple c7000 Enclosures
In this example, a single 8-node Serviceguard cluster spans two c7000 enclosures, with 4 nodes in
each enclosure attached to shared storage provided by an EVA disk array. A Virtual Connect
stacking link is used between the enclosures to provide a private cluster heartbeat network between
the enclosures. A Systems Insight Manager Central Management Server is used to provide overall
management of the systems environment; however this server is not part of the Serviceguard cluster.
Note that it is permissible to mix server blades with other external physical or virtual (i.e., nPar, vPar,
HP Integrity Virtual Machine) Serviceguard nodes in this configuration. However; no enclosure or
complex can contain more than half of the cluster nodes (e.g., server blades, nPars), and requires an
external quorum server to maintain cluster quorum in the event of a blade enclosure failure. A quorum
service, running on a small HP-UX system in this example, is located outside of the blade enclosures to
serve as a tie-breaker in case of an enclosure failure to allow the remaining 50% of the surviving
cluster nodes to successfully form a new cluster. Additional information on Serviceguard cluster
quorum requirements is available in the white paper titled “HP Serviceguard Cluster Configuration for
HP-UX 11i or Linux Partitioned Systems” posted at
http://docs.hp.com/en/6033/HPServiceguardClusterConfig_WP.pdf.
Advantages and Other Considerations
Having a Serviceguard cluster span multiple c7000 enclosures or other non-blade cluster nodes has
many advantages over a “cluster in a box” configuration, and few limitations:
HP Systems Insight Manager (SIM)
Central Management Server (CMS)
c7000 BladeSystem
Enclosure
EVA
Disk Array
c7000 BladeSystem
Enclosure
Serviceguard
cluster
Quorum Service (HP-UX OS)
VC Stacking Link
LAN
SAN