ServerNet Cluster Manual
Upgrading a ServerNet Cluster
ServerNet Cluster Manual—520575-003
4-54
Merging Clusters to Create a Split-Star Topology
Merging Clusters to Create a Split-Star
Topology
To create a split-star topology, you must merge two clusters that use one cluster switch
per fabric. Typically, you will merge two clusters that use the star topology. However,
you can also merge valid subsets of other topologies to create a split-star topology.
This subsection contains the following examples and procedures:
•
Example: Merging Two Star Topologies to Create a Split-Star Topology on
page 4-54
•
Steps for Merging Two Star Topologies to Create a Split-Star Topology on
page 4-60
Example: Merging Two Star Topologies to Create a Split-Star
Topology
This upgrade begins with two ServerNet clusters having up to eight nodes and running
the G06.09, G06.10 or G06.11 RVU. The upgrade:
•
Installs new versions of the software listed in Table 4-6 on page 4-8 on all nodes if
necessary
•
Upgrades the firmware and configuration in the cluster switches if necessary
•
Reconfigures the clusters if necessary so that the ServerNet node numbers do not
overlap
Following the upgrade, the merged cluster will use the split-star topology and support
up to 16 nodes.
Table 4-16 summarizes this upgrade.
Note. HP recommends upgrading to the latest software whenever possible. See
Upgrading Software to Obtain G06.14 Functionality on page 4-34.
Note. The upgrade requires that the G06.12 version of SANMAN (T0502AAE or
superseding) be running in all nodes in the cluster. T0502AAE is required because
versions of SANMAN earlier than G06.12 cannot support ServerNet node numbers 9
through 16. In addition, the cluster switches must be running the T0569AAB firmware and
configuration. T0569AAB is required for the split-star topology (two cluster switches per
fabric).
Note. The migration of a cluster earlier than G06.12 to a G06.12 cluster using the X2/Y2
cluster switches requires shutting down ServerNet cluster communications for all nodes in the
cluster. Shutting down ServerNet cluster communications is required because the ServerNet
node numbers in the cluster must be changed from numbers 1 through 8 to numbers 9 through
16 once both the X-fabric and Y-fabric cluster switches in that cluster are upgraded to the
X2/Y2 cluster switch configuration.