HP Serviceguard Storage Management Suite A.03.02 Release Notes (March 2012)

For more information on AUTOSTART_CMCLD, see the Setting up Autostart Features and Performing
a Rolling Upgrade sections in Managing Serviceguard, Eighteenth Edition, available at
www.hp.com/go/hpux-serviceguard-docs → HP Serviceguard
Cluster Offline Upgrade Option for SG SMS Bundles Containing CVM/CFS
To upgrade a SG SMS CFS bundle (T2775xx, T2776xx, or T2777xx) where the supported upgrade
path in Table 6 (page 22) indicates an offline upgrade is required, there is a procedure that may
work with your configuration to minimize the time that the entire cluster is offline. This is
accomplished by performing an offline upgrade on half of the cluster nodes while the other half
of the cluster nodes remain online.
NOTE: This cluster offline upgrade option is supported for uprade to SG SMS version A.03.01
only from SG SMS versions containing Serviceguard 11.19 (SG SMS versions A.02.01, A.02.01.01
and A.03.00).
This cluster offline upgrade procedure considers the upgrade of SG SMS only. You need to take
responsibility for applications that run in the SG SMS cluster you plan to upgrade. Before starting
this procedure, make certain your computing environment and the applications running on top of
SG SMS in the cluster can accommodate being offline temporarily. You might benefit from using
Dynamic Root Disk (DRD) utilities to perform your uprade, if your upgrade is supported with DRD.
See “Using Dynamic Root Disk Utilities (DRD) to upgrade to SG SMS A.03.02” (page 25) for more
information.
Offline Upgrade Procedure:
Divide the cluster nodes into two groups. For purposes of explanation, we will refer to these groups
of nodes as Group A and Group B. In explaining this procedure we assume groups of equal size
will be the most efficient approach. It is not necessary to divide the cluster nodes into groups of
equal size, if some other arrangement will work better in your computing environment.
• If the cluster has an even number of nodes, assign half of the nodes to Group A and the other
half to Group B.
• If the cluster has an odd number of nodes, subtract one node from the total number of nodes,
divide that number in half, then assign the resulting number of nodes to Group A. Assign the
remaining, larger group of nodes to Group B. For example, if the cluster contains 9 nodes,
assign (9 - 1)/2 = 4 nodes to Group A, and assign the remaining 5 nodes to Group B.
We will use a 3 node cluster as an example for explaining this offline upgrade procedure. For a
3 node cluster, Group A will contain Node 1, and Group B will contain Node 2 and Node 3.
NOTE: Steps 1, and 2 in this procedure can be performed on multiple nodes in parallel, or on
one node at a time, before continuing to step 3.
1. Move all application failover packages that can be moved from Group A nodes, to Group B
nodes.
2. Halt the Group A cluster nodes and upgrade them to the new SMS version:
a. Node 1 > cmhaltnode –f node1
b. Perform an offline upgrade of the Group A nodes (Node 1 in this example).
3. Gracefully halt application packages running on Group B nodes, before halting the cluster
on the Group B nodes.
24 HP Serviceguard Storage Management Suite Version A.03.02 for HP-UX 11i v3 Release Notes