HP XC System Software Installation Guide Version 4.0

The InfiniBand firmware binary files are included in SFS and are automatically installed on
all SFS nodes. However, the firmware on the HCA adapter must be upgraded manually,
using these firmware binary files.
If you need assistance with identifying and upgrading the firmware on the InfiniBand
switches and HCA adapters, contact your local HP support center.
7. If the system is not configured with improved availability of services, determine whether
you want to configure that feature. Improving the availability of services is a feature that
was introduced in HP XC System Software Version 3.1. Improved availability of services
requires the use of HP Serviceguard, a product that is purchased and licensed separately
from HP.
See Section 1.9 (page 26), which describes what improved availability offers, how it is
achieved, the supported availability tools, and the services that are eligible for improved
availability.
8. Read the Version 4.0 HP XC System Software Release Notes before beginning the upgrade
procedure. The release notes describe any last-minute changes to software, firmware, or
hardware that might affect the upgrade process.
The HP XC System Software Release Notes for each HP XC software release are available online:
http://www.docs.hp.com/en/linuxhpc.html
9. Read the HP Scalable Visualization Array, V2.1.3 Release Notes before beginning the upgrade
procedure. Upgrade and reinstallation procedures for SVA are part of Section 9.7 (page 146)
Proceed to Section 9.3 (page 141).
9.3 Preparing the System State
On the head node, follow this procedure to ensure that the system is in the appropriate state for
the upgrade:
1. Set all nodes to network boot so that all nodes can be reimaged after the head node is updated:
# setnode --resync --all
2. Stop all running jobs. See the HP XC System Software User's Guide for specific instructions.
3. If the system is configured with improved availability of services, stop the availability tool:
# transfer_from_avail
4. If you are using Serviceguard as the availability tool, identify and remove the Serviceguard
RPMs:
a. Identify the RPM names of the following Serviceguard RPMs that are installed on the
system:
# rpm -qa | grep pidentd
# rpm -qa | grep qs
# rpm -qa | grep serviceguard
b. Remove the RPM names obtained in the previous step:
# rpm -e pidentd-{version}
# rpm -e qs-A.{version}.rhel{version}
# rpm -e serviceguard-{version}.rhel{version}
5. Shut down all nodes except the head node:
# stopsys
6. If the /hptc_cluster file system is mounted over an SFS server, perform the following
task; otherwise, proceed to step 7.
a. Determine the processes that are using /hptc_cluster:
# fuser -vm /hptc_cluster
9.3 Preparing the System State 141