HP XC System Software Administration Guide Version 4.0

Table Of Contents
0:00:00:00:00:00:00
inet6 addr: fe80::000:0000:0000:000e/64 Scope:Link
UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
RX packets:22989 errors:0 dropped:0 overruns:0 frame:0
TX packets:35 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:256
RX bytes:1299245 (1.2 MiB) TX bytes:5572 (5.4 KiB)
The interface (or interfaces) should be reported as UP and RUNNING. If not, and you are
using a dual ported InfiniBand card, make sure that the InfiniBand cable is connected to
Port 1 of the InfiniBand card. You can try to ping other nodes that are connected to the
network.
8. Find additional information about InfiniBand in the /sys/class/infiniband* directory.
Use the find command to locate the information:
# find /sys/class/infiniband* -type f -print -exec cat {} \;
9. Consult the documentation for the available OFED commands:
You can access the manpages for OFED commands, such as ibv_devinfo and
ibstatus, with the man command.
Additional documentation on OFED is available from the ofed_docs and
kernel_ib_devel RPMs.
The InfiniBand vendors also have OFED documentation available from their web
addresses.
21.5 Improved Availability Issues
This section contains notes that apply when improved availability of services has been configured.
The HP XC System Software Installation Guide describes how to configure services for improved
availability.
21.5.1 How To Start HP Serviceguard When Only the Head Node is Running
Serviceguard does not automatically start up the Serviceguard cluster when only one node is up
and running and the quorum server or lock LUN is not accessible. Serviceguard was designed
that way to prevent a split brain scenario. A split brain scenario could occur if two nodes of a
Serviceguard cluster cannot communicate with each other and each determine that the other
node is down. If both nodes are allowed to start up their own one-node cluster, both nodes could
attempt to start up packages and utilize the same resources. This could lead to a variety of
problems including data loss.
Run the following command if you want to start up Serviceguard when only the head node is
up and running. If you did not install Serviceguard in its default location, you must specify the
full path to the command.
# cmruncl -n {nodename}
You are prompted to verify that starting Serviceguard on this node is correct. At some later time,
when the other configured node of the Serviceguard cluster is booted, it joins the running cluster.
All packages are running on the head node. You must manually relocate any packages that are
required to be running on the other node of the Serviceguard cluster.
21.5.2 Restart Serviceguard Quorum Server if Quorum Server Node is Re-imaged
You must restart the Serviceguard quorum server if the quorum server node is re-imaged while
improved availability is running. Issue the following command on the quorum server node to
restart the quorum server and update the /etc/inittab file:
# /opt/hptc/availability/serviceguard/startquorum
260 Troubleshooting