HP Fabric Clustering System HP-UX Administrator's Guide, March 2008

• Description: IPoIB PPA instance numbers start at 9000, but the current NetTL tool does not
support interfaces whose PPA instance numbers are greater than 255.
When running a trace on the system, the correct PPA number is provided, but when
formatting trace packets for a specific PPA (known as filtering), you need to make an
adjustment in the request to obtain the proper information.
Symptoms: Unless a specific PPA instance number is adjusted by subtracting 9000, the trace
request will not work as expected.
Workaround: When filtering on a particular IPoIB instance, like 9005, specify the instance
number“5” instead of the actual interface number.
Filter Example: In the example below, the filter configuration file may contain the following
line to view trace packet information only for IPoIB PPA 9005.
formatter filter device_id 5 IPOIB
Workaround Side-effect: When the subsystem name (IPOIB) is omitted in the filter
configuration line shown above, the filtering and hence the packets will be shown for the
physical PPA of 5 and also IPoIB PPA of 9005.
• Description: IPoIB MAC Address and header is not properly printed when formatting is in
“nice” or “terse” format.
Symptoms: When formatting trace packet information with “nice” format option, APP
packets are not formatted nicely. For “terse” formatting, the packet header is not printed
properly.
Fix: The problem is fixed in NetTL patch PHNE_30451.
IP Filter Product
• Description: IPoIB interfaces that are SNAP-configured using the ifconfig utility in
combination with the IP Filter product being installed, may cause a panic.
Symptoms: The system may panic with the IP Filter Product version “B9901AA A.03.05.06
HP IPFilter 3.5alpha5”.
Workaround: Update the IP Filter product to the most recent version, which can be found
at http://www.software.hp.com, and is a free download.
Monitoring and Troubleshooting the Switch
The CLI can also be used to monitor the system.
NOTE: HP is in the process of transitioning to a new switch, the Cisco SFS 7000P. For
documentation and ordering information, see
http://www.cisco.com/en/US/products/ps6421/index.html
Health Monitoring
The system is a robust product with fail-over protection. In the event a physical component fails,
a similar one automatically assumes the load of the failed component.
The physical components should be checked on occasion, particularly if there are problems with
the initial setup and configuration.
The HP Fabric switch, and controller may be monitored in several ways. Switch cluster connections
and controller cards include a general system LED to indicate the physical health of the card.
Card status information may be displayed in the CLI that indicates the success or failure of the
boot process.
Ports may be monitored in several ways. You may check a port physically by looking at the port
LEDs. You may also open various graph and statistics windows to notice network traffic across
124 Monitoring and Troubleshooting