HP-UX LAN Administrator's Guide (Feburary 2007)

Troubleshooting HP-UX LANDRAFT COPY
Troubleshooting Q & A
Chapter 542
The file, filename .TRC0 is the most recent trace file. If this file does not contain the trace
information you are looking for, check the
filename .TRC1 file.
To format your LAN trace using a filter file, execute:
netfmt -c filterfile -N -f filename .TRC0 > fmt0
nettl appends TRC0 or TRC1 to the name you give the raw trace file.
Related Documentation: Refer to Chapter 5, “LAN Resources.”
Intermittent networking problems: I'm experiencing intermittent networking problems
on my computer. What should I check to ensure proper operation of my networking software?
Answer: Upper layer software often requires loopback. Check /etc/rc.config.d/netconf to
be sure that the loopback entry is correct. The line in netconf file should read:
LOOPBACK_ADDRESS=127.0.0.1
Related Documentation: Refer to chapter 6, “Network Addressing”.
Performance: I've noticed a significant drop in system response time and performance. What
steps can I take to improve it?
Answer: Performance may be affected by many different factors. Sometimes removing pseudo
drivers from the kernel for networking software that you may not be using improves
performance. The problems may also be in the upper layer software (ftp or telnet).
Also, it is possible that too little memory is allocated to hold fragmented messages in the IP
layer. IP messages may be fragmented into smaller parts when the message is sent through
the system. The fragments must be held in memory for some time so that the entire message
can be reassembled because the fragments arrive at the destination at different times and
possibly out of order. Normally, fragmentation reassembly memory is limited arbitrarily so
that incomplete messages do not consume all of memory, which could cripple the system.
During stressful networking activity, some fragments might never be delivered because they
are typically dropped in transit, for example, due to a collision or resource limitations on an
intermediate system. However, fragments might also not be delivered (“dropped”) if there is
insufficient fragmentation reassembly memory on the destination system during periods of
high network activity. This can degrade performance due to retransmissions of data. If the
problem is due to a high number of fragments dropped after time-out (see the output from the
command netstat -sp ip), you might want to increase the size of the fragmentation
reassembly memory by changing the ip_reass_mem_limit value using the ndd command.
(The default is 2 MB for the system.) Enter the command /usr/bin/ndd -h to display ndd
parameters and their use.
Deferred transmissions/collisions: Why is there a significant increase in the number of
deferred transmissions and collisions on my network?