Pathway/iTS System Management Manual (G06.24+)

Maintaining Pathway/iTS Objects
HP NonStop Pathway/iTS System Management Manual—426748-002
4-22
TCP-Specific Problems
When describing the problem to your service provider, include details of how to
reproduce the problem by navigating to particular screens and pressing particular keys.
TCP Dump
When a TCP detects an inconsistency in its control tables, the TCP dumps
automatically if the dump feature is enabled. To enable the TCP dump feature, perform
the following:
>PATHCOM $<pm>
=CONTROL <tcpname>, DUMP ON (FILE <tcpdump>)
=EXIT
where pm is the name of your PATHMON process, tcpname is the name of the TCP
and tcpdump is name of the file where you want the TCP to dump errors.
If the TCP tables (showing, for example, terminal states or TCP states) appear to be in
error, you can force a dump by performing the following:
>PATHCOM $<pm>
=CONTROL <tcpname>, DUMPMEMORY <option> (FILE <pmdump>)
=EXIT
where option is PRIMARY, BACKUP or BOTH. Use PRIMARY unless a TCP is
reporting errors from its backup.
If the TCP primary process is running in the CPU defined for the TCP backup process
because of a SWITCH command or some other processor change, the BACKUPCPU
attribute causes the PATHMON process to change the PATHMON configuration file so
that it contains the current CPU numbers for both the primary and backup TCPs.
If a TCP was configured with the NONSTOP attribute set to 0, changing the
BACKUPCPU attribute only changes the backup CPU value in the PATHMON
configuration file; no processor change occurs.
When you cool start a PATHMON environment after using the CONTROL TCP
command, the TCPs start according to the changes recorded in the PATHMON
configuration file.
If you request the help of your service provider in analyzing a problem, the
representative will likely require a dump file. It is therefore recommended that you
always set the DUMP option to ON for production systems.
The DUMPMEMORY option is not a substitute for setting the DUMP option to ON.
DUMPMEMORY is primarily useful in a controlled troubleshooting situation where you
need to take a snapshot of the TCP’s internal state at a particular time - before it
encounters a fatal error.