 |
» |
|
|
|
This section gives some suggestions for troubleshooting run-time problems.
It is divided into the following sections:
Remotely Initiated TP Failure
Lost Data |  |
If data loss occurs at either the HP 3000 side or
the remote side, follow these steps: Get a description of what is missing.
Get copies of the original document or file sent and the document
or file that was received with data missing. Use a protocol analyzer to determine whether data
is being transferred. If you suspect that the HP 3000 is causing
the data loss, collect the following things and send them to your
HP representative for analysis: the SNA Transport user trace
(see the SNA Link/XL Node Manager's Guide) copies of the document or file sent and the document
or file received with data missing the APPC subsystem internal trace the NMMAINT.PUB.SYS version stamp
APPC Subsystem Hang |  |
After you issue the APPCCONTROL STOP command to stop the APPC subsystem (or after a TP calls
the APPCStop intrinsic), the message APPC: Subsystem Stopped (APPCLOG 1) should be logged to the console screen. Issue
the APPCCONTROL STATUS command to make sure the subsystem has shut down completely.
The APPCCONTROL STATUS display will list any APPC sessions or transaction
programs that are still active after the APPCCONTROL STOP command has executed. If the APPC: Subsystem Stopped message does not appear, or if the APPCCONTROL STATUS display indicates that APPC sessions are still active,
then the APPC subsystem is probably hung. Follow the procedure below
to collect all needed diagnostic information and bring the subsystem
down. Note that if user and internal tracing were not enabled at
the time the problem occurred, you will have to recreate the problem
with both traces enabled in order to get tracing information. Collect the following information for
analysis by your HP representative: The version stamps from running NMMAINT.PUB.SYS. The NMS log file (NMLGxxxx). The current log file will be locked, so issue
the SWITCHNMLOG command to close the current log file and start a new
one. Any console messages you receive.
Try to force the APPC subsystem to go down. The
APPCCONTROL STOP command allows you to choose among three StopTypes. The highest
priority StopType is K[ILL], followed by P[ROTOCOL], and Q[UIESCE] (the default). Try issuing APPCCONTROL STOP;TYPE=P[ROTOCOL]. If the APPC subsystem still remains active, issue
APPCCONTROL STOP;TYPE=K[ILL]. If you do not see the APPC: Subsystem Stopped message, issue the APPCCONTROL STATUS command to see which sessions are still active. Issue
the APPCCONTROL STOPSESSION command for each active session, then issue the APPCCONTROL STOP command again. If the APPC subsystem still does not come down,
try to bring down the line from the host side. (Be sure to check
with other users of the line.) Collect the user and internal trace files. Submit all diagnostic information to your HP representative.
APPC Subsystem Failure |  |
If the APPC subsystem comes down unexpectedly, follow the
procedure below to collect the data your HP representative will
need to analyze the problem. Note that if user and internal tracing
were not enabled at the time the problem occurred, you will have
to recreate the problem with both traces enabled in order to collect
the tracing information. Issue the SNACONTROL STATUS command to check the activity of SNA Transport. Create
a copy of the display or any error messages that the command generates. Find the APPC dump if there is one. APPC dumps are
created automatically in certain cases. The following message should
appear on the console, containing the name of the dump file: APPC: Object dumped to file APPCDP03 (APPCLOG 29).
|
The file will be called APPCDPxx.APPC.SYS where xx is a number from 00 through 49. To be sure you collect
the most recent file, store off all files with the name APPCDPxx. Collect the following information for analysis by
your HP representative: The user trace files and the
APPC internal trace files. The version stamps from running NMMAINT.PUB.SYS. The NMS log file. The name will be NMLGxxxx.PUB.SYS. The current log file will be locked, so issue
the command SWITCHNMLOG to close the current log file and start a new one.
Submit the data you collected to your HP representative
for analysis.
Transaction Program Hang |  |
When a user reports a deadlocked or hung transaction program
(a program that does not return control to the user), the problem
could be caused by the program or by the APPC subsystem. Follow
the procedure below: Record any messages the user received. Check the error and logging messages. Issue the APPCCONTROL STATUS command. If the subsystem is not active, see "APPC
Subsystem Failure," earlier in this chapter. If the subsystem
is still active, find out which session the deadlocked TP is using. Try to bring down the session with the APPCCONTROL STOPSESSION command. If the session does not come down, see "APPC
Subsystem Hang," earlier in this chapter.
Remotely Initiated TP Failure |  |
If a remotely initiated TP that worked on a previous version
of APPC fails after you install a new version, get a user trace
of the TP and check the status info value returned on the MCGetAllocate intrinsic. If the status info value is -1008, do the following things: Modify the local TP so that it passes
the LocalTPName as an input parameter to the MCGetAllocate intrinsic. On older versions of the APPC subsystem, the LocalTPName was an output parameter. However, on the Node Type
2.1 version of the APPC subsystem, the LocalTPName is an input parameter. Make sure the LocalTPName parameter of the MCGetAllocate intrinsic matches the LocalTPName parameter of the TPStarted intrinsic. Use NMMGR to configure the LocalTPName as a remotely initiated TP. See Chapter 4 “APPC Subsystem Configuration” in
this manual for more information.
System Hangs or Failures |  |
In the event of a system hang or failure, perform the following
steps: Take a memory dump, format the dump
and store off the spool file to tape. Store the following files off to tape and submit
them, along with the memory dump, to your HP representative for
analysis. The APPC subsystem internal
trace and the LU 6.2 product user trace. The version stamps from running NMMAINT.PUB.SYS. The NMS log file (NMLGxxxxPUB.SYS). The current log file will be locked, so issue
the command SWITCHNMLOG to close the current log file and start a new one.
|