If
the system log becomes corrupt, CICS quiesces. After the system log has been
corrupted, it cannot be used again; to get CICS back into production, you
must perform an initial start. To prevent the problem recurring, you also
need to gather diagnosis information that will enable IBM® Service to discover
why the log was corrupted. Unfortunately, performing an intitial start destroys
all information from the previous run of CICS.
One way to gather the required diagnosis information is to scan the failed
system log, using a utility such as DFHJUP. However, the output produced by
DFHJUP in these circumstances is not easy to interpret. To supplement DFHJUP’s
output, you should perform a diagnostic run of CICS,
using the corrupt system log, before performing the
initial start. On a diagnostic run, CICS:
- Produces a dump of the CICS system state, retrieved from the failed system
log.
- Terminates. Note that, on a diagnostic run, CICS performs
no recovery work and no new work.
The output produced by a diagnostic run is usually passed to IBM Service.
Benefits of a diagnostic run
The advantages of performing a diagnostic run are:
- It collects diagnostic information automatically, thus allowing you to
get CICS back into production quickly.
- When CICS failed, a system dump may not have been produced. A diagnostic
run provides one.
- If the diagnostic run is not able to retrieve all the records from the
CICS system log, the last record it retrieves shows the point at which the
log became unreadable, and may indicate the cause of the problem.
- A diagnostic run allows you to capture a dump of the MVS logger address
space--see Getting dumps of the MVS logger and coupling facility address spaces.
How to perform a diagnostic run
For a diagnostic run to take place, AUTO must be specified on the START
system initialization parameter. If the system log becomes corrupt, CICS:
- Sets the recovery manager autostart override record in the global catalog
so that the next automatic restart of CICS is a diagnostic run (AUTODIAG).
- Issues message DFHRM0152, saying that the next automatic restart will
be a diagnostic run, and should be performed before an initial start.
There may be occasions other than when the system log has become unusable
when you feel it would be useful to perform a diagnostic run. You can use
the recovery manager utility program, DFHRMUTL, to specify a diagnostic run.
For information about DFHRMUTL, see the CICS® Operations and Utilities Guide.
For reliable diagnosis, it is important that you have dumps of the MVS
logger address space and (if applicable) the coupling facility structures
used by the system log. This means that, before performing the diagnostic
run, you will probably need to set a SLIP trap, as described in Setting a SLIP trap.
You need to set a SLIP if any of the following are
true:
- You have not applied MVS APAR OW27057.
- MVS did not produce a dump of the logger address space.
- MVS produced a dump of the logger address space but you have not kept
the dump.
- The CICS system log uses coupling facility log streams.
When specifying the SLIP, note the following:
- Set the trap for the specific DFHLG07xx message that CICS issued when
the original failure occurred--see the example SLIP in topic Figure 20.
When the diagnostic run occurs and the failure repeats, the message will drive
the SLIP.
Occasionally, the DFHLG07xx message that was issued at the time
of the original failure is not repeated during a diagnostic run--instead,
a different DFHLG07xx message is issued. Therefore the SLIP is not triggered.
If this happens, perform another diagnostic run. This time, however, set the
SLIP for the DFHLG07xx message that was issued during the first diagnostic
run.
- You do not need to specify a dump of the CICS system, because one is taken
automatically by the diagnostic run mechanism. Change the JOBLIST parameter
in the example SLIP to read:
JOBLIST=(IXGLOGR,XCFAS),
- Specify a dump of the MVS logger address space--see the example SLIP.
(If you have applied MVS APAR OW27057, and the original failure occurred
because the MVS logger was unable to find a specific log stream block identifier,
an extra dump may be produced.)
- If the system log uses coupling facility log streams, specify a dump of
the coupling facility structure. You can get the name of the structure (or
structures) from the two DFHLG0104 messages that were issued when CICS connected
to DFHLOG and DFHSHUNT during the run in which the failure occurred.
If
DFHLOG and DFHSHUNT use separate coupling facility structures, dump both structures.
Specify the names of both structures on the STRLIST parameter.

Shutdown and restart recovery
[[ Contents Previous Page | Next Page Index ]]