CICS® collects statistics
on the data written to each journal and log stream; this data can
be used to analyze the activity of a single region. However, because
general log streams can be shared across multiple MVS images, it can be more useful to examine
the statistics generated by MVS.
About this task
The MVS system logger writes SMF Type 88 records containing statistics for each connected log stream. MVS supplies in SYS1.SAMPLIB a sample reporting program, IXGRPT1, that you can use as supplied, or modify to meet your requirements. Alternatively, you can use some other SMF reporting program. For information about the SMF Type 88 records and the sample reporting program, see z/OS MVS System Management Facilities (SMF).
The
main events to monitor routinely are as follows:
- For coupling facility log streams, the number of
structure
full
events
- For DASD-only log streams, the number of
staging data set full
events.
If these events occur frequently, this indicates that
the logger cannot write data to auxiliary storage quickly enough to
keep up with incoming data, which causes CICS to wait before it can write more data.
Procedure
- Consider the following solutions to resolve problems that
occur as a result of event-full conditions:
- Increase the size of primary storage (that is, the size
of the coupling facility structure or, for a DASD-only log stream,
the size of the staging data set), in order to smooth out spikes in
logger load.
- Reduce the data written to the log stream by not merging
so many journals or forward recovery logs on to the same stream.
- Reduce the HIGHOFFLOAD threshold
percentage, the point at which the system logger begins offloading
data from primary storage to offload data sets.
- Review the size of the offload data sets. Offload data
sets must be large enough to avoid too many “DASD shifts”—that
is, new data set allocations. Aim for no more than one DASD shift
per hour. You can monitor the number of DASD shifts using the SMF88EDS
record.
- Examine device I/O statistics for possible contention
on the I/O subsystem used for offload data sets.
- Use faster DASD devices.
The best CICS system
logs performance is achieved when CICS can
delete log tail data that is no longer needed before it is written
to auxiliary storage by the MVS system
logger. To monitor that this is being achieved, your reporting program
can examine the values in the SMF88SIB and SMF88SAB SMF Type 88 records,
which provide helpful information relating to log data.
- SMF88SIB
- Data deleted from primary storage without first being written
to DASD offload data sets. For a system log stream, this value is
normally high in relation to the value of SMF88SAB. For a general
log stream, this value is normally zero.
- SMF88SAB
- Data deleted from primary storage after being written to DASD
offload data sets. For a system log stream, this value is normally
low in relation to the value of SMF88SIB. For a general log stream,
this value is normally high.
Note: In any SMF interval, the total number of
bytes deleted from primary storage (SMF88SIB plus SMF88SAB) might
not match the total number of bytes written to auxiliary storage.
Data is only written to offload data sets and then deleted from primary
storage when the HIGHOFFLOAD threshold limit
is reached.
- If the SMF88SAB record frequently contains high values
for a CICS system log:
-
Check that RETPD=dddd is not specified on the MVS definition of the log stream. For information about the MVS
RETPD parameter, see Managing auxiliary storage.
- Check that no long-running transactions are making recoverable
updates without syncpointing.
- Consider increasing the size of primary storage.
- Consider increasing the HIGHOFFLOAD threshold
value.
- Consider reducing the value of the AKPFREQ system
initialization parameter.