Previous topic |
Next topic |
Contents |
Contact z/OS |
Library |
PDF
Reports from the Record Selection and Validation Program z/OS DFSMS Managing Catalogs SC23-6853-00 |
|
For complete examples of the reports from Record Selection and
Validation, see Figure 1 and Figure 2. In this section we will
repeat various lines from these examples with explanatory notes.
Each page of reports from Record Selection and Validation begins
with these lines, identifying first the name, release and modification
level numbers of the product writing the reports. The second line
identifies the program by name and the DD name being used. The execution
date is given in both Gregorian and Julian formats followed by the
time at which program execution began. Report pages are numbered consecutively.
The report identification appears on each page.
Each page of report information from Record Selection and Validation lists the execution parameters. Because the results must be viewed relative to these values, review the parameters to make sure that:
or
Record Selection and Validation produces first a summary report
for all systems and then individual reports for each system encountered
in the SMF data. You should first review the report for all systems,
and then only if further information or investigation is needed, proceed
to the individual system reports. The report analysis is basically
the same for all reports, except that the first report will tell you
about the different systems encountered and provides an overall summary.
This
is the maximum condition code encountered during program execution,
for all systems or for an individual system, depending on the report
being viewed.
If
the count of GAPS AND/OR LOST DATA RECORDS is not
zero, you should analyze the following data to locate the cause.
If
the count of RECORDS SELECTED is zero, your subsequent analysis should
try to determine if this could be correct. Since the numbers will
vary by catalog, by elapsed time and by activity, it is not possible
to present guidelines on what these numbers should be. This line should
be the total of the counts below of (TYPE 6x) RECORDS SELECTED.
Normally you might
expect somewhat equal counts for DEFINE and DELETE with a smaller
count for ALTER. This example is not very representative of what you
should expect to see.
These lines appear only on the report for all systems.
Look at the count for SYSTEM(S) that RECORDED CHANGES TO THIS CATALOG
and beneath it, the list of system identifiers (normally more than
one line). Make sure that all systems having access to the catalog
are in the list. However, if a system that had access to the catalog
did not update the catalog it will not be listed here. There will
be a system report for such a system and you should look there to
be sure that that system's data was not omitted.
Note that the next four lines concern only the SMF records for
the catalogs named here.
The
date and time of the OLDEST SMF CATALOG RECORD FOUND should normally
precede the specified recovery start time. It should also precede
the next entry.
The
date and time of OLDEST SMF CATALOG RECORD SELECTED should usually
coincide fairly closely with the specified recovery start time. If
EXPORT is being used for backup, there should be an SMF record a few
seconds after the specified start time, resulting from the catalog
update to turn on the temporary EXPORT indicator.If the OLDEST SMF CATALOG RECORD SELECTED precedes the specified
recovery start date and time, it should be within the interval specified
as the multi-system clock difference (which is used to establish an
effective start time). This condition may account for spurious error
messages, if records in this interval result in reprocessing changes
already represented in the EXPORTed copy of the catalog.
The
date and time of the NEWEST SMF CATALOG RECORD SELECTED for this catalog
should usually be the same as that of the NEWEST SMF CATALOG RECORD
FOUND unless intermediate record collection is being done.The date and time of each should precede the specified recovery stop time. Even though the recovery stop time is adjusted by the multi-system clock difference, the time needed to confirm that the catalog was closed for recovery should be well after the last update to the catalog.
The
date and time of the NEWEST SMF CATALOG RECORD FOUND for this catalog
should usually be the same as that of the NEWEST SMF CATALOG RECORD
SELECTED, unless intermediate record collection is being done. You
would normally want the most current data.
Note that the remainder
of the report concerns SMF records of all types, not just the SMF
catalog records.
The
date and time of the OLDEST SMF RECORD FOUND should normally precede
the specified recovery start time. If it does not, older SMF data
has probably been omitted.
The
date and time of the NEWEST SMF RECORD FOUND should normally be later
than the specified recovery stop time, otherwise the most recent SMF
data has probably been omitted. The next several lines of the report summarize the events that you should investigate for omitted or lost SMF data.
This count
includes all systems from which SMF data was read, not just those
which updated the catalogs named above.
Based on
the experience of your installation, determine whether this count
is reasonable considering the number of systems and the period covered
by the SMF data.
This
count is not very interesting on the "all systems" report. However,
on an individual system report, it tells you how many times the SMF
recording data set was switched, and therefore gives you some idea
of whether all appropriate SMF data sets have been included. In addition,
in the log data set you will find a message for each switch, CRU023I.
The time difference between the switch records tells you how often
the SMF data sets are filling up (or being switched). Based on this
value, you can determine whether the SMF gap time you specified is
sufficiently small to detect the absence of an SMF input data set.
Again,
this value is primarily of interest for a single system. The intention
is that you should be able to spot periods of deliberate system inactivity
and thereby resolve corresponding gaps. Look for message CRU022I in
the log. Normally it will precede the message for the next event.
This
value too is primarily of interest for a single system. The IPL may
be a planned one or it may be the result of an interruption. Look
for message CRU021I in the log. The SMF record for a planned IPL will
normally be preceded by a record for EOD. The accompanying gap in
SMF recording (if any) is thus explained and can be disregarded.If no EOD record precedes the IPL record, there is a good chance that there has been a system interruption and SMF records may have been lost. If the time between the interruption and the IPL exceeds the gap time, an accompanying gap message should be expected. Otherwise, this IPL record may be the only indication of lost SMF data. Consequently, each IPL should be understood, referring to your problem management records and to the system log as necessary to establish the time of any accompanying interruption.
The
next line gives the count of SMF LOST DATA RECORDS FOUND. This value
is primarily of interest for a single system. Look for message CRU202I
for that system in the log. The dumped record accompanying that message
contains the beginning date and time of the resulting gap in SMF records
and also the number of records lost. You will normally want to proceed
with the ICFRU, using the SMF data you do have.
Any gaps that appear should be resolved from the single system reports. A FORWARD GAP condition is recognized when the date and time for the current record is later than the previous record for that system by an interval longer than the gap time you specified. Each such event is signaled by message CRU201I in the log. It is most frequently not associated with a lost data condition, but rather with a normal or planned period of system inactivity. If accompanied by a normal EOD and IPL sequence or if the interval is during a time when this system is normally inactive, the gap is accounted for an may be ignored. If the gap is not accounted for in this way, look for omissions in the SMF input data. Even if all input data is present, it may not have been presented in sequence by data set. In this case, there should be an accompanying BACKWARD GAP (see below). If the gap cannot be explained by any of the previous reasons, you should assume that SMF data has been lost. You will normally want to proceed with the ICFRU, using the SMF data you do have.
A BACKWARD GAP condition is recognized when the date and time for the current record is earlier than the previous record for that system by an interval longer than the gap time you specified. The event is signaled by message CRU200I in the log. It typically results from concatenating newer SMF dump data sets in front of older ones. Thus the principal use of this count and its companion message is the resolution of forward gaps. It could also be caused by reusing a partially filled SMF recording data set before it has been dumped (or emptied). Thus it might accompany a system interruption, which resulted in the use of an SMF recording data set other than the one active at the time of interruption. In this last case you might expect to find an IPL record after the backward gap. |
Copyright IBM Corporation 1990, 2014
|