HP 3000 Manuals HP 3000 Manuals
Other Archiving Techniques
You can use Performance Collection Software log file data in several
ways. Shortly after it is logged, you can use it to diagnose short-term
problems on your system (or systems). To do this, you must be able to
access the latest log file data quickly and in full detail.
If you have direct data communication links between your PC and the host
system, the fastest way to access log file data is by opening a raw log
file directly as a remote log file. This lets you see data that is no
more than 5 minutes old.
If you do not have direct data communications between your PC and the
host system, you can use the EXTRACT program to extract data from the raw
log files for transmission to the PC or to a host that the PC can access
directly. This extraction can be performed on demand either when needed
or periodically in anticipation of a need.
For example, you may need to look at a host system when time is not
critical. To handle questions once a week, you only have to examine the
previous week's data for each system. In this case, you can schedule a
weekly data extraction on each system and transfer the extracted log file
to a central site--a PC or the host system.
Case 1
You want to have the previous week's data from each host system at a
central site. This will help you minimize the demands of making
extractions during prime time shift operations and to give you rapid
access to data. Data from the previous week is satisfactory for analysis
since it will not be affected by aging.
Technique.
You want to retain each system's weekly log files. To make this
manageable, set up each remote system in a different group on the central
system. Run the remote job stream late Sunday night to take advantage of
the WEEKLY extraction command.
A typical job stream is as follows:
:JOB SUNDAY,SCOPE.SYS,SCOPE
:RUN EXTRACT
GLOBAL BOTH
APPLICATION BOTH
PROCESS ON
DISC ON
WEEKLY
EXIT
:DSLINE CENTRAL
:REMOTE HELLO ARCHIVE.LOGFILES,BERT
:DSCOPY RXWEyyww TO ,CENTRAL;MOVE;REP
:REMOTE BYE
:EOJ
Now you can access any week on the BERT system by opening the appropriate
log file. (The third week of 1991 for the BERT system is in a file
called RXWE9103.BERT.ARCHIVE.)
Case 2
Case 2 is the same as case 1, except in case 2 you will overwrite each
system's log files weekly.
Technique.
Construct a batch job to run on each host system at the end of the week
when prime time shift operations have finished. Have the job extract the
data for the previous week and transmit it to the central site for
access.
The new batch job would
be:
:JOB SUNDAY,SCOPE.SYS,SCOPE
:RUN EXTRACT;PARM=7
GLOBAL BOTH
APPLICATION BOTH
PROCESS ON
DISC ON
OUTPUT RXBERT
EXTRACT
EXIT
:REMOTE HELLO ARCHIVE.LOGFILES;DSLINE=CENTRAL
:DSCOPY RXBERT TO ,CENTRAL;MOVE;REP
:REMOTE BYE
:EOJ
TIP You can save disc space and reduce transmission time at the expense
of 5-minute data points on the Performance Collection Software
graphs by specifying global summaries and application summaries to
the EXTRACT program. You can save more disc space and transmission
time by not extracting process data, although this data could be
useful for diagnostic work.
At the central site, you can analyze this system by opening the remote
log file RXBERT.ARCHIVE.LOGFILES, if central-site security allows the
SCOPE.SYS user to read this file.
Case 3
Each host system must ensure that data is not lost even though the raw
log files are busy during the system's daily backups. This allows
SCOPE(XL) to run continuously and be available to analyze any performance
problems, even those occurring during backup.
Technique.
Perform a monthly extraction every day immediately before backup. (The
MONTHLY extract command creates a single log file for each month, but
appends new data to that log file whenever the command is executed.)
Perform the following commands immediately before system backup:
:RUN EXTRACT
GLOBAL DETAIL
APPLICATION DETAIL
PROCESS ON
DISC ON
MONTHLY
EXIT
Remember that the extracted files are generated in the LOGON group.
The file created--RXMOyymm, where yymm is the year and month--is free to
be written to the backup tape. Each month a new file is created with a
higher numerical designation in its name. After a new monthly log file
is created, the previous month's log file can be transferred to tape and
the disk purged to recover disc space. If the stored data is ever
needed, the file can be restored, and data can be extracted from it using
the EXTRACT program.
Case 4
You want to determine long-term trends in a system's data. This allows
you to track application CPU utilization over periods of a year or
longer, or to make effective use of a forecasting tool such as HP
RXForecast. In this case, you will not need detailed data, and the
diagnostic nature of the process data is not worth the disc space it
would require.
Technique.
Save global and application summary data and disc detail data since it
requires little additional disc space and it will be available for later
examination. Stream the job at least once a month (more often if the
size of your raw log files will hold less than two months of raw global
and application data).
The following is a typical job stream:
:JOB SUMMARY,SCOPE.SYS,SCOPE
:RUN EXTRACT
GLOBAL SUMMARY
APPLICATION SUMMARY
PROCESS OFF
DISC ON
YEARLY
EXIT
:EOJ
You can access the remote log file RXYRyyyy (where yyyy is the year
desired) or transfer this file to your PC for local access. Usually, the
level of summarization you select will store a year of data in about 5
megabytes of disc space, or less (consistent with the program's other
space requirements).