# #ident "@(#)sa:README 1.8"
#ident "$Header: /proj/irix6.5.7m/isms/eoe/cmd/sa/RCS/README,v 1.4 1996/05/09 05:01:22 kanoj Exp $"
README 1.8 of 10/4/85
Instruction of daily report generation:
1. Enter
su sys -c "/usr/lib/sa/sadc /usr/adm/sa/sa`date +%d`"
in /etc/rc in state 2 under "$2" = 0, i.e.,
sadc is executed only for the first transition into state 2.
This entry causes a special record to be written on the
daily data file when UNIX restarts.
2. The execution of shell script
sa1 [t n]
causes data collection program to write system activity
counters n times at every t-second interval on data file.
If t and n are not specified, it writes data once.
According to your installation's requirement, enter
entries of executing sa1 [t n] in /usr/lib/crontab
to collect system activity data.
For example, entries
0 8-17 * * 1-5 su sys -c /usr/lib/sa/sa1 1200 3 &" and
0 18-23 * * 1-5 su sys -c "/usr/lib/sa/sa1 &"
cause data collection program to be activated at every hour
on the hour from 8:00 to 23:00 on weekdays.
Moreover, It writes data on data file 3 times at every 20
minutes interval from 8:00 to 17:00 and once at other times.
3. Shell procedure sa2 will invoke sar command to generate
the daily report from the data file. Its usage is
sa2 [-options] [-s hh:mm] [-e hh:mm] [-i ss]
where -s and -e specify the report starting and ending times
respectively, -i specifies the report data interval in seconds.
If they are not specified, all data from the data file are to
be reported. -options are report options, see manual page
sar.1 for description.
Make an entry to execute sa2 in /usr/lib/crontab.
For instance, entry
5 18 * * 1-5 su adm -c "/usr/lib/sa/sa2 -s 8:00 -e 18:01 -i 3600
-ubd &"
causes the invocation of sar command at 18:05. It generates
the daily report that includes the hourly cpu utilization,
buffer usage and disk and tape activities from 8:00 to 18:01.
Notes on sadc/sar record formats (from kanoj@sgi.com):
Till 6.2, sadc would write the foll. fields in the sa file that
sar reads:
1. A tblmap[] array, whose first element gave the # disks in
the system, which was assumed not to change ever.
2. An sa structure written at system startup time (specially
marked as a restart record).
3. A super record consisting of an sa structure followed by
the sgidio structure for each disk in the system. This
super records were written at each invocation of sadc
while the system was running. Since the number of disks
was assumed invariant on the system, this super record
size was also assumed invariant.
From ficus/kudzu, I changed this to take into account the
variability of disk configuration on systems. The sa file
now consists of:
1. An sa structure written at system startup time (specially
marked as a restart record). These sa structures contain
the number of disks in the system at the point of sadc
invocation.
2. A super record consisting of an sa structure followed by
the sgidio structure for each disk in the system at that
point.
