sa(8)sa(8)NAMEsa - Summarizes accounting records
SYNOPSIS
/usr/sbin/sa [-abcdDfijkKlmnorstu] [-v number] [-S savefile] [-U user‐
file] [filename]
OPTIONS
Outputs all command names (including those containing unprintable char‐
acters and commands used only once) in the last column. In the default
format, such commands are summed and the total is written as the entry
***other. Sorts cpu output column 3 according to the sum of user and
system CPU time divided by the amount of CPU time required to execute
the command entered in the last column (6) as many times as is entered
in the first column (1). Adds three percentage columns to the default
format to list percentages as follows: Lists the percentage of the num‐
ber of times each command was executed with respect to the total number
of times all commands were executed (see 1a below). Lists the percent‐
age of the amount of real time required to execute each command the
number of times entered in the first column with respect to the total
real time required to execute the total of all commands entered in the
last column (see 2a below). Lists the percentage of the amount of com‐
mand CPU time required to execute each command the number of times
entered in the first column with respect to the total CPU time required
to execute the total of all commands entered in the last column (see 3a
below). Sorts avio output column (4) in descending order according to
the average number of disk I/O operations. Substitutes tio column (4a)
for the avio (4) column and sorts tio output column 4a in descending
order according to the total number of disk I/O operations. Used with
the -v option to inhibit interactive threshold comparison of commands.
Reads raw database file /var/adm/pacct only. Does not include records
from summary database file /var/adm/savacct. Outputs the average num‐
ber of seconds per command in default columns 2, 3, and 4 instead of
the total time in minutes for the number of calls entered in column 1
for each command. Sorts and outputs records according to the value in
the k output column 5 in descending order. Substitutes k*sec column
(5a) for the k (5) column and sorts the k*sec output column in descend‐
ing order according to the value of the memory time integral. Sepa‐
rates cpu column 3 into two columns. The new column entries are column
3a, which lists the s (system) part of the CPU minutes, and column 3b,
which lists the u (user) part of the CPU minutes. Outputs a 5-column
file, which provides the information in the following table. Listed
below in left-to-right order are the column identification suffixes, or
none when no suffix is used, and the purpose of the column. Some col‐
umns are identical to the default output format described in the table
in the DESCRIPTION section; these are marked with an asterisk (*).
Username or user ID as written in the /etc/passwd file. The total num‐
ber of processes executed by the user during the accounting period.
Same as column 3 in the default output file. Same as column 4a in the
default output file. Same as column 5a in the default output file.
Outputs the default format sorted in descending order according to the
number of times each command was called. Substitutes, in the default
output format, the ratio of user CPU time (u) to system CPU time (s) as
u/s in column 3 in place of the total user and system CPU time (cpu)
for the number of calls entered in the first column. The default for‐
mat, described under Description, is resorted in ascending order
according to the values entered in column 3, cpu time. This sort is the
reverse of the default sort. Merges information in accounting database
file /var/adm/pacct with summary files you specify with the -U and -S
options, or merges the database file information with information in
default files /var/adm/usracct or /var/adm/savacct. After the merge,
database file /var/adm/pacct is truncated. The use of this option also
implies the use of the -a option. Uses savefile as the command summary
file in place of file /var/adm/savacct. Adds the re/cp column (3d) to
the default format. Entries in this column express the ratio of real
time to total (cpu) time, which is the sum of user and system time for
each command entered in the last column. Suspends all other options
and prints the user numeric ID, the CPU time, memory usage, number of
I/O operations, and the command name for each command. Uses userfile
as the user summary file in place of file /var/adm/usracct to record
per-user statistics output with the -m option. Prints, as a query, the
name of each command used number times or fewer to the standard output
as follows:
command--
where command-- is the name of the command written to the stan‐
dard output by sa.
When you respond by typing y to the standard input, the command
record is omitted from a default-formatted list at the end of
the interactive command queries written to the standard output.
The columnar values of the omitted record are totaled in an
added record whose command name is **junk** in the last column
of that list. When you type any other character, the record for
the queried command name remains in the default output list at
the end of the interactive commands written to the standard
input.
OPERANDS
Process accounting file name. If this optional operand is absent, the
sa command uses /var/adm/pacct as the default process accounting file.
DESCRIPTION
The sa command helps you manage the large volume of accounting informa‐
tion that is generated each day when system accounting has been enabled
by the system administrator or by the superuser.
When you use the -s option with the sa command, the information in
/var/adm/pacct is condensed into summary file /var/adm/savacct, which
contains a count of the number of times each command was called and the
amount of time system resources were used.
Condensed information for each user is stored in /var/adm/usracct.
This condensed-information file conserves storage space because on a
large system the /var/adm/pacct daily process file can grow by as many
as 100 blocks per day. Summary files are normally read before account‐
ing files are, so that files produced by sa include all available
information.
When a file name is given as the last argument, the named file is
treated as the process accounting file. The /var/adm/pacct file is the
default process accounting file.
When the sa command is invoked with no options, the default output sum‐
mary is an unheaded 6-column file consisting of, in some cases, infor‐
mation having an identification suffix in the column. The identifica‐
tion suffix may be changed from the default (no options specified) out‐
put format by using various options.
The following table lists the columns with left-to-right reference col‐
umn numbers (not included in output) for the default format on the
left, the identification suffix for the entry when one is used (or none
when one is not in the middle), and the purpose of the information in
that column on the right.
Columns having more than one identification suffix description (2 and
2a, for example) use the alternate suffix designation in the same sa
output printout column for each of the listed alternate entries for the
column. For example, the second column has two possible suffix desig‐
nations: re and %. The re reference in the middle column describes the
information in the second column of the output printed by the sa com‐
mand when this suffix is used. Correspondingly, the % reference
describes the information in the sa output when the % suffix is used.
1 none The number of times the command entered in the last
column (6) was called.
1a % When the -c option is used, sa adds this column
after column 1 to list the number of times the com‐
mand was called (entered in column 1) as a percent‐
age of the total number of times all commands
entered in the last column were called.
2 re The number of real-time (elapsed) minutes required
to execute the command entered in the last column
(6) as many times as is entered in the first column
(1).
2a % When the -c option is used, sa also adds this col‐
umn after column 2 to list the amount of real time
(entered in column 2) required to process the com‐
mand entered in the last column (6) as many time as
is entered in the first column (1), as a percentage
of the total amount of real time required to
process all of the commands listed in the last col‐
umn.
3 cpu The number of CPU (user plus system) minutes used
to execute the command entered in the last column
(6) as many times as is entered in the first column
(1).
3a u The number of user CPU minutes used.
3b s The number of system CPU minutes used.
3c u/s When the -o option is used, substitutes u/s column
(3c) for the cpu (3) column and sorts the u/s out‐
put column in descending order according to the
ratio of user CPU time to system CPU time.
3d % When the -c option is used, sa also adds this col‐
umn after column 3 to list the amount of CPU time
(entered in column 3) required to process the com‐
mand, entered in the last column, the number of
times, entered in the first column, as a percentage
of the total CPU time required to process all of
the commands listed in the last column.
3e re/cpu When the -t option is used, adds the re/cpu column
to the default output format. Entries in this col‐
umn express the ratio of real CPU process time to
total CPU time (cpu), which includes user and sys‐
tem time. These entries appear after entries for
the cpu (3) column. The default output sort remains
unchanged.
4 avio The average number of input/output operations for
each listed command.
4a tio The total number of input/output operations for
each listed command.
5 k The average number of kiloblocks (blocks x 1024) of
memory used for each command process.
5a k*sec CPU storage-time integral in K-core seconds (sec‐
onds x 1024).
6 none The command name. A trailing asterisk (*) indicates
a forked program.
Other considerations for entries in the printed sa output are as fol‐
lows: All times are expressed to nearest one hundredth. The default
format is sorted in descending order according to the values entered in
column 3, cpu time. You should not share accounting files among nodes
in a distributed environment. Each node should have its own copy of
the various accounting files. When you are also using /usr/sbin/acct/*
accounting commands, do not delete accounting records in the
/var/adm/pacct process accounting source file because these records
also provide information for summary data files when the -s option is
used.
EXAMPLES
Use the sa command with the -a option to summarize accounting records
for all commands entered in the /var/adm/pacct process database file:
% sa-a
Commands used only once are summed with the entry ***other in
the last column of the default output format. Use the sa com‐
mand with the -k option to summarize accounting records accord‐
ing to the average number of kiloblocks of memory used for each
command:
% sa-k
FILES
Specifies the command path Default process accounting database file
Default system process accounting summary file Default user process
accounting summary file
SEE ALSO
Commands: acct(8), acctcms(8), acctcom(8), acctcon(8), acctmerg(8),
acctprc(8), fwtmp(8), runacct(8)sa(8)