Running a job at a specific time
During a session on your UNIX system, most of the programs you
run will be executed immediately. Sometimes this will not be
desirable. If a program is likely to take a long time to run, you
may wish it to run overnight since, if a machine tries to execute
too many processes at once, it becomes slow, to the detriment of
all logged-in users. The command at can be used to
schedule a job for a specific time in the future.
In its simplest form, typing at followed by a time
and/or date will cause the standard input to be read, and those
commands entered as input to be executed at that time, thus:
$ at 1530
at> echo "It's half past three"
ctrl-D
job 81 at 2001-12-07 15:30
The number of the job, together with confirmation of the date
and time it has been scheduled, will be printed on the terminal.
The output (both standard output and standard error) will be mailed
to you. Alternatively, you can create a file
(testfile, say) containing commands you wish to be
executed, and then you can use at with option
-f ('file'):
$ at -f testfile 1530
job 81 at 2001-12-07 15:30
The formats allowed for you to specify the time are broad and
easy to use, but rather complex to specify - look in the manual
page for at to check the exact syntax allowed. The
following examples will give the general idea:
1645
16:45
16:45 GMT tomorrow
noon
4am Jan 25 2001
11pm today
now + 30 minutes
now + 1 month
If you take care that your time/date specification is
unambiguous, it will probably be acceptable to at.
When a job has been created using at, it is placed
on a queue. At the specified time and date,
or as soon thereafter as the load on the system permits, an
invocation of the shell will execute the commands given to
at. The jobs on the queue can be examined with option
-l ('list') and jobs on the queue can be removed with
option -r ('remove') by at:
$ at -l
81 2001-12-07 15:30 a chris
$ at -r 81
$ at -l
$
The fourth column of the listing of jobs contains an
a indicating that the queue is named a
('at queue'). You should not normally need to be concerned about
which queue a job has been placed on.
Similar to at is batch (which is
actually the same as batch is the same as the command
at -q b -m now)) This command is
used when you do not wish to specify exactly when a job should run,
merely that the system load should not be high when you do it. A
job submitted with batch is dealt with by the system
in exactly the same way as at, except that
batch will instruct the time of running to be
now and will place the job on a separate queue. Jobs
submitted to this batch queue can be listed
and removed using at:
$ batch <testfile
job 121 at 2001-12-07 15:45
$ at -l
121 2001-12-07 15:45 b chris
Notice that the name of the queue is b ('batch
queue')
The 'jobs' described in connection with at and
batch should not be confused with the 'jobs' in the
section on job control. Jobs in an at-queue can only
be created and removed by batch or
at.
Worked example
Write a script to list all the files in your filespace, and run
that script in one minute's time.
Solution: Using man ls you will
discover that option -R ('recursive') to
ls will list all files in the current directory and
recursively through all subdirectories. So we can pass ls
-R as the command to be processed by at. First
of all, however, we must change directory to the home
directory.
$ at now + 1 minute
cd
ls -R
ctrl-D
|