I need to run a programmme, let say ./a for 10 times in the linux shell. What is the command to do that?
Also, how could I differentiate the output from different processes?
Besides, my ./a is a continual process. How could I terminate it?
Thanks.
+3
A:
Make a loop that starts the program 10 times in the background:
for i in $(seq 1 10) ; do
./a &
done
Also, how could I differentiate the output from different processes?
You can't easily do that, the output can be interwined. Redirect the output of each process to a different file rather, e.g.
for i in $(seq 1 10) ; do
./a >output.$i &
done
As a oneliner, you'd run
for i in $(seq 1 10) ; do ./a >output.$i & done
to kill those processes later:
for i in $(seq 1 10) ; do kill %$i ; done
nos
2010-10-11 07:55:45
doh. you were first.
naugtur
2010-10-11 07:56:36
in bash, you can use {1..10} instead of calling external command
ghostdog74
2010-10-11 09:34:10
+3
A:
In bash you can do:
for ((i=1;i<=10;i++)); do
# run the command ./a and redirect its output to a file named after
# its sequence number...like out1.txt, out2.txt...
./a > out$i.txt &
done
The key here is the & which runs the command in the background. Without the & the 2nd ./a will be started after the fist got completed, effectively making it serial execution. But with &, we don't wait for one invocation of ./a to complete before we launch another one and that is how we achieve parallelism.
codaddict
2010-10-11 07:56:52
Thanks for the reply. My ./a is a continual process. How could I terminate it? It seems like ctrl C does not terminate all the process
Steven
2010-10-11 08:10:51
jweyrich
2010-10-11 08:17:07
If you want to be able to kill the processes on ctrl+c, then use the $! variable to obtain the pids, and then a subsequent loop with pgrep to watch them until they complete, and a trap with pkill in a loop to kill the individual processes. Here's some useful links in case you need them: http://tldp.org/LDP/Bash-Beginners-Guide/html/sect_12_02.html http://www.tutorialspoint.com/unix/unix-special-variables.htm
supercheetah
2010-10-11 09:12:53
You can view all the background jobs in a shell with the `jobs` command. You can bring one of them to the foreground, by running `%4` , 4 is the job number here. You can kill that job by running `kill %4` . To make the current foreground job into a background job, hit CTRL+Z, and run `bg` (that is, if you ran the loop posted here directly in the shell, not in a separate shell script)
nos
2010-10-11 21:32:52
+1
A:
for i in {1..10}
do
./a > "out_${i}.txt" &
done
to terminate a, you can use pkill a
ghostdog74
2010-10-11 08:15:28
@gameover: it does exactly the same, but also accepts the process name as argument (or even a regex).
jweyrich
2010-10-11 08:40:30
And by default `pkill` does a substring match on the process name, so literally using `pkill a` is likely to kill a lot of processes.
al
2010-10-11 09:21:59
of course i am not with the idea of naming the program "a" to begin with.
ghostdog74
2010-10-11 09:33:49
A:
Assuming you use Bash or compatible shell, the special variable $! holds the process ID of the last process you made asynchronous. Augumenting nos' answer, I'd suggest using
for i in $(seq 1 10) ; do
./a 1> stdout.$i 2> stderr.$i &
echo $! > pid.$i
done
Now you have error output in stdout.X, error output in stderr.X and the process ID in pid.X so you can kill it with
kill `cat pid.X`
or even kill all processes with
for i in $(seq 1 10) ; do
kill `cat pid.$i`
done
Christoffer
2010-10-11 08:22:01
A:
All of the other answers are fine so far, but you amy also want to try repeat, if supported by your system:
repeat 10 ./a
Here to differentiate outputs, unfortunately, you won't be able to use a counter variable. You can however use a timestamp:
repeat 10 ./a >> output_`date +%N`
That being said, the behavior of date for anything below the second mark is unspecified, so that's not the most reliable way. A bit easier than to write a for loop though.
haylem
2010-10-11 08:32:45