Parallel Batch Processing

[keziah]

Hello,

I want to do start some jobs in parallel in my script (bash) and return an error code after the end of all process. I want return an error code set to 1 if one of them failed.

How can i retrieve the error code of each process ?
Because with the wait command i've only the error code of the last process if i do:
wait $pid1 $pid2 $pidx

Some idea ?

Thanks for help

 

[PHV]

Can your jobs write some status in sort of log files you can then analyse ?

Hope This Help, PH.
Want to get great answers to your Tek-Tips questions? Have a look at FAQ219-2884

 

[Salem]

Here is a reaper which will wait for all child processes

/* compile this as 
 * gcc -o reaper reaper.c
 */
#include <stdio.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <unistd.h>

int main( void ) {
    int     child_status;
    pid_t   wait_status;
    int     final_status = 0;

    /* wait for all children to exit */
    printf( "Waiting for all children to exit\n" );
    while ( (wait_status = waitpid(-1,&child_status,0)) > 0 ) {
        /* did the child exit normally? */
        if ( WIFEXITED(child_status) ) {
            if ( final_status == 0 ) {
                /* update with the exit status of this child */
                /* but only if all the previous children passed OK */
                final_status = WEXITSTATUS(child_status);
            }
        } else {
            /* a child failed horribly */
            final_status = 2;
        }
        sleep( 1 ); /* take a nap, to allow more children to finish */
    }
    printf("All done, final status is %d\n", final_status );
    return final_status;
}

The shell script looks something like this

#!/bin/sh
./worker &
./worker &
./worker &
./worker &
./worker &
exec ./reaper

The exec at the end is very important. It replaces the shell, and in doing so, the reaper adopts all the children, and can wait for them properly.

--

 

[aigles]

You can write status of each job in a file and anlyze the result file :
[tt]
rm -f status_file
( ./job1 ; echo $? >> status_file) &
( ./job2 ; echo $? >> status_file) &
( ./job3 ; echo $? >> status_file) &

wait # all child processes

grep -c -v 0 status_file | read errors
if [ "$errors" -ne 0 ] ; then
echo "At least one job failed."
end
[/tt]


Jean Pierre.

 

[keziah]

Thanks all for the answer

The C code is cool but i think the solution of Jean-Pierre will is what i want

Regards,