I have a CI script that I want to speed up by running several things in the background. I want the script wait for all processes and check each one to see if it failed.
Here a a simplification:
#!/bin/bash
set -e
bg()
{
sleep .$[ ( $RANDOM % 10 ) + 1 ]s
}
bg2()
{
sleep .$[ ( $RANDOM % 10 ) + 1 ]s
exit 1
}
bg & # will pass after a random delay
bg2 & # will fail after a random delay
# I want the output of the program to be a failure since bg2 fails
Yes.
You can use the wait
command in bash
to wait for completion on one or more sub-processes to terminate in which case we provide the PID
to wait on it. Also wait
can optionally take no arguments in which case it waits for all background process to terminate.
Example:-
#!/bin/bash
sleep 3 &
wait "$!" # Feeding the non-zero process-id as argument to wait command.
# Can also be stored in a variable as pid=$(echo $!)
# Waits until the process 'sleep 3' is completed. Here the wait
# on a single process is done by capturing its process id
echo "I am waking up"
sleep 4 &
sleep 5 &
wait # Without specifying the id, just 'wait' waits until all jobs
# started on the background is complete.
# (or) simply
# wait < <(jobs -p) # To wait on all background jobs started with (job &)
echo "I woke up again"
Update:-
To identify the jobs when the fail, it is best to loop over the list of background jobs and log their exit-code for visibility. Thanks to wonderful suggestion by chepner. It goes like
#!/bin/bash
for p in $(jobs -p)
do
wait "$p" || { echo "job $p failed" >&2; exit; }
done
#!/bin/bash
set -e
bg()
{
sleep .$[ ( $RANDOM % 10 ) + 1 ]s
}
bg2()
{
sleep .$[ ( $RANDOM % 10 ) + 1 ]s
exit 1
}
export -f bg
export -f bg2
parallel ::: bg bg2 || echo $? of the jobs failed