Home / ubuntu / Questions / 1259111
Accepted
An Ignorant Wanderer
Asked: 2020-07-15 13:03:55 +0800 CST

Get process ids of all spawned processes from xargs

  • 772

In my script, I have the following: cat list | xargs -P6 -n2 ./first_queue & where list is just a file with pairs of words that I want to process:

word1 word2

word3 word4

etc...

first_queue is just another script. I am trying to get the pid of all 6 processes that are spawned by this xargs right after I call it so that I can later terminate them. I read that $!gives the pid of the last process that ran in the background. However, I want all 6 of them. How can I get them?

More details:

I wrote the following small script to test what @xenoid suggested:

cat list | xargs -P6 -n2 ./first_queue &
id=$(echo $!)
echo $id
ids=$(pgrep -P $id)
echo $ids
ps aux | grep $id
for x in $ids; do
    echo $x
    ps aux | grep $x
    kill $x
done
kill $id
ps aux | grep $id
for x in $ids; do
    ps aux | grep $x
done

Here is first_queue extremely simplified (to the command that is giving me the trouble, and that keeps running despite running kill on the parent process i.e xargs as well as the child processes):

srr=$1
bioproject=$2

prefetch $srr -O download_dir/$bioproject

prefetch just downloads data from an online database.

bash process

1 Answers

  1. Best Answer

    The PID you get with $! is xargs's own PID (because this is xargs that you are sending to the background...), not the one of your processes. If you terminate it it will terminate all its children (i.e., any of your processes still running).

    If you want the pids of the child processes you can use pgrep -P {parent} where {parent} is xargs PID.

    Edit:

    When you run a command (pid=Y) in a script (pid=X), and send a signal to the script (to pid=X), the signal is not forwarded to the command (so pid=Y doesn't see it).

    You can write a contrived script that starts the command in the background, grabs its pid (Y), traps signals send to pid=X and forwards them to pid=Y.

    But in many cases, the command is the last significant bit of code to run (the script exits with the command's return code). When this is so instead of using:

    some_initialization_code
the_command and the command args
exit $?

    you write your script as:

    some_initialization_code
exec the_command and the command args

    In the first case the script is the process with pid=X and the command is the process with pid=Y. With exec, process with pid=X becomes the command, and everything happens as if you called the_command directly: there is no pid=Y, and signals sent to process X (the script) are received by the the_command when it runs.

    • 0