Take a look at the following nodes:

• Maximum # of concurrent runs (lots of discussion)
• how to limit child processes? (links in discussion)
• Many children, but never more than a fixed number at once. (code snippet)
• A framework for implementing child processes performing file updates, SQL queries, and the passing of results to the parent using Parallel::ForkManager (code snippet)

Update: Added 2 more links

One way is to maintain a counter file that is checked every time a script is run. If the counter file holds less than five, go ahead and increment it by one and continue execution. At the end of the script, decrement the counter file by one.

This can get messed up if people kill-9 one of the scripts.

One way to deal with the possibility of active processes being killed, thus fouling up the count file, is to use timestamp entries in the countfile. If you know that there is no way your script should take longer than 1 minute to run, and upon script startup, you find an entry in the file that is ten minutes old, remove that entry; it's a ghost.

There is the problem of variables shared between all your scripts. Problem? No, in this particular case, you can make good use of it, and use one common variable to count how many copies are running. Each instance of your script should increment that variable, and decrement it when it exits. If there are too many already, exit immediately.

#!/usr/bin/perl

use Proc::Queue size=>5, ':all';

my @script_names= qw(s1 s2 s3 s4 s5 somescript);

my @pids;
for my $s (@script_names) {
push @pids, system_back($s)
}

%results=waitpids(@pids);
[download]