I work with programs that interact across different servers, and recently one of them has been leaving behind processes after exiting in certain ways (that is to say, if I've used the wrong input for it and had to close manually). This is on a server used by several people, though, and I'm never really sure when I've left something behind. Also, closing things by hand was difficult, as I had to check several nodes for processes.

That's the intro - I've written a script to go to each node, find the right processes, make sure they're not supposed to be there, and then kill the offenders. Thing is, I'm not so good with perl, so I figured I'd put this up to see if anyone has critiques so I can learn.

#!/bin/perl
use strict; use warnings; 

# Known problems: 
# -Issues for runs started around the new year
# -Broken runs started at the same time as long 
# good runs will continue as long as the good run
# if the 'ps' output changes order, than may cancel the wrong process 

# Get the currently in-process run times - grep is to specify only act
+ual run lines
my $rundate = `/opt/lsdyna/License/lstc_qrun | grep @ | cut -c 57-68`;
#cleandate subroutine is at the bottom of the file
&cleandate($rundate);
#splitting the dates into an array - allows easier 'foreach' loops
my @rundates = split(/\s+/,"$rundate");

my @nodes = ("node2", "node3", "node4", "node5", "node6", "node7", "no
+de8", "node9", "node10");
#this main, big loop allows each node to be checked
foreach my $nodenum (@nodes) {
    print "Testing $nodenum: ";

    #pull the mpp processes from the node, then take and clean process
+ ids
    my $procid = `ssh $nodenum ps -eo pid,lstart,cmd | grep mpp | cut 
+-c 1-5`;
    $procid =~ s/ //g;
    my @procids = split(/\s+/,"$procid");
    # print "@procids\n";
    
    #These two are purely for output, make sure that it's running
    my $procnum = scalar(@procids);
    print "$procnum processes found.\n";
    
    #pull the mpp processes from the node, then take and clean process
+ dates
    my $procdate = `ssh $nodenum ps -eo pid,lstart,cmd | grep mpp | cu
+t -c 11-22`;
    &cleandate($procdate);
    my @procdates = split(/\s+/,"$procdate");
    
    #This is the comparison loop; each of the dates in @procdates is c
+ompared to
    #each of the dates in @rundates. If all of the comparisons are mor
+e than an 
    #hour apart (tested through $counter), then the process is killed.
+ $procloop 
    #is just an index, so that the corresponding part of $procid is ki
+lled.
    my $procloop = 0; my $counter = 0;
    foreach my $pd (@procdates) {
        $counter = 0;
        foreach my $rd (@rundates) {
            my $timedif = abs($pd-$rd);
            if ($timedif >= 60) { $counter++ }
        }
        if ($counter == scalar(@rundates)) { 
            print "$procids[$procloop] is outside expected time range.
+\n" ;
            my $cmd = "ssh $nodenum kill -9 $procids[$procloop]";
            #print is to see if anything is killed, and make sure comm
+and syntax is correct
            print "$cmd\n";
            system ($cmd);
        }
        $procloop++;
    }
}

sub cleandate {
    #This takes the Mon  DD HR:MN format and makes it into a numeric v
+alue for comparison
    #first line adds leading 0 to single-digit dates
    $_[0] =~ s/\s(\d\s)/0$1/g;
    #change month to a numeric value
    $_[0] =~ s/Jan/01/g;
    $_[0] =~ s/Feb/02/g;
    $_[0] =~ s/Mar/03/g;
    $_[0] =~ s/Apr/04/g;
    $_[0] =~ s/May/05/g;
    $_[0] =~ s/Jun/06/g;
    $_[0] =~ s/Jul/07/g;
    $_[0] =~ s/Aug/08/g;
    $_[0] =~ s/Sep/09/g;
    $_[0] =~ s/Oct/10/g;
    $_[0] =~ s/Nov/11/g;
    $_[0] =~ s/Dec/12/g;
    #remove colons
    $_[0] =~ s/://g;
    #remove spaces
    $_[0] =~ s/ //g;
    #page breaks from grepping are left, to allow splitting above.
}
[download]