RE: [Condor-users] Possible Limitation Condor vs Perl / IPC::open3

["Dave Lajoie" <dlajoie@xxxxxxxxxxxxxxxxxxxx>]

Erik, Filip, John
Here are the files

Submit file
____________________________________________
universe = vanilla
executable = C:\Perl\bin\perl.exe
Requirements = (Has_XSI42_Batch == TRUE) && (Memory >= 1024)
Rank = (machine == "DAVELAJ_M60.PRODUCTION")
output = $(Cluster).$(Process)_out.txt
error = $(Cluster).$(Process)_err.txt
log = $(Cluster).$(Process).log
notification = always
notify_user = dlajoie@xxxxxxxxxxxxxxxxxxxx
copy_to_spool = false
should_transfer_files = false

priority = 10
initialdir = C:\temp
arguments = C:\Condor\wrapper\xsibatch42_wrapper.pl
Queue


Perl Wrapper
__________________________________
use Config;
use IPC::Open3;

use sigtrap 'handler' => \&myINT, 'INT';
use sigtrap 'handler' => \&myTERM, 'TERM';
use sigtrap 'handler' => \&myKILL, 'KILL';
use sigtrap 'handler' => \&myAlarm, 'ALRM';
use sigtrap 'handler' => \&myPIPE, 'PIPE';


sub myAlarm
{
}

sub myPIPE
{
    my $signame = shift;
    print "\n caught SIG$signame";
    exit(0);
}

sub myINT
{
    my $signame = shift;
    print "\n caught CONDOR::KILL \n";
    killChildrenProc();
    exit(0);
}

sub myKILL
{
    my $signame = shift;
    print "\n caught CONDOR::KILL \n";
    killChildrenProc();
    exit(0);
}

sub myTERM
{
    my $signame = shift;
    print "\n caught CONDOR::VACATE \n";
    killChildrenProc();
    exit(0);
}

sub killChildrenProc
{
	if ( $Config::Config{'osname'} eq "MSWin32")
	{
		# first we must kill the shell
		kill 9, $pid;
		open(TLIST,"tasklist 2>&1|");
		while(<TLIST>)
		{
			if ($_ =~ /^XSIBATCH.exe\s+(\d+)/)
			{
				push @tlist, $1;
			}
		}
		kill 9, @tlist;
	} elsif ( $Config::Config{'osname'} eq 'linux')
	{
		open(TLIST,"ps -eo pid,comm |");
		while(<TLIST>)
		{
			if ($_ =~ /(\d+)\s+XSIBATCH/)
			{
				push @tlist, $1;
			}
			sort @tlist;
			kill 9, @tlist;
		}	
	}
}

$cmd = 'C:\Softimage\XSI_4.2\Application\bin\xsibatch.bat -r -scene
"C:\data\customers\condor_tests\Scenes\condor_render_test_v02.scn" -mb
off -startframe 1 -endframe 10 -verbose prog -skip false';

$pid = open3("<&STDIN", \*XSIPROC, \*XSIPROC, $cmd) or die "Cannot run
$cmd: + $!";
	
for(;;)
{
	sleep 1;
	$line_count	= 0;
	while(<XSIPROC>)
	{
		if ($_ =~ /^\'ERROR/i)
		{
			print STDERR "WRAPPER::PROCESS_ERROR " . $_;
		} else
		{
			print STDOUT $_;
		}
		if ( $line_count++ > 10 )
		{
			last;
		}
	}
}


Here is a simpler test you can use to repro the problem

Submit file
______________________________________________________
universe = vanilla
executable = C:\Perl\bin\perl.exe
Requirements = (Has_XSI42_Batch == TRUE) && (Memory >= 1024)
Rank = (machine == "DAVELAJ_M60.PRODUCTION")
output = $(Cluster).$(Process)_out.txt
error = $(Cluster).$(Process)_err.txt
log = $(Cluster).$(Process).log
notification = always
notify_user = dlajoie@xxxxxxxxxxxxxxxxxxxx
copy_to_spool = false
should_transfer_files = false

priority = 10
initialdir = C:\temp
arguments = C:\Condor\wrapper\test.pl
Queue


Test.pl
________________________________________________________
$cmd = 'ping -n 10000 localhost |';
$pid = open(XSIPROC,$cmd) ;

# or die "Cannot run $cmd: + $!";

for(;;)
{
        sleep 1;
        $line_count     = 0;
        while(<XSIPROC>)
        {
                print $_;
                if ( $line_count++ > 10 )
                {
                        last;
                }
        }
        print "____\n";
}