[GE globus] Grid Engine and GT4

David McBride dwm at doc.ic.ac.uk
Fri Jun 16 10:22:23 BST 2006


    [ The following text is in the "X-UNKNOWN" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some characters may be displayed incorrectly. ]

On Thu, 2006-06-15 at 17:53 -0400, Diego Bello wrote:
> With the gt4 integration package, I get this when starting container:

> java.io.IOException: java.io.IOException:
> /opt/globus4/libexec/globus-scheduler-provider-sge: not found

Hi,

The LeSC GT4-SGE adaptor never shipped with an MDS information provider,
which results in the above error.  

This will only affect you if you're trying to run an MDS service; it
doesn't affect GRAM or WS-GRAM.

I'm surprised that the GT4 JobManager doesn't work with when invoked by
the GRAM backend; I didn't think the changes between the two releases
were that significant.

The error you're getting corresponds to
Globus::GRAM::Error::TEMP_SCRIPT_FILE_FAILED -- which is returned if the
JobManager is unable to create the temporary script file for some
reason.

Unfortunately, by default Globus doesn't seem to write out the logging
output generated by the JobManager anywhere.  Rewriting the new()
function in the file JobManager.pm to read as follows will force log
files to be created in /tmp:

$GLOBUS_LOCATION/lib/perl/Globus/GRAM/JobManager.pm:

--------8<---------------------8<---------------------8<-------------

sub new
{
    my $class = shift;
    my $self = {};
    my $description = shift;

    $self->{JobDescription} = $description;

# [dwm] Begin changes
#
#    if(defined($description->logfile()))
#    {

    my $WHOAMI = `whoami`;
    chomp $WHOAMI;

        local(*FH);
        open(FH, '>>'. "/tmp/$WHOAMI-jobmanager.log");
        select((select(FH),$|=1)[$[]);
        $self->{log} = *FH;
#    }
# [dwm] End changes

    bless $self, $class;

    $self->log("New Perl JobManager created.");

    return $self;
}

--------8<---------------------8<---------------------8<-------------

If you're still can't work out what's going wrong, the JobManager perl
script itself will have been installed in
$GLOBUS_LOCATION/lib/perl/Globus/GRAM/JobManager/sge.pm -- if you know
any Perl at all, try adding additional logging statements to the
submit() function to try to narrow down the problem.

I'm afraid I haven't done any work with SGE and GT4 since the adapator
was released, so I probably can't be of much more help;  however, it's
possible that you might find my latest internal revision (which I've
been using with my GT2-based LCG grid work) useful -- see:

 http://www.doc.ic.ac.uk/~dwm/Code/gnu-arch/2006/sge-jobmanager--main/
 http://www.doc.ic.ac.uk/~dwm/arch/2006/ (GNU Arch Repository)

Best of luck.

Cheers,
David
-- 
David McBride <dwm at doc.ic.ac.uk>
Department of Computing, Imperial College, London


    [ Part 2, "This is a digitally signed message part" ]
    [ Application/PGP-SIGNATURE (Name: "signature.asc") 196 bytes. ]
    [ Unable to print this part. ]



More information about the gridengine-users mailing list