[GE users] debugging tight integration

fx d.love at liverpool.ac.uk
Fri Dec 18 12:16:16 GMT 2009

    [ The following text is in the "utf-8" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some characters may be displayed incorrectly. ]

pollinger <harald.pollinger at sun.com> writes:

> So the process chain from the sge_execd to the qrsh_starter is fine, but 
> the job itself (gamess.64.x) is not a child of the qrsh_starter, but a 
> child of init. And I'm missing a shell at the end of the process chain. 
> Did you specify the "-shell no" option to qrsh?

No.  The way it works is that gamess uses its own communication stuff
unfortunately (not MPI) and you can specify the rsh command it uses,
which I set to the SGE mpi/rsh wrapper.

> It seems either the job script exited/died or gamess daemonized itself. 
> But then I'm wondering why the qrsh_starter doesn't quit.
> You could replace gamess by a script like this:
> #!/bin/sh
> echo "starting"
> sleep 100
> echo "done"
> exit 0
> and start it with exactly the same command line. If it works fine and is 
> a child (or a child of a child) of the qrsh_starter, gamess itself does 
> something wrong.

I think that's where the problem is.  Reuti has actually spotted
something odd about the executable, which we'll try to investigate.
I'll write this up if it looks as though there's something useful for
others if/when we've sorted it out.  Thanks.

(Dr) Dave Love
?E-Science?, Computing Services Department, University of Liverpool
AKA fx at gnu.org


To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

More information about the gridengine-users mailing list