[GE users] SGE6.0u1 and user hold and -sync issue

Andreas Haas Andreas.Haas at Sun.COM
Tue Jan 25 13:59:32 GMT 2005

On Mon, 24 Jan 2005, Wilfried Gaensheimer wrote:

> Andreas Haas wrote:
> Andreas,
> ...
> >>I can live with the current situation (our submit wrapper reads the qsub
> >>-sync output and does the right thing).
> >
> >
> > I take your statement as a confirmation users perceive dual use
> > of qsub exit status a useful interface ;-)
> >
> > Thus I believe it is valid to make a distinction with qsub -sync y
> > exit status:
> >
> > - with sequential jobs dual use of qsub exit is useful as
> >   long as qsub -sync y output allows to differ between
> >   job error condition and exit status
> What's your definition of "dual use"? I hope that means the exit code
> will reflect both qsub and batch job exit status information.

Dual use refers to status information indicating both
whether the job actually started and job return code in
case it was started.

> > - with array jobs dual use of qsub exit isn't useful. If
> >   the job exit status is of interest qacct anyways has to be
> >   used for checking each tasks exit status
> My preferred solution would be to get a non-zero return code with
> the -sync option, if any of the job fails. If a user wants to ignore
> that, he can do. For usage in Makefiles it's important to get the
> "effective" return code, which includes the whole chain of qsub + job.

I believe we're actually in agreement on the desired behaviour:
A non-zero exit status will be returned if either the job couldn't
be started or if the job itself returned non-zero exit status.

I filed two issues


for the qrls problem and the qsub -sync bug. Note the workaround
with qrls.

Thanks a lot for reporting those issues!


To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net

More information about the gridengine-users mailing list