Daniel Lorenz wrote:
> Hello,
>
> t hanks for all answers so far.
> As indicated by new tests, the jobs are aborted as they arrive a particular
> WN. They are executed if qsub'ed and aborted when submitted with
> globus-job-run.
>
> It seems not to be a problem to copy back results, nor low disk space or
> memory space.
> What could cause that jobs submitted with Globus are aborted and those
> submutted with qsub not?.
Did you check the environment a pool account gets on that WN?
Compare with another WN.
Maybe X509_USER_PROXY gets set to the default?
|