> > P.S: Be careful with "automatic" retrieval: if two "edg-job-get-output"
> > are running in parallel, they will interfere with each other. Locally, you
> > will get an overwrite. On the RB side, the second edg-job-get-output might
> > arrive at the time where the first one finishes, deleting the Output
> > Sandbox directory. Hence partial retrievals...
>
> Argh. You've got to be kidding. That is a serious problem! How can
> this possibly be? Do they use hard-coded temporary file names or
> something? Of course I do them in parallel! It takes ages to wait for
> each one to complete.
Hum, no , I meant, two edg-job-get-output on the same job (string).
It takes a little time for the job status to be updated from "Done
(Success)" to "Cleared".
While preparing my demo in Cambridge, I made the accidental mistake to have
two "automatic retrieval" processes running closely in parallel, each of
them scanning the same list of job strings using edg-job-status and
retrieving "Done" jobs using edg-job-get-output.
>
> Is this registered as a bug? If not, I will input it immediately.
I considered that was MY mistake, so I did not register it as a bug.
Cheers,
Frederic
>
> Cheers,
>
> Ian.
> --
> Ian Stokes-Rees [log in to unmask]
> Particle Physics, Oxford http://www-pnp.physics.ox.ac.uk/~stokes
>
|