Re: [Opaltoolkit-users] Opal job scheduler benefits

[Luca C. <luc...@gm...>]

On Tue, Mar 5, 2013 at 8:21 AM, Nadya Williams <na...@sd...> wrote:
>
> Hi Guillermo,
>
> On Mar 5, 2013, at 3:19 AM, Guillermo Marco Puche wrote:
>
> Hello Nadya,
>
> I remade my XML and created a wrapper as you told me.
> Asking user for Condor input file works great !
>
> glad it works.
>
>
> The only I thing I don't like about this is that user gets status 8
> (execution completed) right after submitting the job. I mean you cannot
> track the real job status from Opal because of Fork Job Manager.
>
> This is true.
> you still can do something about it in your shell wrapper. After
> condor_submit command you have  to parse stout
> and find a jobid for your submission. you can query with condor_queue and
> looking for a specific jobid
> in a loop and check for the job status on some interval. When a job is
> done you exit the loop.  The interval value is hard to
> guess and may need to depend on overall expected job time. But one can
> choose an intelligent default since condor
> is "high throughput" and not "high performance". May be waiting for
> minutes is not a big deal. So putting a "sleep X"
> in your loop will work fine.
>

Why don't you try to use the condor_wait logfile to do that?

condor_wait should block till the job logging in the logfile is finished.

Luca