Testbed Support for GridPP member institutes
> [mailto:[log in to unmask]] On Behalf Of Ian Stokes-Rees said:
> And this can be very annoying -- of course it makes sense for
> there to
> be no "BrokerInfo" file if there was no broker, but this is
> the best way
> (I know of) to do "reflection" and find out about the site
> where the job
> is running. It would be very nice to have a "dummy" BrokerInfo file
> available with site information. When a site gets a job
> "directly", can
> it distinguish this from jobs received from an RB? If so, could it
> create a BrokerInfo (or equivalent) file for the job to inspect?
The reason for the brokerinfo file is that in general you don't know
which site a job will land on in advance - but if you use -r you clearly
do! The information is just what you can get anyway from the information
system and catalogue queries. I don't think it would make a lot of sense
for a site to do that, it would be a significant amount of work for
something which should be a rare case. If you want to force a job to a
site but have the brokerinfo you can always do it with a requirement,
e.g. on the ceid.
Stephen
|