> -----Original Message-----
> From: Testbed Support for GridPP member institutes
> [mailto:[log in to unmask]]On Behalf Of Ewan MacMahon
> Sent: 08 August 2008 12:43
> To: [log in to unmask]
> Subject: Re: Inefficient Jobs
>
>
> > -----Original Message-----
> > From: Testbed Support for GridPP member institutes [mailto:TB-
> >
> > I have received a response from my ticket...
> > https://gus.fzk.de/pages/ticket_details.php?ticket=39591
> >
> > It states we can kill her jobs... which I have done.
> >
> That response is from a dteam VO person though isn't it? While this is
> clearly
> an inappropriate use of the dteam VO, it's not at all clear to me that
> the
> user actually meant to use the dteam VO. At Oxford we (obviously)
> support
> dteam, but don't support biomed, and I don't see any submissions from
> this
> user, so I'd guess that the RB is only matching the jobs to biomed
> supporting
> sites.
>
I agree, we've only seen jobs via the CE that supports biomed.
> So, I think we've got two problems here:
> - badly written jobs,
> - incorrect VO mapping at some sites (rather than abuse of
> the dteam VO)
>
> It'd be interesting to know why RAL-LCG2 and Liverpool mapped the user
> to a
> biomed account, they're certainly in my grid-mapfile as dteam. Is the
> user
> not using a VOMS extended proxy, or is Durham having trouble with the
> VOMS
> and falling back to the basic grid-mapfile approach?
>
One of the jobs I looked at did have biomed voms extensions in the proxy
attribute : /biomed/Role=NULL/Capability=NULL
attribute : /biomed/lcg1/Role=NULL/Capability=NULL
but we map the user to biomed in the gridmap file anyway on that CE,
their mapped to dteam on the CEs that don't support biomed. I think the
order of the VOs in ed-mkgridmap.conf is significant - we have dteam
right at the end.
Derek
|