Tom,
strictly speaking, the (GridPP) LFC already uses "/grid/full.vo.name/"
for its LFNs, which isn't quite the same (though at least it's trivially
convertible).
Daniela,
Thanks for the explanation. I think I'll bury my head in the sand for now
rather than try to work out how it would work with our stuff on Castor, if
we ever had to move!
Thanks
Henry
On Mon, 8 Feb 2016, Tom Whyntie wrote:
> Thanks for this - seems like a good system for the DFC as we move forward.
>
> I've added and self-assigned an issue to the GridPP UserGuide to add a note to this effect (the /full.vo.name/ LFN convention is already implicitly used, but an info box might be useful).
>
> Cheers, Tom
>
> On Fri, 5 Feb 2016 at 13:06 Daniela Bauer <[log in to unmask]> wrote:
> Hi Henry, Sam,
>
> they have been warned or if not, they will be when they start using dirac.
> So far all the VOs that started using dirac have just asked, as the situation is different for each VO.
> Most of the VOs have little or no existing data. The problem only creeps up when you try and integrate existing data.
>
> The requirement is that in the dirac file catalogue, as it is shared between multiple VOs, the LFN starts with /full.vo.name. As far as I am aware that is the only requirement.
>
> dirac then attaches this to the VO path on a given SE. As DPM automatically uses the VO name in the VO path, you end up with /dpm/home/voname/voname (or so, I can't quite remember the default DPM
> naming scheme). dCache lets you define the VO path on the SE much more freely, so there you have /pnfs/blah/vonickname(or in case of t2k I've seen 'nd280data')/voname.
>
> It was bad enough getting dirac to accept that not every SE uses /one/single/path/+voname (as there has never been a requirement for it) and the current setup, while far from ideal, at least is
> consistent across all SEs and therefore predictable. Which in the long run is what users want.
> If we start introducing exceptions ("if DPM then ..."), that quickly becomes unmaintainable and also we would have to ask everyone who already has data on the SEs according to the current scheme to
> move.
>
> So far the SEs haven't been used much via dirac. Most VOs seems to have one pet SE and as long as that works they are happy, so I am not surprised that Alessandra encounters difficulties.
>
> Regards,
> Daniela
>
>
>
> On 5 February 2016 at 12:36, Henry Nebrensky <[log in to unmask]> wrote:
> Hi,
>
> Basically their naming scheme clashes with the requirement of the dirac file catalogue.
>
>
> Just out of interest, are the details of the clash and the dirac requirement something us other VOs should be warned about?
>
> Thanks
>
> Henry
>
> On Fri, 5 Feb 2016, Alessandra Forti wrote:
>
> Hi Daniela,
>
> it is 1.7TB I've just calculated it.
>
> gfal-ls -l srm://gridpp09.ecdf.ed.ac.uk:8446/srm/managerv2?SFN=/dpm/ecdf.ed.ac.uk/home/lsst|grep DES|awk '{SUM+=$8}END {print SUM/10^9}'
> 1742.55
>
> It is going to be a pain but it is a one off and we probably should do that and take into account the extra directory in future transfers.
>
> cheers
> alessandra
>
> On 05/02/2016 11:16, Daniela Bauer wrote:
> Hi All,
>
> this came up in context of the lsst data at ECDF. Basically their naming scheme clashes with the requirement of the dirac file catalogue. An easy
> fix would be to just move all the files one level down, i.e. from lsstdir/somefile to lsstdir/lsst/somefile.
> If I understand Alessandra's post correctly this would require copying the data out and back in. Can DPM really not handle moving files without
> copying them ?
> Unfortunately the SE bdii seems to not quite setup correctly (at least lcg-infosites just returns N/A for the data size), so I'm not sure of how
> much data we are talking here.
>
> Regards,
> Daniela
>
>
>
> --
> Sent from the pit of despair
>
> -----------------------------------------------------------
> [log in to unmask]
> HEP Group/Physics Dep
> Imperial College
> London, SW7 2BW
> Tel: +44-(0)20-75947810
> http://www.hep.ph.ic.ac.uk/~dbauer/
>
>
> --
> Respect is a rational process. \\//
> Fatti non foste a viver come bruti (Dante)
>
>
>
> --
> Dr. Henry Nebrensky [log in to unmask]
> http://people.brunel.ac.uk/~eesrjjn
> "The opossum is a very sophisticated animal.
> It doesn't even get up until 5 or 6 p.m."
>
>
>
>
> --
> Sent from the pit of despair
>
> -----------------------------------------------------------
> [log in to unmask]
> HEP Group/Physics Dep
> Imperial College
> London, SW7 2BW
> Tel: +44-(0)20-75947810
> http://www.hep.ph.ic.ac.uk/~dbauer/
>
>
>
--
Dr. Henry Nebrensky [log in to unmask]
http://people.brunel.ac.uk/~eesrjjn
"The opossum is a very sophisticated animal.
It doesn't even get up until 5 or 6 p.m."
|