Hi Marcus,
I've replicated the content from UKI-SCOTGRID-ECDF1-disk to
UKI-LT2-IC-HEP-disk (at least all the bits that are contained within
the dirac catalogue):
Number of files copied: 1942
Data copied: 7.041 GB
So they should stay accessible no matter what happens to the ECDF SEs.
The GridPP dirac server is a standard 'generic' dirac server - most
stuff LCHb does eventually trickles down to it, but it's not a
one-to-one mapping. LHCb only solve their specific problems and not
the more generic cases we tend to face. The last time we tried to use
xrootd, the dirac server just broke, so it's turned off. We now have a
dirac test instance which we use to debug such things, so whenever
your SE is ready, we can have a look.
As dirac ships it's own externals which at the moment include lcgutils
it's probably going to stay around for a while.
Cheers,
Daniela
On 30 November 2016 at 15:01, Marcus Ebert <[log in to unmask]> wrote:
> Hi Daniela,
>
>
> On Wed, 30 Nov 2016, Daniela Bauer wrote:
>
>> Hi Marcus,
>>
>> as for the bdii both SEs are still there:
>> lx01:uki-lt2-ic-hep :~] lcg-infosites --vo lsst se | grep ecdf
>> 21614809465 37398041181 SRM gridpp09.ecdf.ed.ac.uk
>> 87167372114 11788674384 SRM srm.glite.ecdf.ed.ac.uk
>> 140325881655 423808716 SRM srm.glite.ecdf.ed.ac.uk
>>
> Yes, that's the confusing part and I'm not sure why srm.glite appears 2
> times there too...
>
>> It might just be cached somewhere, or if it's DPM I remember it being
>> particularly recalcitrant in letting VOs go.
>>
> Ah, yes that could be. I'll look into it more once the files are copied and
> deleted. Maybe the lsst vo has to be removed to not appear on dpm-listspaces
> anymore.
>
>
>> There's no xrootd in dirac, I thought gfal2 was the official
>> replacement for the lcgutils.
>> I'd suggest once the new SE is setup we see how it goes and what kind
>> of adjustments we need to make.
>>
> That's interesting. Is this a feature of GridPP dirac?
> We use xrootd through LHCb's dirac I believe and I can see in their
> configuration for the sites that xrootd is a supported protocol at least for
> their Dirac installation.
>
> For us there will be no new prodcution SE, only the one behind srm.glite
> which will also get the disk space currently used by gridpp09. I recognized
> the problem when looking to the new nagios tests for srm.glite that are in
> place now.
>
>
> Cheers,
> Marcus
>
>
>> Regards,
>> Daniela
>>
>> On 30 November 2016 at 14:17, Marcus Ebert <[log in to unmask]> wrote:
>>>
>>> Hi Daniela,
>>>
>>> Yes, I'm in the process of copying everything over to
>>> UKI-SCOTGRID-ECDF-disk
>>> and then wanted to delete the old files
>>> (all through the dirac tools which should make sure the file catalogue
>>> stays
>>> correct, I hope)
>>>
>>> I know I asked in the past how to make a SE visible for Dirac through
>>> BDII,
>>> but can't figure out right now what is needed to have it removed again. I
>>> removed lsst from /var/lib/bdii/gip/glite-info-service-srm2.2.conf but it
>>> still appears in the list. Do you know of any other place where it needs
>>> to
>>> be removed?
>>>
>>> Another SE related issue that came up while looking to lsst jobs is the
>>> transfer protocol used. On our SL7 worker nodes there are no lcg-utils
>>> available. To configure the site in dirac to use the xrootd protocol
>>> instead
>>> of srm, is this something that also needs to be done through the side
>>> BDII
>>> or centrally on the dirac server?
>>> (sorry for mixing 2 issues here)
>>>
>>> Cheers,
>>> Marcus
>>>
>>>
>>> On Wed, 30 Nov 2016, Daniela Bauer wrote:
>>>
>>>> Hi Marcus,
>>>>
>>>> Do you want these files transferred to UKI-SCOTGRID-ECDF-disk
>>>> (srm.glite.ecdf.ed.ac.uk) ?
>>>> If you remove lsst from the bdii, the SE on dirac will automatically
>>>> update.
>>>> However, the real problem is that the files are registered in the
>>>> dirac file catalogue and they will have to be moved (deleted and
>>>> reregistered) within the catalogue - one of the reasons why switching
>>>> SEs is somewhat frowned upon.
>>>>
>>>> Regards,
>>>> Daniela
>>>>
>>>>
>>>>
>>>> On 28 November 2016 at 18:07, Marcus Ebert <[log in to unmask]>
>>>> wrote:
>>>>>
>>>>>
>>>>> Hi,
>>>>>
>>>>> I need to move and reinstall a SE and in the process the data on it
>>>>> will
>>>>> get
>>>>> lost.
>>>>> It's mostly used for ATLAS tests in the past, but was also integrated
>>>>> into
>>>>> Dirac as a SE for LSST.
>>>>> Does anyone knows what is needed to have it removed from the list that
>>>>> one
>>>>> gets with dirac-dms-show-se-status?
>>>>> I'm writing out the list of files for LSST on that machine right now
>>>>> and
>>>>> will replicated it to our other storage element. Once that is done, the
>>>>> LSST
>>>>> support needs to be removed from that SE(UKI-SCOTGRID-ECDF1-disk).
>>>>>
>>>>>
>>>>> Cheers,
>>>>> Marcus
>>>>>
>>>>> --
>>>>> The University of Edinburgh is a charitable body, registered in
>>>>> Scotland, with registration number SC005336.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>> --
>>> The University of Edinburgh is a charitable body, registered in
>>> Scotland, with registration number SC005336.
>>
>>
>>
>>
>>
>
> --
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
--
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/
|