On 04/05/12 11:16, emyr.james wrote:
> I'm just going to copy /usr/bin/sge_filestaging onto the worker nodes
> and see if that sorts it out. Going by what Stuart sent it may well do
> the trick.
Yes, that's the way we do it.
You need it setup so that the worker node can scp files from the CE
without needing a password.
Chris
>
> On 04/05/12 11:15, Daniela Bauer wrote:
>> As far as I know this is the way QMUL is set up, so they should be
>> able to explain the finer points of it.
>>
>> Cheers,
>> Daniela
>>
>>
>> On 4 May 2012 11:04, Stuart Purdie<[log in to unmask]> wrote:
>>> On 4 May 2012, at 10:58, Ewan MacMahon wrote:
>>>
>>>>> -----Original Message-----
>>>>> From: Testbed Support for GridPP member institutes [mailto:TB-
>>>>> [log in to unmask]] On Behalf Of Daniela Bauer
>>>>>
>>>>> I certainly have it setup that way (we only have one home dir per
>>>>> user on
>>>>> the whole system on some expensive piece of hardware), but I am not
>>>>> sure
>>>>> it's compulsory - maybe Chris would know ?
>>>>>
>>>> It's not compulsory from a grid POV - we, and I think a majority of
>>>> sites,
>>>> have home directories local to each worker node. We do have to
>>>> configure
>>>> things so that torque can use SSH to stage things in and out though; I
>>>> imagine there's a GridEngine equivalent.
>>> Ah! *ding* goes the memory bell.
>>>
>>> There is, but it's an add on to SGE (freely available). Details
>>> escape me, but might be worth checking if that's present. I think
>>> YAIM, left to it's own devices sorts this out - but YAIM left to it's
>>> own devices tends to comandeer SGE, so my understanding is that it's
>>> normally throttled back.
>>>
>>> Something along the lines of
>>> http://gridscheduler.sourceforge.net/howto/filestaging/
>>>
>>> http://grid.pd.infn.it/cream/field.php?n=Main.InstallationAndConfigurationNotesForCREAMUsingSGEAsBatchSystem
>>> (techincally out of date, but ...)
>>>
>>> Not sure exactly how to check the file staging process, but hopefully
>>> that's a hint in the right direction.
>>
>>
|