Hi,
If anyone wishes to use Andrew's patch for the glue1 output, I describe
it here; it's doable to implement (with care) it from this description.
Perhaps it would better to dump glue 1 and concentrate on a rigorous
solution in glue2 that nails the standard output down hard in all
implementations.
Anyway, here it is:
https://www.gridpp.ac.uk/wiki/Example_Build_of_an_ARC/Condor_Cluster#Patch_for_BDII_Job_Count_Breakdown
Cheers,
Steve
On 10/26/2015 11:01 AM, [log in to unmask] wrote:
> Hi Steve,
>
> On each ARC CE I have a cron running every 10 mins which does this:
>
> #!/bin/sh
> condor_q -constraint 'JobStatus==2' -autoformat x509UserProxyVOName | sort | uniq -c > /var/local/condor_jobs_running
> condor_q -constraint 'JobStatus==1' -autoformat x509UserProxyVOName | sort | uniq -c > /var/local/condor_jobs_idle
>
> I then modified /usr/share/arc/glue-generator.pl (attached) so that it reads these 2 files and uses it for the glue1 VO views. The modifications are:
> - added a subroutine near the top getCondorJobsRunning
> - added a subroutine near the top getCondorJobsIdle
> - later made modifications to use these 2 subroutines for running/idle jobs per VO
>
> Regards,
> Andrew.
>
> ________________________________________
> From: Stephen Jones [[log in to unmask]]
> Sent: Monday, October 26, 2015 10:11 AM
> To: Lahiff, Andrew (STFC,RAL,PPD)
> Cc: Alessandra Forti
> Subject: Re: ARC CE publishing
>
> On 10/25/2015 08:19 AM, Alessandra Forti wrote:
>> Hi Raja,
>>
>> I don't know if you eventually subscribed the mailing list but the
>> problem here seems to be that a patch Andrew Lahiff wrote to use
>> VOViews in Glue1 is installed only at CERN, RAL and Oxford.
>>
> Andrew,
>
> what is this patch Alessandra speaks of?
>
> Cheers,
>
> Steve
>
>
>
>> For Glue2 Maria is saying everything is fine, but she only mentioned
>> Alice.
>>
>> cheers
>> alessandra
>>
>> On 23/10/2015 10:47, Raja Nandakumar wrote:
>>> Hi,
>>>
>>> In any case, I am very happy to support a (any) solution which can be
>>> used by us. Is there any mailing list that I can sign up to to be
>>> kept up to date on the progress of this topic?
>>>
>>> And I also really hope the solution can be deployed sooner, rather
>>> than later ...
>>>
>>> Thanks and Cheers,
>>> Raja.
>>>
>>> On 23/10/15 09:11, Alessandra Forti wrote:
>>>> The problem here is that ARC and CREAM publish different things (i.e
>>>> they don't publish the queue jobs on a per VO/per role basis) and
>>>> this affects the way dirac brokers jobs, it infact affects the
>>>> output of lcg-info(sites) too because ARC doesn't publish the roles
>>>> altogether so if I want to know if /*/Role=pilot is enabled I can't.
>>>>
>>>> I don't think this is a generalisation problem. I think this is a
>>>> problem of what we want them to publish. HTcondorCE will have the
>>>> same problem.
>>>>
>>>> cheers
>>>> alessandra
>>>>
>>>>
>>>> On 22/10/2015 20:54, [log in to unmask] wrote:
>>>>> Alessandra Forti [mailto:[log in to unmask]] said:
>>>>>> It seems not. Something else to feed back to the TF.
>>>>> It can be generalised (at least if things are published correctly)
>>>>> but it would be more complicated and needs to be wrapped up in
>>>>> something like lcg-info.
>>>>>
>>>>> Stephen
>>>>>
>
> --
> Steve Jones [log in to unmask]
> Grid System Administrator office: 220
> High Energy Physics Division tel (int): 43396
> Oliver Lodge Laboratory tel (ext): +44 (0)151 794 3396
> University of Liverpool http://www.liv.ac.uk/physics/hep/
>
--
Steve Jones [log in to unmask]
Grid System Administrator office: 220
High Energy Physics Division tel (int): 43396
Oliver Lodge Laboratory tel (ext): +44 (0)151 794 3396
University of Liverpool http://www.liv.ac.uk/physics/hep/
|