So, really to isolate this possibility, Xavier should try to submit jobs
via WMS to ce02.lip.pt using a WMS different from his normal one....
Cheers
Goncalo
Maarten Litmaath wrote:
> Burke, S (Stephen) wrote:
>
>>> Nevertheless, from Xavi tests, his jobs failed both at ce02.lip.pt
>>> and ce101.cern.ch when submitted from the same UI as his pilot jobs.
>>> From standard UI everything is OK at both CEs. This points to
>>> something incorrect in the pilot jobs UI which prevents things to
>>> properly interact with lcg-CE gLite 3.1.
>>
>>
>> I don't think so, I doubt the pilot jobs are being submitted with
>> globus-job-run! I still think the most likely case is that you're
>> seeing errors related to an old proxy with the expired AC and not the
>> newly-created one. (If the jobs go through a WMS, is it using a
>> single delegated proxy, and if so was it replaced?)
>
> Condor-G on the WMS uses a single master proxy per DN+FQAN,
> for which AFAIK it takes a fresh proxy every once in a while,
> so it would be possible for a job to get submitted to the CE
> with a different proxy than the one used to submit it to the
> WMS itself. This may have confused the analysis of the errors
> that were observed.
>
>> PS I seem to remember that proxy renewal on the WMS only looked at
>> the overal proxy lifetime and not the ACs, although I don't know if
>> that's still true.
>
> That bug is still open indeed:
>
> https://savannah.cern.ch/bugs/index.php?28167
|