On 10/24/2012 06:13 PM, Torsten Harenberg wrote:
> Hi Massimo,
>
>
> Am 24.10.2012 um 17:59 schrieb Massimo Sgaravatto <[log in to unmask]>:
>
>> "Killed by CE admin" means that the job was killed, but not because a cancel command issued by the user (e.g. it was killed by bkill/qdel/...).
>
> Ah .. Ok.
>
>> You are using (s)ge right ?
>
> Yes.
>> I remember that there were cases where this message was reported because of a problem/misconfiguration in the bupdater. The cesga/lip guys know more for sure ...
>
> Hmm... Is maybe the output of qacct parsed?
Yes. something like that, as far as I remember
The CREAM-SGE guys can comment
If so, could it has something to do with the segfault I reported in the
first mail?
>
You mean the "cannot find grid-proxy-info" error ?
No, for sure they are not related
> Surprisingly, it really only affects Atlas production. We are even just set back online for User Analysis, as the test jobs succeeded. So I cannot imagine that it is a general issue.
>
Is the same queue used for both atlas analysis and atlas production ?
Is the same set of WNs used in both scenarios ?
Is the submission method (condor) always used ?
Cheers, Massimo
> Thanks, cheers
>
> Torsten
>
|