Hi,
Looking at this closely, the problem comes from this:
On Oct 22, 2010, at 21:28 , Maarten Litmaath wrote:
>> in /opt/glite/etc/lcg-info-dynamic-scheduler.conf we had:
>> neutrinos:/supernemo.vo.eu-egee.org
>> nemosgm:/supernemo.vo.eu-egee.org/Role=lcgadmin
>> nemoprd:/supernemo.vo.eu-egee.org/Role=production
>> neutrinos:/supernemo.vo.eu-egee.org/*
>>
>> due to this groups.conf:
>> "/supernemo.vo.eu-egee.org"::::
>> "/supernemo.vo.eu-egee.org/Role=lcgadmin":::sgm:
>> "/supernemo.vo.eu-egee.org/Role=production":::prd:
>> "/supernemo.vo.eu-egee.org/*"::::
>
The problem is that "neutrinos" is mentioned twice, l-i-d-s will only use the last one, hence there will be no field in the VOViews like
VO:supernemo.vo.eu-egee.org
A single unix group should never be mentioned twice in the lcg-info-dynamic-scheduler conf file. The function of that mapping in that file, is to say for jobs with unix group X, what ACBR field should be printed in the output? It is a one-to-many mapping, ie I can have
neutrinos
nemosgm
nemoprd
all mapped to a single output ACBR "VO:supernemo", but I cannot have a many to one mapping.
JT
> That is wrong. In groups.conf the first field must be a VOMS FQAN
> or a wildcard for a set of FQANs, i.e. it should start with a '/':
>
> https://twiki.cern.ch/twiki/bin/view/LCG/YaimGuide400#groups_conf
>
> If that causes a problem in lcg-info-dynamic-scheduler.conf,
> there is a bug in the YAIM function that creates that file.
>
> Currently YAIM _may_ happen to do the right thing if the first field
> is just the name of the VO (to be checked), but you should not count
> on such a feature.
>
> While the bug is there, it would be safer to create a post-configuration
> function to correct lcg-info-dynamic-scheduler.conf, as explained here:
>
> https://twiki.cern.ch/twiki/bin/view/LCG/YaimGuide400#Use_local_functions
>
> At CERN we have such functions on our CE and WMS nodes to deal
> with various bugs and other issues specific to our environment.
|