On Wed, 15 Feb 2012 09:44:14 +0000
Stephen Burke wrote:
Hi,
> LHC Computer Grid - Rollout [mailto:[log in to unmask]]
> > On Behalf Of Jeff Templon said:
> > If "VO:atlas" appears in the static LDIF file, and "atlas" is the
> > name of a unix group, then some lines about atlas should be printed
> > by the dynamic scheduler program. I *think* that if /atlas is
> > listed in the static LDIF file, but not "VO:atlas", that it also by
> > default prints for /atlas. But maybe not.
>
> In the past pic had a problem with publishing whole-VO FQANs, i.e.
> things like VOMS:/atlas, which led to this request for yaim to flag a
> warning:
>
> https://savannah.cern.ch/bugs/?func=detailitem&item_id=54755
>
> At a quick look it does now have the normal VO:atlas etc, plus FQANs
> for a load of roles and groups which looks a bit strange - and it
> hence has lots of VOViews including DENY rules. Is there a good
> reason for configuring it like that?
I've always seen the warning, but I though it was due to my wildcards
entries: (groups.conf):
"/atlas/ROLE=lcgadmin":::sgm:
"/atlas/ROLE=pilot":::pilot:
"/atlas/ROLE=production":::prd:
"/atlas/ROLE=software":::sgm:
"/atlas"::::
"/atlas/*":::: <--------------------------------
I wanted to say: if found any of the defined roles, map it to specific
group (sgm,pilot,prd,sgm), if no role ("/atlas"::: , old
ROLE=NULL/Capability=NULL) map to regular UNIX group (atlas), if any
other role, map to unix atlas group too.
Isn't it incorrect?
After removing last line from groups, and removing 'atlas entry' from
QUEUE_ENABLE:
GSHORT_SL5_GROUP_ENABLE="
/ops/ROLE=pilot
/ops/ROLE=lcgadmin
/cms/ROLE=production
/cms/ROLE=lcgadmin
/cms/ROLE=t1access
/cms/ROLE=t1production
/cms/ROLE=hiproduction
/cms/ROLE=hit1production
/cms/ROLE=pilot
/atlas/ROLE=lcgadmin
/atlas/ROLE=production
/atlas/ROLE=pilot
/atlas/soft-valid/ROLE=production
/lhcb/ROLE=lcgadmin
[...]
If I reconfigure, the warning is still there.:
[...]
WARNING: Queue gshort_sl5 was enabled for one or more FQANs which specify an entire VO
WARNING: Queue gshort_sl5 was enabled for one or more FQANs which specify an entire VO
[...]
2 questions here:
1) "/atlas"::: enables the whole VO?
2) If new queue_enable looks as above, why is the warning shown?
> Stephen
Cheers,
Arnau
|