re:
> Use of uninitialized value in concatenation (.) or string at
> /opt/bdii/sbin/bdii-update line 748, <ERROR> line 596.
> Error nearby
in new CE bdii.log
> Date: Sun, 13 Dec 2009 08:39:12 +0000
> From: Daniela Bauer <[log in to unmask]>
> Subject: Re: bdii error, adding new CE
>
> Hi Winnie,
>
> I haven't got access to my bdii right now, but try looking in the
> actual ldif file it makes (something with '1' and 'cache' in the
> path). Often the error message in there (if there is any) is far more
> instructive than the logs.
Thanks Daniela, took a look. All looked unsuspicious except this:
gluelocationpath: $VO_LHCB_SW_DIR
Seems strange. It's in a unique-looking stanza named VO-lhcb-pilot:
dn: GlueLocationLocalID=VO-lhcb-pilot,GlueSubClusterUniqueID=lcgce03.phy.bris.
ac.uk,GlueClusterUniqueID=lcgce03.phy.bris.ac.uk,mds-vo-name=resource,o=grid
objectclass: GlueClusterTop
objectclass: GlueLocation
objectclass: GlueSchemaVersion
objectclass: GlueKey
gluelocationlocalid: VO-lhcb-pilot
gluelocationname: VO-lhcb-pilot
gluelocationversion: Prod
gluelocationpath: $VO_LHCB_SW_DIR
gluechunkkey: GlueSubClusterUniqueID=lcgce03.phy.bris.ac.uk
glueschemaversionmajor: 1
glueschemaversionminor: 2
I'd never deployed pilot roles before but set up the groups & users on the
new CE for pilots for lhcb, atlas, cms & few others. But there's only the
one stanza in bdii GIP.ldif about pilot, & only for lhcb. Do other
sites with polot roles supported have the same?
Then spotted this strange animal at the end of software versions:
objectclass: GlueSchemaVersion
gluechunkkey: GlueClusterUniqueID=lcgce03.phy.bris.ac.uk
gluehostapplicationsoftwareruntimeenvironment: GLITE-3_1_0
gluehostapplicationsoftwareruntimeenvironment: GLITE-3_2_0
....
gluehostapplicationsoftwareruntimeenvironment: VO-lhcb-pilot eh??
lhcb has installed software, but all the vo-tags files in
/opt/glite/var/info/${ce}/${vo} are empty.
(think the ${ce} part is a recent lcg-CE change, other CEs don't have it)
Is this all pointing to some site-info.def or vo.d/lhcb error? Clues/hints
welcome.
But then compare with other 2 prod CEs (with no concat errors in bdii.log)
root@lcgce02> grep -r SW_DIR /opt/bdii/var/cache/ | wc -l
123
those are all
GlueLocationPath: $VO_CMS_SW_DIR
root@lcgce01> grep -r SW_DIR /opt/bdii/var/cache/ | wc -l
1372
Atlas, CMS, lhcb.
gadzooks. I really don't pretend to understand any of this.
Does it look normal that $VO_LHCB_SW_DIR is in the bdii
cache/{1,2,3}/GIP.ldif file? Do other sites supporting lhcb pilot
roles have this too?
|