Hi again,
Further to the last email....
The way I had things set up was like this -
globus-mds (2135) on each of the CE's - fine. I could query each via
ldapsearch ....:2135 mds-vo-name=local and see the expected info.
what I didn't realise is that I also needed a site giis ( lcg-bdii
(2170) ). When I realised this, then, ldapsearch .... : 2170
mds-vo-name=SITE-NAME worked!!!!
So now, instead of pointing the resource broker at the site GRIS
(2135), I point it at the site GIISes (2170) and edg-job-list match
shows me all available compute elements!
Thankyou *HEAPS* to Mario and Stephen for your questions which got me
thinking... : )
Regards, Marco.
On Sun, 2005-04-10 at 11:55 +1000, Marco La Rosa wrote:
> Hi Mario,
>
> On Fri, 2005-04-08 at 09:07 +0100, Mario David wrote:
> > when you do
> > ldapsearch ... ldap://lcg-host:2135...
> > do you see the
> > dn: GlueCEUniqueID=lcg-host (what does it publish in total cpus etc?)
> >
>
>
> On the CE which does NOT get listed in the bdii:
>
> lcg01$ ldapsearch ldap://lcg01.vpac.org:2135 -b
> "mds-vo-name=local,o=grid"
>
> --> dn: GlueCEUniqueID=lcg01.vpac.org:2119/jobmanager-lcgpbs-atlas,
> mds-vo-name=local,o=grid
> GlueCEUniqueID: lcg01.vpac.org:2119/jobmanager-lcgpbs-atlas
> GlueForeignKey:
> GlueCEUniqueID=lcg01.vpac.org:2119/jobmanager-lcgpbs-atlas
>
> lcg01$ ldapsearch ldap://lcg01.vpac.org:2135 -b
> "mds-vo-name=VPAC-LCG,o=grid"
>
> --> no output
>
> On the CE which does get listed in the bdii:
>
> lcg01$ ldapsearch ldap://lcg01.hpc.unimelb.edu.au:2135 -b
> "mds-vo-name=local,o=grid"
>
> --> dn:
> GlueCEUniqueID=lcg01.hpc.unimelb.edu.au:2119/jobmanager-lcgpbs-atlas,
> mds-vo-name=local,o=grid
> GlueCEUniqueID:
> lcg01.hpc.unimelb.edu.au:2119/jobmanager-lcgpbs-atlas
> GlueForeignKey:
> GlueCEUniqueID=lcg01.hpc.unimelb.edu.au:2119/jobmanager-lcgpbs-atlas
>
> lcg01$ ldapsearch ldap://lcg01.hpc.unimelb.edu.au:2135 -b
> "mds-vo-name=MU-hpc,o=grid"
>
> --> no output
>
>
> > now the second part
> > when you do
> > ldapsearch ... ldap://lcg-host:2170 -b "mds-vo-name=YOUR-SITE-NAME,o=grid"
> >
> > does the dn: GlueCEUniqueID=lcg-host appear at all
>
>
> grickle$ ldapsearch ldap://grickle.ph.unimelb.edu.au:2170 -b
> "mds-vo-name=local,o=grid"
>
> --> GlueForeignKey:
> GlueCEUniqueID=lcg01.vpac.org:2119/jobmanager-lcgpbs-atlas
> GlueForeignKey:
> GlueCEUniqueID=lcg01.hpc.unimelb.edu.au:2119/jobmanager-lcgpbs-atlas
> dn:GlueCEUniqueID=lcg01.hpc.unimelb.edu.au:2119/jobmanager-lcgpbs-atlas,mds-vo-name=local,o=grid
> GlueCEUniqueID: lcg01.hpc.unimelb.edu.au:2119/jobmanager-lcgpbs-atlas
>
> grickle$ ldapsearch ldap://grickle.ph.unimelb.edu.au:2170 -b
> "mds-vo-name=MUPHYS-LCG,o=grid"
>
> --> no output
>
> I forgot to point out that I used YAIM-2_4_0 to install each node (OS =
> SL-3.0.4 with updates from linuxsoft.cern.ch). Also, my aim is to build
> LCG 'front-ends' to integrate into existing PBS clusters. That's why
> initially I am trying to integrate CE, SE and UI's on the 'front-ends'
> whilst having a separate II to which they report.
>
> As a side note, and following on from the mail:
>
> Subject:
> [LCG-ROLLOUT] : Different LCG
> services on the same node
> Date:
> Fri, 8 Apr 2005 11:17:04 +0200
> (19:17 EST)
>
> I initially started with LCG-2_3_0 and managed to install *all* of the
> services on one physical machine (RH-7.3 and SL-3 cern version). This
> node was pointed at a PBS cluster and worked fine.
>
> LCG-2_3_1 was another matter however. No matter what I tried, I could
> not get all of the LCG services to exist happily on one machine.
>
> Anyway, I think that's everything.
>
> Regards, Marco.
>
>
> >
> > OK, presently thats my problem
> > all the GlueCEUniqueID part is not being feed into the BDII
> >
> > cheers
> >
> > Mario
> >
> > Quoting Marco La Rosa <[log in to unmask]>:
> >
> > > Hi,
> > >
> > > My question relates to the RB seemingly not publishing all of its
> > > information. We have two LCG front-ends (CE's) which report to the same
> > > resource broker.
> > >
> > > An ldap query on each front-end (ldapsearch -x -H ldap://lcg-host:2135
> > > -b "mds-vo-name=local,o=grid") shows me everything about that host.
> > > (Note that these hosts are CE, SE and UI's combined)
> > >
> > > An ldap query on the RB (ldapsearch -x -H ldap://rb:2170 -b ...)
> > > shows me everything about the two front-ends.
> > >
> > > The problem is that "edg-job-list-match test.jdl" only shows one of the
> > > CE's as matching my requirements. Given that I have not specified any
> > > requirements, I'm wondering why they are not both shown?
> > >
> > > The problem is not a firewall one as these have full access to each
> > > other. It's not a service one (globus-gatekeeper running on each
> > > front-end, lcg-bdii running on RB).
> > >
> > > Thanks in advance.
> > >
> > > Marco
> > >
|