|
Atlas at PDSF
PDSF for Atlas
PDSF is a computing facility available to High Energy and Nuclear Physics
users. To learn more about the facility itself visit out
home page
. Amount of resources available to any user group depends on group's financial
contribution towards procurements of the cluster hardware and software. Current
Atlas contribution guarantees 2% of total CPU and some
storage space
. Rather than being an allocation, this is closer to a guaranteed current
minimum. Each of the user groups (including atlas) can make use of any idle
resources exceeding the original share without incurring any penalty. Each
atlas user has also access to the
NERSC mass storage
.
How to gain access
In order to use PDSF resources either via GRID or locally, atlas collaborators
need to obtain
regular unix accounts
. Additionally, to be listed in PDSF grid mapfile a user needs to
submit a request
with relevant information.
Globus Software
PDSF has a Globus
Toolkit 2.0 installed and a gatekeeper running on
pdsfgrid01.nersc.gov, pdsfgrid02.nersc.gov and pdsfgrid03.nersc.gov.
Globus clients can be run on any of the PDSF interactive nodes
(pdsfint.nersc.gov). Users can configure their Globus client
environment by loading a globus module:
module load globus
Atlas Software
PDSF offers its users a standard set of
HENP software
(root, cern libraries, lhc++). We also offer access to both
/afs/usatlas.bnl.gov/ and /afs/cern.ch AFS cells.
Atlas software was also installed and built locally(versions 3.2.0, 4.0.0 and 4.1.0).
Athena Atlfast Standalone Software recommended by the
U.S. Atlas Grid
Testbed Software has been installed as well.
Cluster Status
Information about the
status of our facility
and current load
are also available.
Atlas-specific PDSF FAQ's
- How to access Atlas cvs k-server?
To access the Atlas cvs k-server, a user has to secure first a Kerberos-4
ticket (at CERN this is done automatically when you log in, but PDSF uses a
different authentication model so you have to do it "bu hand"). A Kerberos-4
ticket is generated by running:
/usr/local/pkg/krb4/bin/kinit username@CERN.CH
(watch out for the upper case) and giving your CERN password.
After that you need to use a cvs client
that was built against Kerberos-4 libraries. Since the RH 7.2 upgrade it is our
default version of cvs, so no need to load the cvs module any more. Simply
follow the instructions for
Working with the ATLAS cvs repository.
When you are done with your cvs transactions
it is a good practice to destroy the ticket:
/usr/local/pkg/krb4/bin/kdestroy
We did not put Kerberos-4 kinit and kdestroy into your path because many of
you use Kerberos-5 in transactions with FERMILAB and that could cause
confusion.
- How to set up local Atlas development environment
Add to your .cshrc
(from Paolo)
module load gcc/2.95.2
setenv CMTSITE LBNL
(from Iwona)
module load binutils
The last load gives you a loader from RH 6.1. The RH 6.2 loader
cannot handle some of the Atlas linking and produces really bizzare errors.
Paolo also "ported" David's super-setup system to PDSF. Look into his pdsf
~calaf/.cmt/requirements and drq_setup.csh. If you plan to use it, add
this two aliases to your cshrc:
alias setup "source ${HOME}/.cmt/drq_setup.csh"
alias setuphere 'setenv CMTPATH `pwd`:$CMTPATH'
so that you can e.g. say:
setup 4.1.0
cd your_test_release
setuphere
|