Special for RAL
Special instructions for users running at RAL.
Babar members can find out how to get a RAL account from reading this web page
The general setup at RAL is made to be a close as possible to the
setup at SLAC.
- The machine to log into at RAL is called
babar.gridpp.rl.ac.uk (RedHat 6.2) or babar-new.gridpp.rl.ac.uk (RedHat 7.2).
- Watch out for errors when logging in as they indicate some
error in the setup.
- The default shell at RAL is tcsh. If you wish to
change to ksh then log into babar.gridpp.rl.ac.uk or babar-new.gridpp.rl.ac.uk, issue a ypchsh command and then follow the instructions. You must logout then logon again for the change to take effect. It may take up to ten minutes for the change to become visible to the whole Linux farm.
The network connection between UK and SLAC is generally quite good. Most people already know this but for new users a few hints on how to work over the network connection between SLAC and the UK might be helpful.
- When editing files in the SLAC AFS area from the UK you will need an AFS token in the SLAC cell so you can then edit it as a
local file. On your local machine this could be something like.
klog -cell slac.stanford.edu
emacs /afs/slac.stanford.edu/u/ec/egede/.cshrc
However, there is token passing at RAL so if you logged in from a machine
which already had a SLAC token, you will have that same token at RAL and
will not need to run the above klog command. At present there is no token
passing to the batch machines at RAL. So be aware when submiting jobs were
an AFS token is required.
- When copying files to or from AFS areas just use the normal copy
command cp. When copying between NFS areas use the
secure copy command with compression. Both methods
gives faster transfer than ftp and are at the same time more
secure. As an example to copy a root file from SLAC to RAL log into
babar.gridpp.rl.ac.uk and do (on one line):
scp -C olaiya@noric01.slac.stanford.edu:/afs/slac.stanford.edu/g/babar/work/o/olaiya/myevents.root .
- The user home directories are on NFS-mounted disks. This means
that an AFS token is not obtained at login. Your
personal AFS area in the RAL cell is located as
/afs/rl.ac.uk/user/<initial>/<username>.
- To do addpkg and other commands which use the
CVS repository you need to issue a klog
command first.
- The stage disks /stage/babar-user1/<user> and
/stage/babar-awg1 can be used to store large files. There is
no automatic deletion from these disks and more disk space should be
available on request. This is a good place to store root
files.
There is no fixed policy for when BaBar s/w releases are copied to
RAL. Loosely it is something like:
Releases are deleted when we run out of disk space
and always after a warning to the RAL Tier A centre and UK farms mailing list. The releases which reach the current status at SLAC are kept for as long as possible.
At RAL there is a mirror of the SLAC CVS repository
which is updated each night.
- For checking out packages use the addpkg command
exactly as at SLAC.
- To write to the repository use the bcvs command which
interacts with the CVS repository at SLAC directly. First you will need to obtain a SLAC token (
klog <user>@slac.stanford.edu).
- Use bcvs update -A <package> to get the head of the very
latest version of a package (less than 24 hours old). Or you can run the command without the package name in the actual directory
If you develop code at RAL and need to run it both at RAL and at
SLAC it can be very convenient to share the checked out
source you edit between the two sites. To do this:
cd /afs/slac.stanford.edu/<mytestreldir>
ln -s /afs/rl.ac.uk/<mytestreldir>/[A-Z]* .
You will get a warning about GNUmakefile which you can just ignore.
- Make sure your AFS directory at RAL where you have your release has read access without a token. You can do this by changing the permissions with the
commands
cd /afs/rl.ac.uk/<mytestreldir>
klog <user>@rl.ac.uk
find . -type d -exec fs setacl -dir {} -acl system:anyuser rl \;
- Currently, in order to run jobs on the batch machines with your test release on AFS, you will need to move your workdir to an NFS disk. You have to do this from a computer at RAL (babar.gridpp.rl.ac.uk) in order to be able to see the NFS disk. This example will use your home area. The area /stage/babar-user1/<user> is also an alternative option.
mkdir -p /home/csf/<user>/<mytestreldir>
mv /afs/rl.ac.uk/<mytestreldir>/workdir /home/csf/<user>/<mytestreldir>/workdir
- Change the RELEASE soft link in workdir
cd /home/csf/<user>/<mytestreldir>/workdir
ln -nfs /afs/rl.ac.uk/<mytestreldir> RELEASE
The source you edit is now kept updated at both SLAC and RAL since
it is located in the same directory but when linking and compiling
only the small source files are copied across the network.
Jobs can be submitted to the Linux farm which consists of 310 1.4 GHz PIII
machines. Jobs access these machines via the babar queue and
their
priority is partially determined by the amount of CPU requested. More
information on the job priority can be found here:
http://hepunx.rl.ac.uk/BFROOT/csflnx.html
Submission to the babar queues is default when using the bbrbsub
command as follows:
bbrbsub -l cput=hh:mm:ss file.job
Support is available for running the BaBar software at RAL:
- The BaBar computing pages at RAL
http://hepunx.rl.ac.uk/BaBar/software_ral.html
- For general problems with the hardware, disk quotas, forgotten
passwords etc. contact the DCI help desk <support@csfsun.rl.ac.uk
>.
- For problems with the setup at RAL and to request new releases
to be installed contact Tim Adye <T.J.Adye@rl.ac.uk>
and Emmanuel Olaiya <E.Olaiya@rl.ac.uk>
Check the release pages for any hints before reporting a problem.
- For general BaBar problems, i.e. problems which are the same
at RAL and at SLAC use the HyperNews.
