Daresbury Laboratory Condor Pool
This static information was last updated 3/3/2009. This pool is currently maintained by Rob Allan and Dave Cable.
Name OpSys Arch State Activity LoadAv Mem ActvtyTime
animal.dl.ac.uk LINUX INTEL Unclaimed Idle 0.000 2048 0+02:30:04
beaker.dl.ac.uk LINUX INTEL Unclaimed Idle 0.000 2048 0+01:05:04
bunsen.dl.ac.uk LINUX INTEL Unclaimed Idle 0.000 2048 0+01:15:04
fozzy.dl.ac.uk LINUX INTEL Unclaimed Idle 0.000 2048 0+01:10:04
kermit.dl.ac.uk LINUX INTEL Unclaimed Idle 0.000 2048 0+01:05:04
firstname.lastname@example.org LINUX INTEL Unclaimed Idle 0.000 1643 0+01:10:04
email@example.com LINUX INTEL Unclaimed Idle 0.000 1643 0+23:15:28
firstname.lastname@example.org LINUX INTEL Unclaimed Idle 0.000 1643 0+01:10:04
email@example.com LINUX INTEL Unclaimed Idle 0.000 1643 0+23:15:28
statler.dl.ac.uk LINUX INTEL Unclaimed Idle 0.240 2048 0+00:00:04
waldorf.dl.ac.uk LINUX INTEL Unclaimed Idle 0.010 2048 0+00:00:04
firstname.lastname@example.org LINUX X86_64 Unclaimed Idle 1.000 1024 0+02:40:04
email@example.com LINUX X86_64 Unclaimed Idle 3.420 1024 1+02:40:49
rjavig6.dl.ac.uk WINNT51 INTEL Unclaimed Idle 0.020 2030 3+22:01:52
Total Owner Claimed Unclaimed Matched Preempting Backfill
INTEL/LINUX 11 0 0 11 0 0 0
INTEL/WINNT51 1 0 0 1 0 0 0
X86_64/LINUX 2 0 0 2 0 0 0
Total 14 0 0 14 0 0 0
Recent changes include:
- Move RMCS server to rmcs.dl.ac.uk in Grid DMZ
- Insall Globus on rmcs
- Add dl1.nw-grid.ac.uk head node
- Install jobmanager-condor on dl1 and rmcs servers
- Add machines in A27A.
- Add machines in A27. 2x machines to add plus desktop systems
- Add machines in A33. New desktop PC plus existing PC.
Installation Procedure on Linux Systems
Further instructions on UW Madison Web site for 7.2.1
- Setup up the condor user, on DL systems this is uid 14168 in group condor which is gid 14168. I've created a script to do this.
- As root, download appropriate distro from Condor Web site into /home/condor or /usr/local/src. If the O/S is not matched, try the generic glibc32 tarball. On Gentoo the RHEL3 dynamic version seems to work. Here is the site for Stable release v7.2.1
- Unpack the tarball. This will create something like /home/condor/condor-7.2.1
- In /home/condor/condor-7.2.1 run "condor_install --prefix=/opt/condor-7.2.1"
- You may have to install libstdc++ or a compatibility layer if you dont already have it. On Gentoo "emerge libstdc++"
- cd to /opt and create a soft link by doing "ln -s condor-7.2.1 condor". This will be used in subsequent steps and facilitates easy updates. This is also done by the condor_setup_script mentioned below.
- Edit /opt/condor/etc/condor_config to allow appropriate hosts e.g. "HOSTALLOW_READ = *" and "HOSTALLOW_WRITE = *" is suitable for a submit node inside a firewall (note: in 7.4.1 it was found that these settings have been changed to ALLOW_READ and ALLOW_WRITE, but the syntax is the same). Everything else is over-ridden in condor_config.local (see below).
- You will need to create the init script. On Gentoo this is /etc/init.d/condor and is of SysV style. A sample can be found in /opt/condor/etc/examples/condor.boot which can be copied over and edited. It is however better to use John Kewley's modified script which has start, stop and restart options. It can be found here.
- You need a directory called /opt/condor/local.$(HOSTNAME). This should contain condor_config.local of which this is an example modified for tardis.dl.ac.uk on the DL pool.
- If you do not have an /etc/profile.d/condor.sh you need to create one, here is an example. Note, this should have been created by the condor_install command.
- "chown -R condor:condor /opt/condor/local.$(HOSTNAME) /opt/condor/etc/condor_config". I've created a script to copy the examples and do this part of the setup and it also does the soft link mentioned above.
- Do "/etc/init.d/condor start" and sit back. You can access the Web site at http://tardis.dl.ac.uk/Condor to check the status or do "condor_status -pool condor-main.dl.ac.uk"
- Don't forget to put ntpd on the machine too for any distributed computing activity.
Using and Testing
Need to source /opt/condor/condor.sh in user's environment somewhere. On some machines we set it in /etc/profile.
Some useful condor commands include:
condor_submit -- submits a job
condor_q -- see your own jobs
condor_q -g -- see all jobs
condor_rm -- kill a job
condor_q -better-analyze -- find out why a job isn't running
A sample job submission script which includes file transfer stuff and submits a linux job from a linux node is:
universe = vanilla
should_transfer_files = YES
when_to_transfer_output = ON_EXIT
requirements = ( Arch=="X86_64") && ( OpSys=="LINUX" )
executable = hostname.sh
output = host.out
error = host.err
log = host.log
notification = Error
Note: executable must be executable!
For more examples see http://www.cs.wisc.edu/condor/quick-start.html
Information about Globus jobmanager-condor
cp /home/globus/Growl/packages/gt4.2.1-all-source-installer//source-trees/gram/jobmanager/setup/condor/condor.in /usr/local/globus-4.2.1/lib/perl/Globus/GRAM/JobManager/condor.pm
or wherever you have the installed job managers. Edit the BEGIN initialiser as follows:
# RJA done this config by hand
$condorbin = '/opt/condor/bin';
$condor_submit = "$condorbin/condor_submit";
$condor_rm = "$condorbin/condor_rm";
$condor_config = '/opt/condor/etc/condor_config';
You now need to add some files:
For additional information on this see http://www.cs.wisc.edu/condor/condor_globus.html
You can set the ephemeral port range used by the Condor jobmanager to the same as Globus by setting LOWPORT=64000 and HIGHPORT=65256 in the local config file.
Topic revision: r7 - 23 Feb 2010 - 09:17:20 - DaveCable