Before you begin, if you plan to join the UW-Madison Computer
Science Condor pool, you must send email to condor-admin@cs.wisc.edu
and give us the hostnames of the machines you intend to add to our
pool, your name, email address, and a phone number where we can reach
you. Without this infomation, we won't allow your machine(s) to
join our pool and you won't be able to submit jobs and have them run.
The administrator of the Central Manager of a given Condor pool
(condor-admin@cs.wisc.edu in our case) has to specifically add
hostnames to the Condor config files to allow machines to join
the pool.
NOTE: When installing as a submit-only node in a Condor
pool, you won't be able to submit "vanilla" jobs, since vanilla jobs
almost always depend on a shared filesystem between the execute and
submit machines. This is certainly not going to be the case for your
machine and our machines in our pool. Read section 2.5 on
"Job Preparation" from the Version 6.2 User's Manual
for details on vanilla and standard jobs.
For the purposes of these instructions, we'll assume you are setting
up a single machine, of a single platform, for a single user to submit
jobs into the UW-Madison pool. If this isn't the case, some of the
answers to the questions explained below will be different. For
details on all the questions and possible answers to them, consult the
INSTALL file provided with the binary distribution, or the
Administrator's Manual, available on the world wide web at:
http://research.cs.wisc.edu/condor/manual/
You should also read over the overview of Condor and the User's
Manual to learn how to use Condor once you have completed the
installation.
For the purposes of these instructions, we'll use "condor-6.2.0.irix6-SGI.tar.Z" as the filename you downloaded.
Now, untar the binary distribution:
% tar -zxvf condor-6.2.0.irix6-SGI.tar.ZIf your version of tar doesn't support the "-z" option, you can do this, instead:
% cat condor-6.2.0.irix6-SGI.tar.Z | uncompress | tar -xvf -This will create a "condor-6.2.0" directory. Go into that directory now:
% cd condor-6.2.0Now, run the condor_install perl script:
% ./condor_installIf perl isn't installed in /usr/bin/perl, you might have to start up perl explicitly, like this:
% perl ./condor_install
like thisTo accept the default listed in []'s, just hit enter. For responses that ask for a path or username, this sheet might contain "username" in the appropriate response. However, you should substitute your username for "username" when you actually type in your responses.
First, it will display a screen of information. Read it, then hit enter to begin the installation.
Continue with submit-only installation? [yes]
yes
Are you planning to setup Condor on multiple machines? [yes]
no
Have you installed a release directory already? [no]
no
Where would you like to install the Condor release directory?
[/home/username/condor]
[hit enter to accept the default, which is a directory called "condor" in your home directory].
If this directory doesn't exist, the script will create it for you. The "release.tar" file that comes with the binary distribution will be untarred into the directory you specify. This will create the "bin", "sbin", "lib" and "etc" directories, containing all the Condor daemons, tools, libraries and supporting files.
For installations involving multiple users or multiple machines, this directory can be shared. If it is going to be shared, don't use the same directory for the "Condor needs some directories for its own use..." question below.
If something goes wrong with Condor, who should get email about it?
[hit enter to accept the default, which is your user name, at the machine you are using].
What is the full path to a mail program that understands "-s" means
you want to specify a subject? [/usr/bin/mail]
[hit enter to accept whatever default is given]
When I'm done, I'm going to need to install some scripts and links.
...
Where should I install these files?
[/home/username/bin]
[hit enter to accept the default, which is the "bin" subdirectory in your home directory].
If you don't have a bin directory already and/or it's not in your PATH, condor_install will complain that the directory isn't in your PATH. Tell it you want to use that directory anyway. If the directory doesn't exist, the script create it for you. If you are only joining one pool, you won't need this directory anyway, and you can delete it and all the files within it once condor_install has completed.
Condor needs to create some directories for its own use. Where would
you like to put these directories?
[/home/username/condor]
[hit enter]
This directory is important, and you will not be able to remove it at any point while Condor is running on your machine. If it doesn't already exist, the script will create it for you. This directory cannot be shared among multiple users. It is referred to as the "local directory" or "LOCAL_DIR" in the manual and config file.
How many different pools would you like to add this machine to? [1]
1
What is the full hostname of the central manager?
...
[condor.cs.wisc.edu]
[hit enter]
Where do you want to put Condor's configuration file for this pool?
[/home/username/condor/etc/condor_config]
[hit enter]
This should be the "etc" subdirectory of the LOCAL_DIR you specified above.
What name do you want to use for your schedd in this pool?
(If you don't understand this question, just accept the default.)
[hit enter]
That's the end of the installation.
If you are only joining our pool, simply set the CONDOR_CONFIG environment variable to point to the config file you specified above, something like "/home/username/condor/etc/condor_config". You do this with either:
setenv CONDOR_CONFIG /home/username/condor/etc/condor_configor
CONDOR_CONFIG=/home/username/condor/etc/condor_configdepending on your shell. You will probably want to put this into your "dot file", such as ".cshrc", which gets loaded automatically when you log in. In addition, you will want to make sure that the "bin" directory of the Condor release directory is in your path:
setenv PATH $PATH:/home/username/condor/binor
PATH=$PATH:/home/username/condor/binNow, start up your personal Condor daemons:
% condor_master
That's it. You can view your Condor daemons now with the ps command:
% ps uwx | grep condor_or
% ps -ef | grep condor_depending on your flavor of UNIX.
Try running "condor_status" to see the status of the UW pool:
% condor_status -total
If you this works, you should be all set to start submitting jobs.
You will need to compile Solaris binaries of your jobs, since the overwhelming majority of machines in our pool are Solaris machines, both Sparc and Intel Solaris. If you do not have access to either kind of Solaris machine to compile your programs, contact condor-admin@cs.wisc.edu. There are also a few Alphas in our pool running Digital Unix, so if you'd like to submit binaries for that platform, that will work as well.
Do not try to submit Linux or SGI jobs to our pool. Even though you might see some Linux and SGI machines in the pool, these machines are on a network that cannot reach the outside world, so your jobs will not run.
If you have any questions or problems with this, please contact
condor-admin@cs.wisc.edu and your email will get entered into an
automated tracking system so we can assign your request to one of the
Condor developers and we will be sure to follow up with you.