Building a Pool with a Shared Release Directory

Now, we're going to create a pool with a shared release directory. infn-corsi98 will remain the central manager. However, we will now use binaries and config files installed on NFS instead of the local versions you installed earlier.

In the last excercise, we completely shutdown all the Condor daemons on your machine. Now, we're going to reconfigure your machine to use a release directory installed on nfs.

All you have to do is point your condor_config file to the one on NFS, and that global config file has all the correct settings for the global pool:

% cd ~
% rm condor_config
% ln -s /usr/local/condor/nfs/etc/condor_config .

Now, if you use condor_config_val, you can see the settings for the new pool:

% ccv DAEMON_lIST

Notice that you now only have MASTER, SCHEDD, and STARTD. There are no longer any entries for COLLECTOR or NEGOTIATOR. Let's see what file is being used as the local config file for your machine:

% ccv LOCAL_CONFIG_FILE

Now, lets look at the DAEMON_LIST for the central manager:

% ccv -name infn-corsi98 DAEMON_LIST

You'll notice that this includes everything, just like you'd expect.

Finally, you should start up the condor_master again, so that we actually have all the machines in the pool. We'll use another trick here. In UNIX, if you enclose commands in "back-ticks", using the ` mark, the output of the command is taken and used in the calling command. So, we'll ask condor_config_val the path to the condor_master, and give that directly to the shell to have the shell start up the condor_master:

% `ccv MASTER`

You can run ps again to see the Condor daemons running:

% ps auwwx | grep condor_

Now, run condor_status to see the status of the big pool again:

% condor_status

That's it! You're done.