You will find the Clust Name Space daemon that must run on each data server in
the bin subdirectory of this package. It is called XrdCnsd. Some architectures
have been pre-compiled. If you do not find the one you need or you wish to
recompile the xrootd package. The easiest way to do this is to first execute
"configure.classic --build=debug --disable-mon --disable-perlint" on the host
for whose architecture you would like to have binaries. The "disable" entries
turn off compilation for components we have found to be problematic on some
architectures.

Typiclly, you will need two configuration files. The first is used for the
standard xrootd cluster (i.e., the redirector and all of the data servers).
The file "example.xrdcluster.cfg" gives you a template typically used for the
Atlas experiment. It equally applicable for other experiments. After modifying
the file and distributing it to all nodes, start the xrootd and cmsd with the
configuration file in the start way (see StartXRD and StartCMS scripts in the
Xrd/Xrootd Reference): 

http://xrootd.slac.stanford.edu/doc/xrd_config/xrd_config.htm

For very simple installations you can use a stripped down version of the start
scripts that will start xrootd and cmsd in one go assuming standard defaults.
See the xrdcluster.sh in this directory.

The second configuration file is used when you start the xrootd that maintains
the shadow namespace for the cluster. This xrootd usually runs on the same
host as the redirector but uses a different port number and log directories.
The file "example.cns.cfg" gives you a template typically used for the
Atlas experiment. It equally applicable for other experiments. After modifying
the file, use it to start the "special" namespace xrootd. Again, you can use
the StartXRD script (specifying a unique instance name via the -n option) or
the simpler all-in-one start-up script in this directory called "cns.sh".
