  Cleo is a batch processing system for miltiprocessors or/and clusters
(queue system).

  This software is made for Joint SuperComputer Centre (Moscow Russia)
and for Science Research Computer Centre of Moscow State University
(Moscow Russia). Author of this software is Sergey Zhumatiy (serg@parallel.ru).

  You can use it ONLY after contacting author and getting the agreement.
It is free.

  Please, see the LICENSE and GPL files.

INSTALLATION

  1. You must have perl5 or higher installed.

  2. Ensure, that all your header files are converted to perl or just run
      cd /usr/include; h2ph -r -l .

     If some header file fails, then simply MOVE it in other place, run h2ph
       again and tthen return header file back.

  3. Edit Makefile.conf file, to set proper directories for
     installation. You also can edit files owner and permissions.

  4. If you are ugrading from version 4 or earlier and want to save
     old executables, run 'make backup'. This saves your old
     executables with '.version4' extensions. Then you must copy current
     'tasks' and 'mpirun' executables to safe place.

  5. If you want to save compatibility with old commands names (such
     as qmode), you can run 'make compatible'.

  6. Save your original 'mpirun' script, if it is placed in the same
     dir, as cleo executables will be installed. Cleo will use its own
     mpirun script, but you must save the original.

  7. Run 'make', then 'make install'

  9. If you didn't use automatic copying to nodes, copy
     /usr/local/bin/cleo-mon to the same dir on computational nodes,
     and file /etc/init.d/cleo-mon.rc to /etc/init.d/cleo-mon on nodes.
     Then you must create directory /usr/local/share/Cleo (if you had
     changed SHAREDDIR, then $SHAREDDIR/Cleo), and copy
     /usr/local/share/Cleo/Conn.pm on every node.

  10. Link /etc/init.d/cleo.rc to your system start script directory(s)
     with accordance to used runlevel. On different systems it may be
     /etc/rc3.d/S99cleo, or /etc/init.d/rc3.d/cleo, etc...

  11. Link /etc/init.d/cleo-mon.rc on computational nodes to
     your system start script directory(s).


  Now customize your /etc/cleo.conf file (see documentation), and start
monitors, if you need (/etc/init.d/cleo-mon.rc start) on nodes.
  You can use saved mpirun script to run mpi appliactions from
cleo. To prevent run this script directly, you can change its group to
special group (like mpi), permit to read and execute it only for this
group, and specify this group in cleo.conf by 'group = ...' in
[global] section.

  To start system just type '/etc/init.d/cleo.rc start'.


TUNING

  You can tune the system by adding new modules and scedulers. You can
read about creating new scedulers in file README-sceduler-create. An
example sceduler is presented by file example-sceduler, you can use
it. Also you can use base_sced - the default sceduler.
  You can create new modules, which are called on each task run or
end. Detailed info and simple examples may be found in file
Modules-howto. For example, you can copy some files on nodes before
task run and clean them after task end, etc.

