Testing BOSCO

Here's a tarball: ftp://ftp.cs.wisc.edu/condor/temporary/bosco/2012-01-25/condor-7.7.5-UW_development-rhel5.7-stripped.tar.gz

To install, untar, cd into the top directory, and run ./condor_install --bosco Then set CONDOR_CONFIG=$HOME/bosco/etc/condor_config in your environment and add $HOME/bosco/bin and $HOME/bosco/sbin to your PATH. Then run condor_master.

Then try running bosco_add_host and submitting a job.

Questions

What if the home directory is shared? What if you can already login? Still add bosco key/pass because the auth may be temporary Tell that that bosco key is added to auth keys Tell the abs path of glite directory What if there is already a Condor running? Choose a different port ? use the system collector and negotiator?

Reinstall to verify if the glite dis is overwritten or if it is a new one in the same path

/home/marco/tmp-bosco/condor-7.7.5-UW_development-rhel5.7-stripped/libexec/glite/etc/glite-ce-blparser
/home/marco/bosco/libexec/glite/etc/glite-ce-blparser
/home/marco/bosco/glite/etc/glite-ce-blparser

Shared directories

File to change in the original BLAH bundle (wget http://jra1mw.cvs.cern.ch/cgi-bin/jra1mw.cgi/org.glite.ce.blahp.tar.gz?view=tar&pathrev=glite-ce-blahp_R_1_16_3_3): org.glite.ce.blahp/config/blah.config.template

Uncomment line and set it to $HOME (replaced by an install script) + other common directories:

#blah_shared_directories=/home:/users

My preference goes for option 1 or 2 since all files have to be assumed shared

Options

Change the function in src/scripts/blah_common_submit_functions.sh:

function bls_test_shared_dir ()
{
  local file_name
  local test_dir
  local shared_dir_list

  file_name=${1:?"Missing file name argument to bls_test_shared_dir"}

  bls_is_in_shared_dir="no"

# Test only absolute paths
  if [ "${file_name:0:1}" != "/" ] ; then
      return
  fi

  #TMP:BOSCO Assuming that all directories are shared
  bls_is_in_shared_dir="yes"

#  shared_dir_list="$blah_shared_directories"
#  test_dir=""
#
#  until [ "$shared_dir_list" == "$test_dir" ]; do
#    test_dir=${shared_dir_list%%:*}
#    shared_dir_list=${shared_dir_list#*:}
#    if [ "${file_name:0:${#test_dir}}" == "$test_dir" ]; then
#      bls_is_in_shared_dir="yes"
#      break
#    fi
#  done 
}

Change the config file template config/blah.config.template (v1):

#blah_shared_directories=/home:/users
blah_shared_directories=/

Change the config file template config/blah.config.template (v2):

#blah_shared_directories=/home:/users
tmp1_blah_shared_directories="$HOME/"
tmp2_blah_shared_directories=${tmp1_blah_shared_directories#/*/}
blah_shared_directories=${tmp1_blah_shared_directories:0:${#tmp1_blah_shared_directories}-${#tmp2_blah_shared_directories}-1}

PBS info

Maui/Torque

Mohab/Torque

See MarcoPbsNotes

TODO

- traceroute for the job
are jobs running
is my job moving 
is anybody moving?
which PBS commands do that?
How other queue mgr (LSF, SGE) do that?



Available clusters:
- PADS, Beagle Torque + Moab
- ITB, (MWT2_UC, TeraPort), Torque + Maui
- Ask Joe U for SGE access


Scalability and unit testing

-- MarcoMambelli - 14 Dec 2011