Difference between revisions of "Cluster:Using the F@C PBS script"

From Earlham CS Department
Jump to navigation Jump to search
 
(One intermediate revision by the same user not shown)
Line 1: Line 1:
* The F@C PBS script lives in <tt>/cluster/generic/bin/fatc_pbs.sh</tt>. To use it, you have to specify the number of nodes, the processors per node, the molecule, the database table name, and the test set label. You should also set redirection files for STDOUT and STDERR. Do so like this:
+
* The F@C PBS script lives in <tt>/cluster/generic/bin/fatc_pbs.sh</tt>. To use it, you have to specify the number of nodes, the processors per node, the molecule, the database table name, and the test set label. Do so like this:
  
qsub -l nodes=n:ppn=m -v MOLECULE=molecule_name,TABLE_NAME=table,TEST_SET_LABEL=label -o out.txt -e err.txt /cluster/generic/bin/fatc_pbs.sh
+
<tt> for((i=1;i<13;i++));do qsub -l nodes=$i:ppn=2 -v MOLECULE=$molecule,TABLE_NAME=results_skylar,TEST_SET_LABEL=Parse_Test,FATC_RELEASE=fatc-a3.1-linux-powerpc.gz,FATC_DIR=fatc-a3.1 -N fatc.$molecule.nodes$i /cluster/generic/bin/fatc_pbs.sh;done</tt>
  
* Make the submission from the head node (probably a0, b0, or c0) of that cluster. Make sure that your current working directory is inside a fatc directory.
+
* This for loop simulates the behavior of the run-fatc script on the cairo cluster. You should adjust the upper bound and lower bounds of the loop as needed of the loop as needed. You of course should also adjust TABLE_NAME, TEST_SET_LABEL, and FATC_RELEASE as needed. Remember to export $molecule before doing the run.
 +
 
 +
* Make the submission from the head node (probably a0, b0, or c0) of that cluster. You no longer have to be in a fatc directory to make the submission; the submission script automatically generates and cleans it up.
  
 
* The log parsing is done from within the submission script, and is stuck in the fatc-testing database.
 
* The log parsing is done from within the submission script, and is stuck in the fatc-testing database.
 +
 +
* For general information on using PBS/Maui, see [[Using PBS/Maui|here]].

Latest revision as of 17:44, 20 July 2005

  • The F@C PBS script lives in /cluster/generic/bin/fatc_pbs.sh. To use it, you have to specify the number of nodes, the processors per node, the molecule, the database table name, and the test set label. Do so like this:

for((i=1;i<13;i++));do qsub -l nodes=$i:ppn=2 -v MOLECULE=$molecule,TABLE_NAME=results_skylar,TEST_SET_LABEL=Parse_Test,FATC_RELEASE=fatc-a3.1-linux-powerpc.gz,FATC_DIR=fatc-a3.1 -N fatc.$molecule.nodes$i /cluster/generic/bin/fatc_pbs.sh;done

  • This for loop simulates the behavior of the run-fatc script on the cairo cluster. You should adjust the upper bound and lower bounds of the loop as needed of the loop as needed. You of course should also adjust TABLE_NAME, TEST_SET_LABEL, and FATC_RELEASE as needed. Remember to export $molecule before doing the run.
  • Make the submission from the head node (probably a0, b0, or c0) of that cluster. You no longer have to be in a fatc directory to make the submission; the submission script automatically generates and cleans it up.
  • The log parsing is done from within the submission script, and is stuck in the fatc-testing database.
  • For general information on using PBS/Maui, see here.