How to Submit PAUP Jobs in the Standard Universe (Checkpointing)

Prepare a PAUP batch file

For this example you'll need two files -- one containing a NEXUS dataset and the other containing the PAUP commands. The data set looks like this:

#NEXUS 
Begin data;
        Dimensions ntax=8 nchar=200;
        Format datatype=dna interleave;
        Matrix
A CGAATATAACGGAGCCAGTACTCAGACGCACTGCCAACCCAGCGAAGCCCGATACGCCGT
B CGAATATAACGAAGCCAGTATTCAGACGCACTGCTAACCCAGCGGAGCCCGGTACGCCGT
C CGAATATAACAAAGCCAGTACTCGGACGCACTACCAACCCAGCGGAGCCCGATACGCCAT
D CGAATACAACAAAGCCAGTATTCAGACGCACTGCCAACCCAACAGAGACCGGCGTGCTAT
E CGAATACAACAAAGCCAGTATTCAGACGCACTGCCAACCCAGCAGAGACCCCCACGCTAT
F CGAATACAACAAAGCCAGTATTCAGACGCACTGCCAACCCAGCAGAGACCCACACGCTAT
G CGAATACAACAAAGCCAATATTCAGACGGACTGCCAACCCAGCAGAGACCGACACGTCAT
H CGAATACAACAAAGCCAATATTCAGACGGACTGCCAACCCGGCAGAGACCGACGCGTCAT 

...  

;
end;

Download a copy of the sample data file here.

The commands used to run a specific analysis are kept in a separate NEXUS file, which will reference the data set given above.

#NEXUS;

Begin paup; 
    log file=paup.log replace;
    execute example_data.nex;
    
    [build a tree using the neighborjoining algorithm]
    nj; 
    savetrees file=paup.tre replace;      
    quit;  
end; 

Copy the paup block given above and paste it into a new file named example_paup.nex.

Test the batch file

Before launching your job under condor, test the batch file at the console to make sure that it is working properly. At this point you should have two file: example_data.nex and example_paup.nex.

Type:

condor_paupdev example_paup.nex

Because this is a short analysis, the program should execute and terminate within a second or two, saving a single tree and log file to the current directory. In reality, you will be testing analyses that might run several hours or days before completing. If this is the case, you will want to terminate the analsysis after making sure that PAUP properly executes the file. To interrupt a PAUP process, simply press control+C.

Create a submit file

Now that you know you paup job will run without errors, you need to create a condor submit file. More specific information is on how to create a submit discription file is given in the UsingCondor topic.

########################################
#
# PAUP run in the Condor standard universe
#
#########################################

Universe     = standard  
              
Executable   = /usr/common/i686-linux/bin/condor_paupdev
             # Use command "which condor_paupdev" to find out the path of "condor_paupdev"
Arguments    = example_paup.nex -n -f
             #  -n   Run in "non-interactive" mode (no prompts)
             #  -f   Ignore file-locking 

output       = example_paup_condor.out
error        = example_paup_condor.error
log          = example_paup_condor.log

Queue

Copy the text given above and paste it into a new file named example_paup.cmd.

Logon to an SCS submit node

SCS maintains several submit nodes that give users a way to access SCS computer resources. The general access submit node is named phoenix. There are also two other submit nodes (anfinsen and petal), which are part of restricted access research clusters. Special permission from the resource owner is required to access the petal and anfinsen submit nodes.

$ ssh <username>@phoenix.scs.fsu.edu

Submit the job

To submit a job to the condor cluster you will use the condor_submit command. For example:

$ condor_submit example_paup.cmd

You should see the following output:

 
Submitting job(s).
Logging submit event(s).
1 job(s) submitted to cluster 56776.

Check the status of a job

After submitting your job to the condor cluster you can check on the status of your job by using the condor_q command. For example:

$ condor_q <your user name>

The output from this command should look something like this:

-- Submitter: petal017.csit.fsu.edu : <144.174.160.147:10076> : petal017.csit.fsu.edu
 ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD               
56776.0   yanfeng         5/22 18:22   0+00:00:00 R  0   1.8  condor_paupdev exa

1 jobs; 0 idle, 1 running, 0 held

Don't be surprised if your job remains idle (designated by an I under ST) for serveral minutes or longer. If your job does not run right away it most likely means that you have a low priority on the cluster and the cluster is being heavily utilized or it may mean that someone job with a lower priority is taking a while to vacate a node so that your job can run. Remember, condor is based on a High Throughput Computing (HTC) model and not a High Performance Comuting (HPC) model.

You can also see what has happened to your job by looking at the condor log file. Remember the condor log file was defined in the condor submit file. To look at the file you might use the cat command. For example:

$ cat example_paup_condor.log 

The output from this command should look something like this:

000 (617.000.000) 10/19 14:34:02 Job submitted from host: <144.174.160.169:11297>
...
001 (617.000.000) 10/19 14:34:32 Job executing on host: <144.174.160.207:9705>
...
005 (617.000.000) 10/19 14:34:32 Job terminated.
        (1) Normal termination (return value 1)
                Usr 0 00:00:00, Sys 0 00:00:00  -  Run Remote Usage
                Usr 0 00:00:00, Sys 0 00:00:00  -  Run Local Usage
                Usr 0 00:00:00, Sys 0 00:00:00  -  Total Remote Usage
                Usr 0 00:00:00, Sys 0 00:00:00  -  Total Local Usage
        3346  -  Run Bytes Sent By Job
        2000108  -  Run Bytes Received By Job
        3346  -  Total Bytes Sent By Job
        2000108  -  Total Bytes Received By Job
...

Moving on

After the analysis is complete, you should receive and email with more information regarding the execution of our job. In addition, the file that PAUP created during the analysis will be located in your directory. You will also find the the .out and .error files in your directory. These files contain the standard out and standard error generated by the executable.

This is a barebones example submit file. See the UsingCondor topic for more information on creating submit files.