Running and Verifying SGE

Procedure

1. Use PuTTY to log in to the SGE master host as the root user.
2. Run the following command to add the execution host:

   qconf -as armnode4

3. Query the default host name.

   qconf -shgrpl

   @allhosts

4. Modify the host group information and add the execution host to the host group.
   1. qconf -mhgrp @allhosts
   2. Press i to enter the insert mode and enter the following information:

      group_name @allhosts
      hostlist armnode4

   3. Press Esc, type :wq!, and press Enter to save the file and exit.
5. Switch to a non-test user.

   su - test

6. Load the SGE environment variables as the test user.

   source /path/to/SGE_ROOT/default/common/settings.sh

7. Add the SGE environment variables to the .bashrc file of the test user for the environment variables to take effect permanently.

   echo "source /path/to/SGE_ROOT/default/common/settings.sh" >> /path/test/.bashrc

8. Create an execution script run.sh.

   Run the vi run.sh command.

   Press i to enter the insert mode and add the following information:

   #!/bin/bash
   #$ -S /bin/bash
   nodeinfo=`hostname`
   echo "This is the SGE test from $nodeinfo" >> sge-test.log

   Press Esc, type :wq!, and press Enter to save the file and exit.

9. Submit a job.

   qsub -V -cwd -o stdout.txt -e stderr.txt run.sh

   Table 1 describes the command parameters.

   Table 1 Common parameters in the qsub command

   Parameter	Description
   -V	Exports the environment variables in the current shell to the job to be submitted.
   -cwd	Runs the program in the current working directory. By default, it is the home directory on the compute node of the current user.
   -o	Adds the standard output to the end of the specified file. The default file name is $job_name.o$job_id.
   -e	Adds the standard error output to the end of the specified file. The default file name is $job_name.e$job_id.
   -q	Specifies the queue to be delivered. If this parameter is not specified, the system searches for the queue with the permission and minimum load to execute the job.
   -S	Specifies the software that runs the commands in run.sh. The default value is tcsh. You are advised to use bash. Set this parameter to /bin/bash, or add "#$ -S /bin/bash" at the beginning of the run.sh file. If the parameter is not set to bash, the output contains "Warning: no access to tty (Bad file descriptor)."
   -hold_jid	Specifies the jobs to be executed before the current job. A comma (,) is used to separate multiple job_ids.
   -N	Sets the job name. The default job name is the input file name of qsub.
   -p	Sets the job priority. The value ranges from -1023 to 1024. A larger value indicates a higher priority. However, a higher permission is required to set this parameter to a positive number. Common users cannot set this parameter to a positive number.
   -j y|n	Specifies whether to merge the standard output and standard error output streams into the -o parameter result.
   -pe	Sets parallel environment (PE) information.

10. Manage jobs. Table 2 describes the commands and parameter for job management.

    Table 2 Commands and parameter for job management

    Command	Description
    qstat -f	Queries all jobs submitted by the current user on the current node. The job status can be: qw: waiting state. A job is in qw state after being submitted. The job runs immediately once computing resources are available. hqw: pending state. A job in this state starts to run only after the previous jobs are executed. The job submitted by using qsub with -hold_jid specified is in this state. Eqw: job waiting in error state. r: running state. s: job suspended temporarily because resources are used by the job with a higher priority. dr: job deleted after a node exits unexpectedly. The job disappears only after the node is restarted.
    qstat jobID	Queries information by job ID.
    qstat -u user	Queries information by user.
    qdel jobID	Deletes a job.