Sorry for the inconvenience, but this page is currently being updated.
Below are the instructions for submitting jobs to LoBoS2, the ones for LoBoS4 will be up shortly



Submitting Jobs

In general, users prepare a script to be executed on LoBoS and then submit it with the lobosq2 command. A typical script includes directives for the queue manager, followed by unix level commands which are executed when the job is run. A user's interaction with LoBoS will normally occur in the following sequence:

  1. User submits a job with the lobosq2 command
  2. A file is created in the current directory with the form username.yymmddhhmmss
  3. The job file is then copied (rcp) to the master nodes and scheduled.
  4. When the job begins execution the file username.yymmddhhmmss.run will be created.
  5. Output from the job begins to appear in username.yymmddhhmmss.log, which may be monitored.
  6. Upon completion or cancellation username.yymmddhhmmss. run is deleted.

The script inherits the following environment during execution:

  1. Execution occurs in the directory where the user issued the lobosq2 command
  2. Environment variables NODE0, NODE1,..NODEn which are assigned the symbolic name (pe1,pe2 etc.) of the processors where the job is executing.
  3. All disk space requirements impact the user's local disk, unless the user specifies use of temporary space on the master or compute nodes.

To Submit a job

Type the commands interactively:

lobosq2 sub <CR>

Enter the commands

<ctrl>-d

or redirect the input file directly:

lobosq2 sub < filename <CR>

To check on job status

Use the username.yymmddhhmmss.log file to monitor the job's output during execution

Type:

lobosq2 list <CR>

To cancel a job

Type:

lobosq2 can username.yymmddhhmmss <CR>