Overview of Portal Job Submission
Prior to submitting jobs via the Grid Engine Portal, some basic information about projects and directories is necessary. When a user logs into the portal, a predetermined list of applications that are available is proffered. In order to submit jobs, a project must be made from the application, i.e. the application must be instantiated. The first time a project is made, a new directory is created in the users home directory, this directory suntcp will hold all the files related to each project you create such as output.txt and error.txt. Also, there will be three files, .suntcp-qsub, .suntcp-project and .suntcp-su. Do not remove these, they are necessary for proper job submission. See below for an example of HOME directory structure.
$HOME/
/.bash_history
/.bashrc.d/
/suntcp/ <--- Directory automatically created for all projects
/.suntcp-list*
/P1052500340231/ <--- Project Directory
/P1053109842674/ <--- Project Directory
/.suntcp-project
/.suntcp-su*
/.suntcp-qsub
/output.txt
/error.txt
Once you have created a project, you may actually submit a job. This is done via the portal as well. Most output created by that job will exist within the projects directory in your $HOME/suntcp directory. Depending on the script you have submitted, you can output files elsewhere. For example, ncverilogs' log file, .nclog could written into the directory in which you submitted script is executed.
- Step 1: Point Browser to URL: https://hpcsrv1.hpcvl.queensu.ca. Use a reasonably modern browser.
- Step 2: Login using Username and Given Password.
- Step 3: Accept Certificate.Grid Engine Portal page will load.See sample of Front Page.
- Step 4: Click on 'Grid Engine Portal' tab. This will bring up a listing of available applications to use. See sample of Submit Page. This page has three main channels. The bottom box lists the applications you may use in order to submit jobs to the Grid Engine Portal. Currently, there is only one method of submitting jobs. In a nutshell, this method is getting the Portal to execute a user's script file which exists in their HOME directory on the Sun Fire Cluster.
- Step 5: In order to submit jobs through the Grid Engine Portal, we must first create an instance of an application. This is called a Project. You may have numerous versions of each Application, i.e. multiple projects for each application. The idea behind this is that you can customize each project such as setting different Environment variables for each project.See sample of Project creation here.
- Step 6: The final step in submitting a job through the Grid Engine Portal is the creation an instance of a Project. See example of job submission.. In the example, the user enters a path to the script file. Typically, this is within the user's HOME directory. Secondly, the user enters the name of the executable Script. In the example, running 'make' suffices, but this this is completely user dependant. Please Note: The appropriate environment setting scripts are automatically run prior to the job executing. This should take care of most PATH issues as well as checking out the correct Licenses.
- Step 7: The now submitted job will show up in the Job List panel as being submitted. When the job has been completed, this listing will automatically be removed. Once this job has finished, clicking on the Project Name, i.e. cmc_test2, will give a listing of everything in the project directory. Please see example of Project Listing. Notice that amongst the files is an error.txt file which is useful for debugging. Also, there is an output.txt which is unsurprisingly, an output file for the job. Any files listed here are downloadable by clicking/right-clicking on the file name.
|