Australian Research Collaboration Service

This page is a very simple walk through launching two simple jobs from Grisu. You must have already obtained a certificate and be in the StartUp Virtual Organisation. All Australian academic certificate applications are automatically put into StartUp when issued.

First Time Setup

When you start Grisu for the first time (or after removing your grisu settings directory) there are currently a few steps you need to to get started. (Before too long, some or all of these steps will be automated using data distributed via MDS.)

1. The first grisu page you see has two fields. First, set the lower one to the VPAC Grisu Server. Second, enter your (very secret) pass phrase in the upper one. Press Login.
2. Next you see a dialogue that asks about what sites you will launch from. For now, select 'Brecca'.
3. Now change your VO (virtual organisation) and set it to 'StartUp'- this grants you some limited access to run test jobs. Most new users will only be able to choose StartUp but you will probably want to be in other VOs later.
4. Grisu now needs to be restarted to makes sure the new mount points and VO memberships work cleanly. Login again.

Running your first Job

1. After you restarted Grisu and logged on again, you'll see a screen a bit like the one shown here. We are going to load a totally trivial application template and run a totally trivial job.
2. Click on the add (+) button below the (currently empty) Applications list. You will see a list of ten or more templates listed. Choose generic_no_mds and press Add. "Generic" means you can run almost any job available on the machine, and "no_mds" means that it won't be filtered through the ARCS Information Service that normally checks the availability of things.
3. Click on the generic_no_mds entry you just added to Applications. Some early Java versions have a problem here- if you don't see some content appear on the right hand side, just click there. Things will appear ...
4. You now have two tags on the right, General and Rest. Select Rest. In this particular template, you can type in the command you want to run in the top field. Type "date" (without the inverted commas). You don't need to set any modules nor upload any files.
5. Select the General tag. Here, you set some general things that are not dependant on the template in use. Change the wall time to three minutes, the remainder is probably OK. Its not an MPI job, the Site is VPAC Brecca, the Queue is dque@brecca-m, and you can change the job name if you wish.
6. Click on Submit ! You'll see some progress reports, and it will run almost immediately in most cases.
7. Move to the (top of Grisu screen) Monitoring tag. Here you see a list of jobs you have run, probably just one there, and double-click on it to see the files generated. Under JobDirectory, you should see the four files the job generated, this is the supercomputer's scheduler, PBS, stuff, not grid by the way. The interesting (?) one is stdout.txt, double-click on it to see its contents in the right-hand panel.
8. Congratulations- it shows you the date and time on the machine in question. That's probably the most expensive and complicated clock you have ever used !

Running your second Job

You could be justified in thinking that you just went to extraordinary lengths to find out the time. I guess that's the nature of demonstrations; the thing we are doing is not of any importance, its just that we are, in fact, doing it. For this, the second demo, we'll do just a bit more, still just using a generic Unix command but we will upload a couple of files, compare them and then get an output file which shows us the difference between them.

1. Click the plus button under the application panel again, this time select "diff" and press "add" to add it to your list of applications. Click on the entry you just added to applications, and, if you have that Java bug, click the right-hand-side of the grisu window as well.
2. Again you see two tabs in the right-hand-side; General has not changed, but Rest has. Its a customised panel specifically for the Unix diff command. It's quite easy to make customised templates for what ever command you might be planning to use.
3. Diff requires two text input files from your desktop computer that it will compare. Press the Browse buttons and find some suitable files. [[ David - we must provide two suitable files for people to download, not everyone will necessarily have something handy]] If you are a unix/linux user and familiar with the diff command, feel free to put some command line switches in the upper field (you don't need to however).
4. The General Tab will be fine as it is, just press Submit.
5. Again this job will run in a couple of seconds (unless you selected some very big files!), and you can move to the Grisu Monitor tab to see progress. Maybe click the refresh button...
6. Now you see a number of files, the four generated by PBS plus the two you uploaded to be worked on. And again, your output is in stdout.txt. Right-click to view or download, your choice.
7. When you have seen all you need, please be a good user and clean up your files from the Monitoring tab.

Where to from here ?

While the above two demonstrations were trivial in the extreme, what you have seen is a framework for launching much more complicated jobs. The templates are easy to customise and very versatile. You can easily make your own or ARCS will be glad help. ARCS already has made a number of templates for more serious jobs, for example the namd template is routinely being used to launch 64 cpu jobs running for many days. It features a nice little graphical viewer you can use to watch the progress of your job.