How-To: Access Hellbender Open OnDemand

How-To: Access Hellbender Open On Demand?

The Open OnDemand (OOD) service on the Hellbender HPC cluster is a web-based service that enables convenient graphics-based user interface (GUI) to the Hellbender cluster. 


Unlike the terminal based ssh access to Hellbender which requires that your connecting computer either be on a campus network, or use a VPN, or that you have an ssh key registered with your account – connection through OOD does not require any of that – it requires only your CILogon credentials. 


Even though OOD is very convenient, it requires an additional layer of service that needs to be running to access the Hellbender. Namely, if the OOD web server is down for some reason, you might still be able to access the cluster from the command line terminal and ssh command. Also, not all the HPC resources you might use in the future will have OOD installed while all of them will for sure have the ability of ssh connection. For that reason, we advise you to learn how to operate on a cluster via simple terminal – in Windows operating system we recommend using MobaXterm for that purpose. 


You can access Open OnDemand for different purposes which will be demonstrated here

  1. Upload/download files
  2. Submit jobs from the login node
  3. Get an interactive session with different applications – the most generic being Hellbender Desktop

 

Access Hellbender’s Open OnDemand Webpage

To access Hellbender’s Open OnDemand web page, open the following link in your browser:

https://ondemand.rnet.missouri.edu/pun/sys/dashboard

You should see something like this (using Firefox browser on Windows OS):

Select your identity provider and click Log On button – This should bring you through the authentication workflow for your home institution (or ORCID).

If successful, you should get to the following main OOD page


If you get a page response like this one:

You will need to delete the cookies in your browser for the OOD site – or as a simpler work around open an incognito/private browser window.


Once in the OOD webpage you can click Files:

And then Home Directory which will lead you to the following page:

This will by default show the content of your home directory – you can navigate to different subdirectories, and you can use Download or Upload buttons to either download files from the Hellbender to your local computer or to upload files from your local computer to Hellbender.


Going back to the main OOD page when you click on Clusters you will see a page like this:

Clicking Hellbender Shell Access will take you to the following page:

You are now in the command line prompt on the login node of the Hellbender cluster. You can use this to submit jobs to Hellbender. Keep in mind that you should not run any code while you are on a login node – your account might get locked if you do.

 

Get an Interactive Session in Open OnDemand

To start an application in an interactive session go to the main OOD page:


There are several predefined applications to run, and we will select the most generic one from which you can run any software that is installed on the Hellbender. Select the Hellbender Desktop icon:

 

 

In the drop down menu, select the options that you want. Probably the most important one to select is the partition. Do not select the requeue or gpu_requeue partition for the interactive session because your session might get terminated unexpectedly due to preemption. We recommend selecting partitions gpu or general.

 

Notice, since the cluster is usually rather busy, you might need to wait to get resources – while the waiting time will strongly depend on resources you selected – if you require lots of CPUs, memory, and/or GPU cards your waiting time might be significant.
Select reasonable parameters. To get GPU options, you must first select a node type to be GPU and then the GPU card type.

 

Notice that general partition does not actually have GPUs even though GPU dropdown menu will offer to select a GPU node and a GPU card – but if you try to launch it, you will get an error message. Requeue, gpu_requeue and gpu partitions offer GPU cards.  In the near future requeue partition will not have GPU cards!
 

After you are satisfied with the selections click the launch button and it will take you to the following page:

Here you can see your expired (old) sessions with the button delete next to them, and current sessions waiting for resources with the cancel button. You can request multiple sessions at the same time.
 

Once your session gets computing resources, it will show a blue Launch button beneath it. 
(You can adjust the compression and image quality setting, but we recommend not changing it except if your Internet connection is slow - then give it a try.)

Click the launch button and you should get to the following page:

Which looks like a regular computer desktop.
The most important thing for HPC work is the terminal – black screen icon at the bottom of the screen – circled red.

 

Clicking the terminal icon will open a new terminal:

Notice, in the terminal we are not on a login node – we are on a compute node. Meaning you can now load any Hellbender module and run the code that you want.
 

Also, you can open multiple terminals by clicking terminal icon several times.
In one of the terminals, we launched Matlab for demonstration:

After some time, the Matlab GUI should show up – and it should be relatively responsive to work with – especially compared to the option of working with ssh connection with X tunneling.

 

If during your interactive session your computer (laptop) shuts down or you must close it and go somewhere else – as long as the OOD session does not expire, you should be able to reconnect to it once you have internet access again. Click on the MyInteractiveSessions on the main OOD page:


 

and you should see all your active sessions:

And you can click the blue launch button to connect to the OOD session – circled in red.

 

For Additional Assistance

For additional assistance please e-mail itrss@missouri.edu or open a general support ticket via TDX.