Porting existing virtual environments to Open OnDemand

IMPORTANT: Because of the change of operating system and software stack, your exisiting virtual environment kernels for Python may not work right away in the OnDemand site. You should be able to activate your python environments in a Trillium terminal (see Terminal access), with all required Trillium modules loaded, and then issue the command venv2jup to get them working correctly.

Introduction

This guide will walk you through the basic steps to get started with the SciNet Open OnDemand portal. Open OnDemand (OOD) is a web-based platform that provides access to a wide range of scientific applications and computing resources, such as Jupyter Lab, R Studio, and VS Code. It allows you to interact with Trillium through a web browser instead of via a terminal, without the need to install any software on your local machine. You will be able to perform file management, submit/monitor jobs and run applications interactively. More information on this project can be found at https://openondemand.org.

Logging into the Open OnDemand portal

Figure 1: Dashboard.

To access the Open OnDemand portal, open a web browser and navigate to the following page: https://ondemand.scinet.utoronto.ca. You will be prompted to enter your Alliance username and password, followed by a second factor authentication via Duo or Yubikey. Once you have logged in, you will be taken to the Open OnDemand dashboard. From here you can access the various tools and applications available on the platform.

File management

Figure 2: File browser.

The Open OnDemand platform provides a file browser that allows you to manage your files and directories on the filesystem. To access the file browser, click on the Files tab and select which directory you want to manage from the drop-down (HOME, SCRATCH or PROJECT). You will be taken to the file browser interface, where you can:

Navigate through your directories
Upload/download files
Create new files/directories
Delete files/directories
Edit existing files

Uploading Files

The current file size upload limit is 10GB, if you need to upload a file larger than this or are facing upload issues due to a bad internet connection for example please try using Globus. There is a Globus button in the file browser at the top right, , which will take you to the Globus web interface where you can login with your Alliance username and password. The path navigated to in the Open OnDemand file browser will be the same path opened in Globus.

Job submission

Open OnDemand also provides a job submission interface that allows you to submit batch jobs to Trillium.

The Open Composer app provides a suite of Slurm job template scripts that can be submitted directly to the Trillium scheduler. It also provides an interface to monitor your submitted jobs, via the History tab. You can access Open Composer by navigating to the Jobs drop-down menu and selecting Open Composer or by clicking on one of the Slurm job templates e.g. MPI Slurm Job, OpenMP Slurm Job and Hybrid MPI/OpenMP Slurm Job.

Figure 3: Open Composer shortcuts.

Once you have selected a job template, you will be taken to the job submission page. This is split between the job parameters on the left and the job script itself on the right. The job parameters let you control how many resources your job will use, such as the number of nodes, number of tasks per node, wall clock time and output file name. The job script section displays the script that will be submitted to the scheduler. Any changes made to the job parameters will be reflected in the job script automatically. You may also edit the job script directly if you wish.

The extra fields at the top of the page allow you to change how your job is submitted:

Script Location: specifies the directory where the job script will be saved and where your job will be run from.
Script Name: specifies the name of the job script file.
Job Name: specifies the name of the job that will appear in the job queue.

Once you are happy with your job script, click on the Submit button to submit the job to the scheduler and save your script to the Script Location. If your job was submitted successfully, you will see a confirmation message at the top of the page with your job ID: .

Figure 4: Open Composer job script editor.

Note: The template scripts provided in Open Composer are basic examples to get you started. You will need to modify the job script further to suit your specific needs, such as loading your required modules and specifying input/output files.

Monitoring Jobs in Open Composer

To monitor your submitted jobs in Open Composer, navigate to the History tab. This will display a list of all your submitted jobs, along with their status: Queued, Running, Completed, Failed. You can filter the jobs by using the Filter text box at the top right or by using the checkboxes below. Clicking on different column fields will give different information about the job:

Job ID: displays detailed Slurm information about the job
Application: opens the job script editor of the template you used
Script Location: opens an OOD file browser window at the location of the job script. Clicking on the the small terminal icon will open a terminal in the job script location.
Script Name: displays the job script that was submitted to the scheduler

To resubmit or modify a previously run job click on the job script under the Script Name column and click Load Parameters. This will take you back to the job submission page where further modifications can be made to the job.

Figure 5: Open Composer job monitoring.

Supported Applications

Open Composer currently supports the following applications:

MPI Slurm Job
OpenMP Slurm Job
Hybrid MPI/OpenMP Slurm Job
Python Slurm Job

Job monitoring

To get an overview of all your jobs in the queue you can use the job monitoring interface. Navigate to the Jobs tab and select Active Jobs. You can filter the jobs by using the Filter text box at the top right. Columns can also be sorted by clicking on the column headers, for example you can sort by job status (running, completed, failed, etc.). Clicking on > to the left of a job will show you more details about the job, such as the start/end time, node list and account charged etc. You might also want to show all jobs in the queue, you can do this by clicking on the drop-down menu at the top right and selecting All Jobs. A more detailed view of your jobs can still be found using the myscinet portal.

Figure 6: Active Jobs.

Interactive applications

Open OnDemand also features interactive applications that can be run directly from your web browser. To access the applications, navigate to the Interactive Apps tab and select the application you want to run from the drop-down. This will then bring you to the job submission page where you can choose job parameters such as:

Length of job in hours
Number of cores
Amount of memory to allocate (GB)
Notify me by email when the job starts

When you have chosen your job parameters click on the Launch button to submit your job to the queue. You will be taken to the My Interactive Sessions page where you can see the status of your job, i.e. queued, running or completed. Once the job has been assigned a node and is running, you can click on the Connect to ... button to launch the application. The application will open in a new tab in your browser, and you can interact with it as if it was running locally.

If you would like terminal access to the node where the application is running, to monitor the performance for example you can click on the button beside Host starting with >_. This will open a terminal window in your browser where you can run commands on the node directly.

If for whatever reason you would like to kill the job, you can do so by clicking on the red Delete button in the job panel in the My Interactive Sessions page.

Figure 7: Interactive app submission form.	Figure 8: Interactive sessions page. Figure 9: Interactive app session.

Installed Applications

We currently support the following applications:

If you would like an application installed please email us at: support@scinet.utoronto.ca.

Terminal access

Figure 10: Terminal

Sometimes you might prefer to use a terminal to interact with Trillium, Open OnDemand provides a web-based terminal that you can use to access the command line interface. To access the terminal, navigate to the Clusters tab and select Trillium Shell Access. This will open a new tab in your browser with a terminal window where you can run commands as you would in a regular terminal session.

Differences compared to the Jupyter Hub

feature	Jupyter Hub (decommissioned)	Open OnDemand
authentication	password	password + MFA
first installed	2017	2025
last update	2021	2025
supports	Jupyter Notebook, JupyterLab (R, Python, Julia)	Jupyter Notebook, JupyterLab (R, Python), Rstudio, VSCode, Desktop, SAS⁴, Stata⁴, ParaView, Forge DDT/MAP
start and continue later	Yes	Yes¹
command terminal	No	Yes
file management	Yes (limited)	Yes
monitor jobs	No	Yes
submit jobs	No	Yes
core limit	8 cores²	20 cores (8 for high memory)³
memory limit	48 GB²	85 GB (500 GB for high memory)³
time limits	3 days²	3 days³
operating system	CentOS 7	RockyLinux 9
software stack	NiaEnv, CCEnv	CCEnv
system issue alerts	No	Yes
user quota alerts	No	Yes
error logs	No	Yes
hardware	1 x CPU with 40 Intel "CascadeLake" cores at 2.5 GHz, 1TB RAM	62 x CPU with 40 Intel "CascadeLake" cores at 2.5 GHz, 180GB RAM (default) 3 x CPU with 40 Intel "CascadeLake" cores at 2.5 GHz, 1TB RAM (high memory)

¹ Within the requested limits.

² Limits on JupyterHub were not implementated very strictly, so you could temporarily exceed these.

³ Limits need to be requested before starting an application.

⁴ Only for users with a license for these products.

Debugging Errors

Figure 11: Accessing error logs via the session card

If you encounter any errors while using an interactive Open OnDemand job, you can check the logs for more information. To access the logs, navigate to the My Interactive Sessions tab and find your active session. Click on the output.log link (see Figure. 8) to open a separate tab which displays the output of your job. This file contains the standard output and error messages generated by the job, which can help you identify any issues that may have occurred during the session. When submitting a ticket to SciNet support, please include the output.log file, your Session ID, which is displayed as a long string of characters, e.g. 8feb45fa-bc65-4846-8398-2a73c1bf8e5a, and any other relevant information to help us assist you more effectively.

Open OnDemand Quickstart

Contents