Open OnDemand
Overview¶
Open OnDemand (OOD) provides browser-based, interactive access to Savio at https://
Some apps and services that are provided are:
OOD Desktop
Jupyter notebooks
RStudio
MATLAB via a Desktop environment
VS Code
File browsing
Slurm job listing
Terminal/shell access
Logging In¶
Visit https://
ood .brc .berkeley .edu. Authenticate via CILogon (choose University of California, Berkeley unless otherwise directed) and login with your UC Berkeley credentials.
Service Unit Charges¶
Most OOD apps launch Slurm interactive sessions behind the scenes. They incur SUs just like regular Slurm jobs.
Two shared nodes—n0000.savio4 and n0001.savio4 (partition: ood-inter) are provided for low-intensity work and do not consume SUs. Treat these like login nodes (no heavy computation). Time is charged from the moment a node is allocated until you stop the session, even if the browser tab is closed.
To check usage, go to My Interactive Sessions and view all currently running jobs under Jobs → Active Jobs.
Using Open OnDemand¶
Files App
Navigate to Files → Home Directory to:
View, create, rename, and delete files/folders
Upload/download files (For large transfers, use Globus instead)
View Active Jobs
Jobs → Active Jobs lets you view/cancel Slurm jobs, including those created by sbatch/srun and sessions launched via OOD.
Shell Access
Open a cluster shell via Clusters → BRC Shell Access.
Interactive Apps
Launch from Interactive Apps (top menu). Available options include:
Desktop App (full GUI desktop on Savio)
Jupyter Server
RStudio Server
Code Server (VS Code)
Desktop App
Use the Desktop App for GUI software (e.g., MATLAB GUI), light visualization, and file management.
To start a desktop session:
Log in at https://
ood .brc .berkeley .edu. Go to
Interactive Apps → Desktopand choose shared node (exploration/debugging; no SUs) or compute via Slurm (SUs apply).Fill the form → Launch.
Adjust image quality/compression if needed → Launch Desktop, which will open a new tab with your desktop session.
To open a command line terminal, use the file manager in the desktop or right-click the desktop → Open Terminal Here
Jupyter Server
Choose Jupyter Server under Interactive Apps. You will have the option to run on shared nodes (no SUs; low intensity) or via Slurm partitions (SUs apply). When finished, go to My Interactive Sessions and Delete the session to stop charges
Don’t submit Slurm jobs from the terminal inside Jupyter. The Jupyter server itself is running inside a Slurm job; launching jobs from within it (job-in-job) can cause failures. For job submission from a browser, use Clusters → BRC Shell Access instead.
RStudio Server
Launch via Interactive Apps → RStudio Server. There are three available modes:
Compute via Slurm (SUs apply; choose resources)
Compute on shared OOD node (no SUs; ≤ 8 GB RAM; a few cores only)
Stop charges by Deleting the session under
My Interactive Sessions
For packages needing compilation or external libs, install from a command-line R session (outside OOD RStudio). Once installed, they’ll be available in RStudio.
Some variables (e.g., SLURM_CPUS_ON_NODE) are not visible inside OOD RStudio via Sys.getenv() or system().
Code Server (VS Code)
Launch Code Server (VS Code in the browser). There are two available modes:
Compute via Slurm (SUs apply)
Exploration/debugging on shared nodes (no SUs; ≤ 8 GB RAM; a few cores)
VS Code Remote SSH is not allowed for Savio for security reasons. Use Code Server via OOD instead.
For more details on using Open OnDemand, look at the full documentation provided by Research IT.