Scientific Computing and Data / High Performance Computing / Frequently Asked Questions
Frequently Asked Questions
Q: How can I request a Minerva account?
Q: I am a Mount Sinai employee and I lost my password. How do I reset it?
Q: I am an external collaborator on Minerva and I lost my password. How do I reset it?
Q: I lost my token, replaced or lost my cell phone! What should I do?
Q: My account is locked. What should I do?
Q: How do I deactivate or delete my Minerva account?
Q: Can I set up my own module system?
Q: The package I need isn’t available! Will you install and maintain it?
Q: When I load a module, it gives me an error. What should I do?
Q: I get errors using java when I submit a job, but not when testing locally.
Q: How can I install software packages such as R or Python?
Q: How do I submit, cancel, modify, and monitor jobs on Minerva?
Q: How can I get internet access on a Minerva compute node?
Q: How do I change, grant, or remove read/write permissions for files and directories?
Q: How do I request access to a specific project, add members, or remove a member?
Q: How do I check the status of a ticket?
Q: How do I access and run RStudio through OnDemand?
Q: How do I start RStudio sessions on Minerva with different R versions and install new packages?
Q: How do I publish applications using RStudio Connect on Minerva?
Q: How do I create an Conda environment on Minerva, including package installation?
Q: Why is my submitted GPU job stuck in a pending state?
Q: When I am on a GPU node, how can I check the GPU status, usage, memory, or driver information?
Q: How can I use a newer version of the Ollama image than the default with the wrapper script?
Q: How do I transfer data between my local computer and Minerva?
Q: How do I recover deleted or lost files from Minerva?
Q: How do I access someone else’s archived data?
Q: I don’t like web- or e-mail-only communication. What can I do?
Accounts and Access
Q: How can I request a Minerva account?
A: To request a Minerva account, please complete the online account request form available here.
Q: I am a Mount Sinai employee and I lost my password. I need to reset it.
A: The password you use to log into Minerva is your Mount Sinai password. We do not reset passwords. If you have forgotten your Sinai password, reset it here.
Note that employees from SLR, BI, and NYEE cannot use this link — they must call their respective IT Help Desks.
Q: I am an external collaborator on Minerva and I lost my password. How do I reset it.
A: If you are an external collaborator, you will need to email us at hpchelp@hpc.mssm.edu to reset your password.
Q: I lost my token, replaced or lost my cell phone! What do I do?
A: Contact the ASCIT help desk, HPC Help does not have the privileges to make necessary changes.
Q: My account is locked. What do I do?
A: The most common reason for failure to log in is a locked two-factor authentication token. This usually happens after successive failed attempts to use the token. Please remember that you can only use a two-factor token output one time. If your password does not work, you will need to wait 30 seconds for the next two-factor token output. We can check if your token is locked, and unlock them. Just send us a message to hpchelp@hpc.mssm.edu.
Q: How do I deactivate or delete my Minerva account?
A: We do not remove Minerva user accounts. When users leave Sinai, their AD network account is terminated by DTP, which also revokes access to Minerva. A monthly cron job checks the status of AD accounts. If an account is inactive, we remove any hosted webpages and Posit Connect applications associated with it.
Software Installation
Q: Can I set up my own module system?
A: Yes! Many of our more advanced users maintain their own packages while in active development, or if they are frequently making changes. Please reference our MODULES documentation, which can be found on this website, for information about setting up your own modules.
Q: The package I need isn’t available! Will you install and maintain it?
A: Probably. We do our best to meet the module needs of all of our users, and the vast majority of packages will be supported if they are useful to more than just an individual user, and have achieved a certain level of maturity. For example, a perl script you wrote yesterday for your own downstream analysis will not be supported, but any community software will be. This is true of top-level packages, as well as R or python packages, for example.
Q: When I load a module, it gives me an error. What do I do?
A: Most of the time this happens it is because of conflicts with already loaded modules. For example, most of the packages we build depend on gcc 4.8.2, our default compiler. If you’ve loaded another compiler toolchain for your own work and then load a package that depends on a conflicting compiler toolchain, the module system will complain to you. Starting fresh by issuing a module purge command will typically fix any problems you have. If you continue to see errors, or need two conflicting packages to be loaded simultaneously, just send in a ticket and we’ll get it sorted out.
Q: I get errors using java when I submit a job, but not when testing locally.
A: Nodes have their own java versions which are local to the operating system installed on the node. Users should always load java using the module file system, which will ensure uniform behavior across the entire cluster. The java packages on Minerva are in the “common” module repository.
Q: How can I install software packages such as R or Python?
A: We provide a wide range of commonly used R and Python packages that are already installed on the system. If you require a package that is not available, you may install it yourself by following the instructions on our documentation site:
If you encounter difficulties installing a package or need additional support, please open a support ticket, and our team will assist you.
Job Submission
Q: How do I submit, cancel, modify, and monitor jobs on Minerva?
A:
- Use bsub to submit a job.
- Use bkill to cancel a batch job.
- Use bmod to modify the resource requirements of a pending job.
- Use bjobs to check the status of your job.
- Use bhist to view a job’s history and retrieve records from the LSF database.
Please check the documentation link to learn more about how to use the above commands.
Q: How can I get internet access on a Minerva compute node?
A: Please load the following module as shown below:
ml proxies/1
Projects Related
Q: How do I change, grant, or remove read/write permissions for files and directories?
A: File and directory permissions are typically managed through group membership and Access Control List (ACL) settings by the project or directory owner. If you need assistance, please contact the project or directory owner directly. You may also open a support ticket by emailing hpchelp@hpc.mssm.edu
Q: The allocations process is cumbersome – it takes too long to get my work up and running. Can you help?
A: We’re working on a brand new allocations process for the next cycle which should significantly streamline and simplify the procedure for obtaining an allocation.
Q: How do I request access to a specific project, add members, or remove a member?
A: To request access to an HPC project (including adding or removing members), please open a ticket by emailing hpchelp@hpc.mssm.edu. Approval from the project’s Principal Investigator (PI) or a delegate is required.
Ticket and Documentation
Q: How do I check the status of a ticket?
A: To check the status of your ticket please visit link here.
- Click on “Check Ticket Status.”
- Enter your Email Address and Ticket Number on the next page.
- You will then receive an access link via email, which you can use to view the status of your ticket.
Please note: There is a known issue with this service. If you are unable to access your ticket using the link above, kindly contact us directly at hpchelp@hpc.mssm.edu
Q: How can I find information about Minerva’s technical specifications, policies, and user documentation?
A: Please check the following links:
RStudio, Jupyter, Conda
Q: How do I access and run RStudio through OnDemand?
A: To access and run RStudio through OnDemand, first log in to the Minerva OnDemand site with your credentials. Then, go to Interactive Apps and select RStudio Server. Fill in the form with details such as the project account, number of cores, memory, runtime, and R version, and click Launch. Once the session starts, a button will appear to Connect to RStudio Server, which will open the RStudio interface directly in your browser.
Q: How do I start RStudio sessions on Minerva with different R versions and install new packages?
A: Please use the following command to start RStudio with a specific R version.
minerva-rstudio-web-module.sh -mm please_mention_module_name_here
Please check our RStudiodocumentation site with detail instruction about how to start RStudio session and install new packages locally.
Q: How do I publish applications using RStudio Connect on Minerva?
A: Please check the following link: R Documentation (section: RStudio Connect Server)
Q: Can multiple Jupyter notebooks be run concurrently using the minerva-jupyter-module-web.sh script?
A: Yes, you can run multiple Jupyter Notebooks concurrently using the minerva-jupyter-module-web.sh script, since each execution submits a separate LSF job that generates its own unique URL and port for access. This means users can start multiple sessions in parallel. The script ensures that each job is isolated, so links won’t conflict, and users can also adjust memory or GPU allocations per session if needed.
Q: How do I create an Conda environment on Minerva, including package installation?
A: On Minerva, you can create, activate, and manage Conda environments for different projects, adding the software you need and switching between environments as required. Within these environments, you can use Conda to install the necessary libraries. A step-by-step guide is available here
AI and GPU
Q: Why is my submitted GPU job stuck in a pending state?
A: A GPU job may remain pending for several reasons, most commonly because the requested resources are not currently available. To increase the chances of faster scheduling, request only the resources you truly need. Start by requesting a single GPU and testing your job with a small dataset to measure actual resource usage before scaling up. This approach helps the scheduler place your job more quickly and ensures efficient use of cluster resources.
Q: When I am on a GPU node, how can I check the GPU status, usage, memory, or driver information?
A: Please use the nvidia-smi command.
Q: How can I use a newer version of the Ollama image than the default with the wrapper script?
A: Please use the following steps.
Start an interactive job:
bsub -P acc_hpcstaff -q interactive -n 4 -W 00:40 -R rusage[mem=8000] -R span[hosts=1] -Is /bin/bash
In the interactive session, run the following commands to build the latest Ollama Singularity image:
module load singularity/3.11.0
singularity build ollama_latest.sif docker://ollama/ollama:latest
Exit the interactive session.
From the login node, start the Ollama script using your Singularity image path as follows:
sh minerva-ollama-web.sh --image full_path_of_the_Singularity_image_file
Data Transfer and Recovery
Q: How do I transfer data between my local computer and Minerva?
A: Please use Globus. For detailed instructions, please check here
Q: How do I recover deleted or lost files from Minerva?
A: We do not back up users’ data. It is the users’ responsibility to back up their data. Please use our TSM Service
Q: How do I access someone else’s archived data?
A: Please find the directions for recovering this data here
Communication
Q: I don’t like web- or e-mail-only communication. What can I do?
A: The HPC staff is always available to talk 1-on-1. You can come talk to us in person on the 3rd floor of Icahn – Greenland Rm. (1425 Madison Ave.), or send a message to arrange a meeting. The best way to contact the team, including reporting issues or asking questions, is to email hpchelp@hpc.mssm.edu