Warning

This is as a Lab Notebook which describes how to solve a specific problem at a specific time. Please keep this in mind as you read and use the content. Please pay close attention to the date, version information and other details.

Extra

The information here specifically pertains to the New OS transition and OnDemand outtage in summer 2024

Introduction

The HPCC is experiencing issues with our service related to upgrading the system OS and an unrelated, unplanned issue with the filesystem which is making the system unresponsive and impacting the OnDemand service. Below we will document workarounds to some common issues users have been encountering due to these events.

New OS

The transition to the new OS has dramatically change how the module system works on HPCC and some software may be available in different (often newer) versions or yet to be installed on the new OS. Our ongoing plan is support users by contiuning to install software on the new OS and address issues with existing one, but this process may be slower than usual due to the number of request and issues with the file system.

The easiest workaround for any issues related to the new OS is to make use of the backwards compatibility containers.

OnDemand

OnDemand is periodically experiencing a communication error between it and the rest of the gateway nodes. When this happens, home or the scratch system becomes temporarily unavailable, making the service inaccessible as it stores certain files in the users home directory. Because this issue is intermittment, users may notice that OnDemand is available at time as we work on it, but be aware that service may drop again without warning. Below we have some alternatives for commonly used servies on OnDemand.

File Upload/Download

File transfer can be done via the Globus service, which runs from a web browser independent of OnDemand. Documentation for this service can be found here

Jupyter Notebooks

Jupyter notebooks running from the command line on HPCC may be accessed on your local computer using port-forwarding. Detailed instruction for the process are available here.

Please me aware that this process depends on maintaing two open sessions with HPCC, so it may be unstable if you are frequently being disconnected.

RStudio Server

This is only possible on nodes running the new operating system. First, follow the Pre-Requisites on the Jupyter Port-Forwarding page. Then load the powertools, RStudio-Server and R module you wish to use.

module purge
module load powertools
module load RStudio-Server
module load R/4.3.2-gfbf-2023a  # For example

Then run

1	`rstudio_server`

and follow the onscreen instructions to connect.

Rstudio

Outdated

Since August 6, 2024 legacy nodes are no longer available, and RStudio (the non-Server version) cannot be launched using the new operating system. Please use the RStudio Server option above or contact ICER support if you are experiencing issues.

The legacy nodes (see the New OS section) can launch Rstudio session which you can access on your local computer. To run Rstudio interactively you will need to:

Have a X server running on your local machine (MobaXterm on windows, XQuartz for mac)
Access HPCC using ssh with the -X option (i.e "ssh -X username@hpcc.msu.edu)
Load R and rstudion (for example: "module load GCC/11.2.0 OpenMPI/4.1.1 R/4.2.2 rstudio/2022.07.2-576")
Run "rstudio" on the command line, which should open a Rstudio window on you local machine

As a warning, please be aware that multiple session of Rstduio can have the same script, creating a danager of unintentionally overwriting change in your files. Considering manually backing up file your are actively working, particularly if they were open in an OnDemand Rstduio session before starting a command line session.

VS Code

Visual Studio Code may fail to correctly set up its terminal, resulting in module and quota commands failing, among others. As we investigate this issue, please run source /etc/profile in the VS Code terminal after you connect to the HPCC.