Gilbreth Cluster Maintenance
Link to update at February 26, 2025 9:54pm EST UPDATE:
As of 9:50pm, we have concluded maintenance on Gilbreth and you can now login and submit jobs to the scheduler. As a reminder, this maintenance involved a major operating system upgrade, and the available software on Gilbreth has changed. Please be sure to double check the module versions your job scripts are loading and rebuild your python environments if necessary.
If you have any questions regarding this maintenance, please reach out to us at rcac-help@purdue.edu
Link to update at February 26, 2025 9:00pm EST UPDATE:
As of 9pm, we are wrapping up some final items before bringing Gilbreth back online. We will provide another update by 10pm.
Link to update at February 26, 2025 7:58pm EST UPDATE:
As of 8pm, we are wrapping up some final items before bringing Gilbreth back online. We will provide another update by 9pm.
Link to update at February 26, 2025 4:57pm EST UPDATE:
As of 5pm, we are still working to bring Gilbreth back online. We will provide another update by 8pm.
Link to update at February 12, 2025 5:49pm EST UPDATE:
This is a gentle reminder that in two weeks the Gilbreth cluster is scheduled for maintenance and will be unavailable on Tuesday, February 25th, 2025 at 8:00am EST through Wednesday, February 26th, 2025 at 5:00pm EST. This maintenance will allow us to upgrade the operating system on the cluster from CentOS 7 to Rocky Linux 9. This migration will also result in changes for many applications on Gilbreth.
We will be making a subset of test nodes running the new application stack available during the week leading up to the maintenance. These nodes can be used to re-build your applications and test the aforementioned changes to applications on Gilbreth.
If you have questions about this upgrade or need help from our support staff, please reach out to us at rcac-help@purdue.edu.
Link to original posting ORIGINAL:
The Gilbreth cluster is scheduled for maintenance and will be unavailable on Tuesday, February 25th, 2025 at 8:00am EST through Wednesday, February 26th, 2025 at 5:00pm EST. During this maintenance, Gilbreth will have its software stack modernized.
What is being upgraded?
During this maintenance, Gilbreth's operating system will be upgraded from CentOS 7 to Rocky Linux 9 and CUDA drivers will be updated to the latest available release from NVIDIA. This means that the system libraries on Gilbreth will be updated, in some cases by multiple versions, and the applications, libraries, and compilers provided by RCAC through the "module" command will be upgraded. Other user experiences such as SSH, web access, and job scheduling will remain unchanged.
How does this affect you?
- Gilbreth will be unavailable during the mainteance window.
- Slurm jobs that are still queued when this maintenance begins on Tuesday, February 25th, 2025 at 8:00am EST will be deleted.
- A reservation will be created that will prevent jobs from starting if their end time would take the job past the start of maintenance. Because pending jobs will be deleted, these jobs will never run.
- Any applications you have built against the system libraries that are being replaced may need to be recompiled. This may include applications you did not explicitly build yourself such as your R and Python environments.
- The "anaconda" modules will be replaced with a "conda" module. This "conda" module will expose the same "conda" interface you are familiar with, however it will install your packages from the "conda-forge" channel rather than the "default" channel in Anaconda.
- User files will persist through the upgrade.
How can you prepare for these changes?
In order to minimize disruption in researcher workflows, we will make a subset of test nodes running the new application stack available during the week leading up to the maintenance. These nodes can be used to re-build your applications and environments against the upgraded Gilbreth nodes prior to maintenance so that you can resume your work as quickly as possible.
Additionally, because the versions (as well as defaults) of modules on the cluster will change, be sure to update your job scripts to load the new versions of the modules post-maintenance.
If you have questions about this upgrade or need help from our support staff, please reach out to us at rcac-help@purdue.edu.