ARC system notices

[Advanced Research Computing]

Summary Outline:

1. V100 nodes on Cascades are being migrated to the Infer cluster
   
2. Huckleberry cluster decommission SP22
   
3. BeeGFS mount removal April 15th
   
4. /fastscratch to have regular, automated file cleanup

1. V100 nodes on Cascades being migrated to Infer cluster

This week, half of the V100 GPU nodes from Cascades have been migrated to the Infer cluster. These nodes are newer than the rest of Cascades and will remain in service longer, so they are being reprovisioned in the same paradigm as ARC's newer systems, Infer and Tinkercliffs.

What you need to know about migrating workloads from Cascades to Infer:
 - Begin using /projects storage instead of /groups 
 - /work is not available on Infer (see this webpage for details)
 - Find suitable software modules and request software installations if needed (see this webpage)
 - Build new Anaconda environments for the reprovisioned V100 nodes (see this webpage)

The remaining V100 nodes (20 of them) on Cascades will be migrated to Infer at a later date, tentatively April 15th.

2. Huckleberry Decommission
This is a reminder that the Huckleberry cluster will go offline at the end of the Spring 2022 semester. Infer and Tinkercliffs clusters both have GPU nodes which can be considered as destinations for any workloads that need to be migrated.

3. BeeGFS mount removal
During the maintenance outage in early March, some important storage system changes were implemented and this is a reminder that you may need to take action as a result. Notably, the out-going BeeGFS storage system has been available read-only at /beegfs to allow retrieval of any needed data from what was previously available as /work on Tinkercliffs and Infer. Two weeks from now, on April 15th, the storage target will be removed. Please use the intervening time to verify that you have recovered any data you need from that location.

4. /fastscratch automated cleanup
The /fastscratch storage target on Tinkercliffs will soon begin operating as a true "scratch" file system. This means that files in that location will be subject to regular, automated deletion based on their age. If you have any static data in that location that you wish to keep, please move it to an appropriate, more permanent location soon. More detail on this topic will be provided in subsequent communications from ARC. 

Use the links below to get help or more information about ARC systems:

Documentation site: https://docs.arc.vt.edu
Helpdesk Tickets: https://www.arc.vt.edu/help
Zoom Office Hours Calendar: https://arc.vt.edu/office-hours/

Regards,

Advanced Research Computing

Virginia Tech