05/03/2024
Note the following changes to the atlas compute cluster:
OS: Rocky 9.1 linux distribution is now utilized
SW: a new software stack has been installed with many version changes.
Use the module commands to see available software.
examples: “module av” and “module spider pkg”
It should be expected that most user codes will need to be recompiled.
Submit a help ticket for any software issues or to request any additional software.
Partition changes:
- development-gpu partition has been removed and the node added to the gpu-v100 partition. A combination of the other gpu partitions and the qos’s should now be utilized
- gpu partition renamed to gpu-v100 to better reflect the gpu type utilized
- gpu-a100-mig7 partition created to better differentiate between the A100 GPUs that are configured for multi instance graphics. The number of nodes in this partition has been reduced from 4 to 2 with the nodes added to the gpu-a100 partition
- gpu-a100 now has 3 nodes. Each with 8x full A100 GPUs available.
ssh host keys:
All service nodes (atlas-login-[1-2],atlas-dtn-[1-2],atlas-devel-[1-2])
have updated ssh host keys and will require the acceptance of these new keys.
This can be done issuing the following command on linux systems.
Old keys can be removed using the following commands:
ssh-keygen -R atlas-devel.hpc.msstate.edu
ssh-keygen -R atlas-devel-1.hpc.msstate.edu
ssh-keygen -R atlas-devel-2.hpc.msstate.edu
ssh-keygen -R atlas-dtn.hpc.msstate.edu
ssh-keygen -R atlas-dtn-1.hpc.msstate.edu
ssh-keygen -R atlas-dtn-2.hpc.msstate.edu
ssh-keygen -R atlas-login.hpc.msstate.edu
ssh-keygen -R atlas-login-1.hpc.msstate.edu
ssh-keygen -R atlas-login-2.hpc.msstate.edu