Results: the cuda role doesn't template in /etc/profile.d/cuda.sh
Any thoughts on this? We have a CUDA module where there might be a conflict.
B) Sites would need to update their group_vars.
Just updating pull_variables.yml is not enough as then the variable is different for compute.yml vs local.yml
Pointing to the pull_variables.yml is also problematic as we'd need to fetch that templated file from the compute nodes to where ansible push is run from.
Perhaps one could start using a vars_files pointing to a file we are using to override role's variables in.
Two things:
A) Disable the cuda.sh bash profile script?
Before we update the cuda role we could set
Results: the cuda role doesn't template in /etc/profile.d/cuda.sh
Any thoughts on this? We have a CUDA module where there might be a conflict.
B) Sites would need to update their group_vars.
Just updating pull_variables.yml is not enough as then the variable is different for compute.yml vs local.yml Pointing to the pull_variables.yml is also problematic as we'd need to fetch that templated file from the compute nodes to where ansible push is run from.
Perhaps one could start using a vars_files pointing to a file we are using to override role's variables in.