Environment loading in parallel/SSH connection

For clusters with different environments, it would be convenient to allow a profile to be explicitly loaded when new ssh connections are started as sub-processes. My understanding is that environment variables in the current process do not propagate to new ssh connections, which will probably source .bash_profile or equivalent.

Being able to pass along a few lines to run in the ssh terminal like: "module load XXX module load YYY source XXX.sh" would allow for the correct environment to be set for each slave process.

Comments (2)