Dear all,

I've just committed (version 1.3.dev11) a few changes that affect all
batch-queuing deployments, with the objective of speeding up the
set-up of a "vanilla" cluster:

- The "HPC common" software set is now an optional add-on, see:
https://elasticluster.readthedocs.io/en/latest/playbooks.html#hpc-common

- The TCL-based "environment modules" tool is now the default, instead
of "Lmod"; Lmod can be requested by adding
`global_var_modules_tool=lmod` to the `[setup/*]` section.

- OpenMPI is the only part of the old "HPC common" role that is still
deployed on all batch-queuing clusters.

To get the pre-1.3.dev11 software set on batch-queuing clusters, you
need to add the `hpc` group to nodes.  For example, in a SLURM
cluster:

     [setup/slurm+hpc-common]
     # ... same as usual SLURM setup, but:
     master_groups=slurm_master,hpc
     worker_groups=slurm_worker,hpc

If you have any questions, remarks, or run into bugs with the new
version please let me know.

Ciao,
R

-- 
You received this message because you are subscribed to the Google Groups 
"elasticluster" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to