[Beowulf] anyone using SALT on your clusters?
hahn at mcmaster.ca
Thu Jun 27 15:39:07 PDT 2013
consider this a bit of devils advocacy and/or display of hickishness:
why the big deal about cf-engine/salt/puppet/chef/etc?
to me, a sensible cluster is nfs-root based, so there is
hardly ever any mass-configuration to do. when there is,
it's usually something like "pdsh -a check_something.sh".
I can easily imagine large web-based sites that might have
lots of machines in different roles, for which these auto-devops
systems would make a lot of sense. but even a large HPC cluster
might consist of one admin node, and identical N-1 compute nodes,
and one login node (compute node with a couple tweaks.)
an organization might well have multiple such clusters,
though that begs the question of why they are separate...
regards, mark hahn.
More information about the Beowulf