On Wed, 2007-03-07 at 22:56 +0000, John Hodrien wrote: > On Wed, 7 Mar 2007, Jeffrey Law wrote: > > > Ah. Does the headnode ever need to do anything to control the stateless > > nodes without human intervention (ie, does the head node ever fire off > > commands to the clients via SSH?) > > Yes. > > > As an admin, do you ever remotely log into the clients via SSH? If so, then > > you probably need a persistent SSH private key for the clients. > > Due to the way it works there's not really any need to have different keys on > the slaves, so in the past I've just used the same keys for all slaves Ah, yes, that will work too -- you make the cluster-wide key part of the system profile. Totally forgot about this approach.
> > The only other bit I have at the moment is X related. Due to hardware > failures I've got two different types of motherboards which require different > X configs. Currently this is simply detected at runtime, which works out as a > better solution than using something like puppet. Yea. Are you using the newer X auto-config stuff, or are you just runtime selecting from a set of config files you've specified? The X config stuff is a long standing wart -- while the autoconfig stuff has improved greatly, I'm still seeing better results with storing the good X configuration on the puppet master and downloading it to the clients. Of course, I'm reimaging clients more often than most :-) > But as I said, partly I wanted to see how far StatelessLinux had got since the > last time I looked at it. It's nice to know what it can do, so that when I > find myself in a situation that suits it later down the line, I notice. Understood. The focus right now is more around making it work with things like puppet and cobbler/koan. That takes us another step forward in terms of integrating into the overall systems management direction within Red Hat. Jeff _______________________________________________ Stateless-list mailing list [email protected] http://www.redhat.com/mailman/listinfo/stateless-list
