I'm pretty new to the IT world as far as the heavy duty stuff goes...so scrolling through the vast pages of EE (just registered yesterday) has been pretty overwhelming for me. Continue to keep this in mind while you answer please :)
I've recently started a research position as a SysAdmin for a cluster of machines running Linux here at my university. Our goal is to try to automate the kickstart process as much as possible as well as have some way to monitor all the machines effectively. After a bit of research I figure Rocks Linux (www.rocksclusters.org
) might be a good way to go, seeing as how it's based from Redhat which wouldn't change much of our implementation. However to my dismay, but team leader explains that we won't be able to use it since our machines have public IPs as opposed to a private network connected through a frontend to the internet.
I'm a "little" familar with what he said, but is this true? Do you absolutely need to have private IPs to implement such a thing, in which case we'd have to revamp our entire network infrastructure (highly unlikely). If so, is there some other solution we could look into? I know Ganglia is something Univ. of California - Berkely is working with, and that's now implemented into the Rocks kernel, but how about using it standalone?
Any suggestions about the automation of the kickstart process, or even large-scale node monitoring would be highly helpful.