[Beowulf] Which do you prefer local disk installed OS or NFS rooted?

[Sean Dilda]

On Wed, 2004-07-14 at 16:38, Brent M. Clements wrote:
> Good Afternoon All:
> 
> Let me start by giving a little background.
> 
> Currently all of our clusters on campus have local disks which by using
> the systemimager suite of tools has an os image installed
> There are one or two clusters that are nfsrooted.
> 
> I'd like to know from all of you, which way is everyone leaning when it
> comes to clusters and os distribution.
> 
> Do you a. Install your OS on each local disk?(using whatever method,
> kickstart,systemimager, etc) or b. do you nfsroot and then use the local
> disk space as scratch and swap.
> 
> What benefits and drawbacks have all of you seen by doing either?
> 
> The Research/linux support group(which I am team lead of) here is leaning
> towards moving all of our clusters to model b. We've done both and from a
> sysadmin perspective we've found doing most things by nfs root a heck of
> alot easier than maintaining a systemimager configuration.

I tend to prefer the local disk route.  Part of that is because my
cluster is setup such that over time it will grow in leaps and bounds. 
Due to this I need to make sure everything I do is as scalable as
possible.

As for maintaining system images, you're right, it is a pain.  That's
why I don't do it :)  What OS are you using?  My cluster is based off of
Red Hat Linux.  This means that instead of maintaining a system image,
all I have to maintain is a package tree (mostly done by someone else)
and a kickstart config.  I also use yum to do updates across the system,
or I have a few shell scripts to quickly ssh to all my nodes and do the
yum update when I want to force an update out.  Kickstart and yum are
tools that were designed to maintain large scale deployments, and as
such work extremely well in maintaining beowulfs.