[Beowulf] distributions

Robert G. Brown rgb at phy.duke.edu
Fri Jan 27 10:13:30 PST 2006


On Thu, 26 Jan 2006, Warren Turkal wrote:

> Is it ok to mix linux distributions when building a cluster? I am wondering
> for migration purposes. For instance, it the current cluster had FC2 and I
> wanted to move to FC3, would it be okay to install new nodes as FC3 and
> gradually migrate?

It's "OK", but sure, you'll have some problems.  In particular, PVM
would be happy enough with it if you build separate binaries (or verify
binary compatibility).  MPI I think you'd have to end up with binary
compatibility -- it isn't as happy with heterogeneity (right, o MPI
experts?).

Other than that, PVM would let you run linux and Windows and AIX on
different nodes (so FC-whatever would surely work).  As would raw
sockets.

The other problems are all management ones -- it is a PITA to maintain
different linux variants (speaking from much experience and pain).
Different repos, different update streams, missing or incompatible
libraries, yuk.

What >>I'd<< recommend is the following, especially if you use FCx.

Construct kickstart files that build a "standard cluster node".  You can
use tools (Duke provides some nice ones at the Linux at Duke site) that
can even customize kickstart installs per node/ip in real time through
the magic of cgi.  Build an FC4 node, not FC3 -- why update to something
that is already obsolete?  Then pick a day and flash over the entire
cluster to FC4.  With kickstart/pxe and said scripts, this should take
maybe a few hours total, and will leave you all happy and everything
with ONE repo to support, ONE yum-update to manage, ONE set of libraries
and no binary compatibility issues.

That is, you maybe "can" do this, but do you want to?

    rgb

>
> wt
>

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu





More information about the Beowulf mailing list