Two heads are better than one! :)

[Donald Becker]

On Thu, 31 Oct 2002, Robert G. Brown wrote:

> On Thu, 31 Oct 2002 robnash at rogers.com wrote:
> 
> > Hello Everyone,
> 
> > Does anyone know if it's possible to have two active head nodes in a
> > Beowulf style cluster? I would like to have two physical access points
> > into one cluster.  Any help regarding this would be greatly appreciated.
> 
> Beowulf "style" cluster, sure.  A Scyld beowulf I don't know (but don't
> think so).

Yes, it's possible, but it requires special configuration with the
current release.  (Note the word "current".)

Some key elements to consider:
   Unified user set or independent administrative domains
      (the master-per-department model) 
   What to do with the jobs controlled by a master when it fails.
   Cases
        fail: kill the master's jobs
	reparent: re-parenting is done to a "adopting" master
	detach: Continue running the application.

> The original >>beowulf<< concept was more like "the COTS cluster is the
> supercomputer", and that is what scyld and the nasa goddard folks have
> largely focused on, so most "true beowulfs" have a single head node,

The first few NASA Beowulf systems all had two "worldly" nodes that were
connected to external networks.  We didn't bother doing that with later
clusters because the systems never crashed.  (Kernels such as 2.0.30
were astonishingly reliable.)

-- 
Donald Becker				becker at scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Scyld Beowulf cluster system
Annapolis MD 21403			410-990-9993