[Beowulf] Re: quick and dirty method for starting job on another node?

Ben Mayer bmayer at gmail.com
Tue May 10 18:33:15 PDT 2005

This brings up an interesting point. How many people are running with
their entire cluster exposed to the big bad internet. From thinking
about it for a minute I would simply expose the head node to the
network/internet and everything else would be a royal pain to route a
packet to.

Why does this not work in practice?


On 5/10/05, David Mathog <mathog at mendel.bio.caltech.edu> wrote:
> > > > Any good systems person would shoot you, of course, for putting
> any such
> > > > thing on an open network.
> > >
> > > rsh  :)
> I missed the original post somehow.  If that poster wants something
> that's really, really fast but hideously insecure for his tiny
> feeble nodes consider using nettee like this:
>   nettee -stm EOS -next $NEXTNODE | execinput
> See the distribution here for more info:
>  http://saf.bio.caltech.edu/nettee.html
> I've messed with this a little where the top node reads
> from a fifo in -stm mode and a command sent to that fifo
> with "echo" executes on every node in the chain very, very quickly,
> essentially they run in parallel.  No rshd required.  NO
> SECURITY TO SPEAK OF EITHER!  Good enough for a private subnet
> though. If all he wants to do is contact
> one node then change the nextnode part to _EOC_.
> To shut down the session echo "EOS" to the fifo.  (Your
> choice of string, it doesn't have to be EOS.)
> Regards,
> David Mathog
> mathog at caltech.edu
> Manager, Sequence Analysis Facility, Biology Division, Caltech
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

More information about the Beowulf mailing list