[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

routing problem



On Sat, Oct 05, 2002 at 04:30:58PM +0000, Doug MacFarlane wrote:
> On 04 Oct 2002, 19:35:14, Kourosh wrote:
> > On Fri, Oct 04, 2002 at 08:24:51PM -0600, dave mallery wrote:
> > Have you enabled IP forwarding on buster?  Do you have firewalling
enabled
> > on buster?
> 
> The answer is that you need to configure Buster to function as a
router.
>  Just configuring Buster to be a part of both networks does NOT mean
that
> Buster will automagically send packets originating on one network to
the
> other network.
> 
> There are several ways to accomplish this.  You can enable IP
Forwarding,
> as suggested, or you can run routed, an IP routing daemon.
> 
> Since you only need to forward/route the one segment, IP Forwarding
will
> be much more straight-forward, although every host on your 10.42.42
subnet
> will need a static route to buster for the 10.42.43 subnet, otherwise,
they
> will send everything to the firewall, and the firewall will put it back
on
> the 10.42.42 subnet addressed to buster (provided you have that rule
configured
> on the firewall).
> 
> madmac

thanks very much guys! 

the truth is that buster needs to route.

i have routing configured in the kernel, but have not really looked at
the problem directly.  due to other complications (not enuf disk space) i
will have to re-work buster, probably tomorrow, with a much larger disk.
(the cluster has 16 nodes and that's a lot of nfs roots!)

the story behind the cluster:  we have some great surplus auctions here
in NM.  last month, i became the owner of two "alta cluster" boxen from
los alamos. (2 0f 8.. the others scrapped) each has 8 dual pii 333s on a
nice asus mobo with antek power.  all assembled into these beautiful
nearly cubic yard boxen.  so rather than scrap them to ebay, i decided to
see if i could make em run again. (all the hi-power interconnect had been
removed)  now i don't  need to do nuclear calculations at home, but i
have done a lot of seti (>10000) and at this level, i need a boost in
production!  so the individual nodes need to get out to send results and
get work units.

thanks again, what a great list!

dave 


-- 
Dave Mallery, K5EN          (debian testing & woody)  
PO Box 520         .~.    _      Ramah,  NM  87321     
                   /V\   -o)
no gates...       /( )\  /\\     running Debian GNU/Linux
  no windows!     ^^^^^ _\_v        free at last!



Reply to: