[lug] routing

Hugh Brown hugh at vecna.com
Thu Apr 25 08:16:21 MDT 2002


Thank you.  I'll look into bonding and into putting them all on separate
subnets.

Hugh


On Wed, 2002-04-24 at 18:42, Charles Morrison wrote:
> Hugh Brown wrote:
> 
> > I have 4 identical machines that will be part of a cluster eventually
> > 
> > Each has an onboard 10/100 nic and 3 10/100/1000 cards.
> > 
> > Each of the three gig cards is connected via crossover cable to one of
> > the other three machines.  each machine has all three gig cards sharing
> > the same ip address.  I was able to ssh to node2 from node1 yesterday. 
> > I was trying to set up rsh/rlogin with rhosts for an oracle clustering
> > product that depends on rsh (I know it is insecure...).  The rsh stuff
> > from node1 to node2 wouldn't work at all.  I figured that the routing
> > was probably all messed up and decided to tackle it today.  I rebooted
> > node1 (they were all headless) after putting a keyboard, mouse and
> > monitor on it.  Now, I couldn't even ping node2 let alone ssh to it. 
> > All the gig cards showed that they were connecting at 1000 and full
> > duplex.  I brought a laptop over and connected it to node1 via crossover
> > cable to the onboard nic and that worked fine.  I connected the same
> > crossover cable to one of the gig nics and the nic recognized that it
> > was connected at 100 full duplex.  I pinged from the laptop and node1
> > saw the arp requests and even responded a few times but none of the
> > responses showed up on the laptop (via tcpdump).
> > 
> 
> > I'm at a loss.
> 
> 
> 
> I can see why. I don't think you can get there the way you have it set 
> up. Each gig card must either have it's own IP address OR they would 
> have to be bonded and go to the same destination. I think your being 
> able to do anything yesterday was a fluke.
> 
> If what you're trying to do is have one card dedicated to communicate 
> soley to one other machine, use different class C networks for each 
> destination and it's associated card and have them be members of one 
> network. This way you don't have to muck with routing tables.
> 
> 
> > 
> > What's the next step?
> > 
> > Is there a way for a machine to ping itself from eth0 to eth1 (e.g. the
> > onboard nic to a gig nic)?
> > 
> > The nodes right now are vanilla redhat with the updates applied.
> > 
> > Hugh
> > 
> > 
> > 
> > _______________________________________________
> > Web Page:  http://lug.boulder.co.us
> > Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
> > Join us on IRC: lug.boulder.co.us port=6667 channel=#colug
> > 
> > 
> > 
> 
> 
> 
> _______________________________________________
> Web Page:  http://lug.boulder.co.us
> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
> Join us on IRC: lug.boulder.co.us port=6667 channel=#colug





More information about the LUG mailing list