chris@asylum.gsfc.nasa.gov (Chris Shenton) (08/31/90)
I've got 4 SR10.2 machines on a ring, addresses 128.183.30.1, 2, 3, 4. I'd like to route through # .30.1 and out via its ether, address .10.55. I wasn't able to get it running under 9.7 and haven't been able to get it to work under 10.2. The gateway node gets to the rest of the ether world as well as the nodes on the ring. The ring nodes can telnet to the ring (.30.1) and ether (.10.155) side of the gate, but can't get past that to anywhere useful. What the solution, subnets? I'm using gateway masks: /etc/ifconfig dr0 128.183.30.1 netmask 255.255.255.0 /etc/ifconfig eth0 128.183.10.155 netmask 255.255.0.0 broadcast 0 and the other ring nodes are likewise masked 255.255.255.0 If I mask the ether side also 255.255.255.0 then the rest of the ethernet people yell at me about bogus broadcasts. The default route on the ring nodes is to the ring address of the gateway (.30.1). Some networkers here have mentioned Proxy ARP, but I'm not sure what it is, or how to do it. I can't be the only one using this ring/ether gateway configuration... How did you solve it? TIA! -- ____________________________________________________________________________ INET: chris@asylum.gsfc.nasa.gov (128.183.10.155) NASA/GSFC: Code 735 UUCP: ...!uunet!asylum.gsfc.nasa.gov!chris Greenbelt, MD 20771 SPAN: PITCH::CHRIS Fax: 301-286-9214 Phone: 301-286-6093
thompson@PAN.SSEC.HONEYWELL.COM (John Thompson) (09/01/90)
> I've got 4 SR10.2 machines on a ring, addresses 128.183.30.1, 2, 3, 4. > I'd like to route through # .30.1 and out via its ether, address .10.55. > I wasn't able to get it running under 9.7 and haven't been able to get it > to work under 10.2. > > The gateway node gets to the rest of the ether world as well as the nodes > on the ring. The ring nodes can telnet to the ring (.30.1) and ether > (.10.155) side of the gate, but can't get past that to anywhere useful. > > What the solution, subnets? I'm using gateway masks: > /etc/ifconfig dr0 128.183.30.1 netmask 255.255.255.0 > /etc/ifconfig eth0 128.183.10.155 netmask 255.255.0.0 broadcast 0 > and the other ring nodes are likewise masked 255.255.255.0 > > If I mask the ether side also 255.255.255.0 then the rest of the ethernet > people yell at me about bogus broadcasts. The default route on the ring > nodes is to the ring address of the gateway (.30.1). Some networkers here > have mentioned Proxy ARP, but I'm not sure what it is, or how to do it. <Open mouth, insert foot if I'm all fuc?ed up> Yell at the ethernet people. You're correct in your subnetting. The guys (nodes) on the ethernet are busy thinking that packets going to 128.183.30.2 (in the ring) are host 30.2 on network 128.183 -- THEIR OWN NETWORK! Because of this, they don't bother sending it to your gateway! > I can't be the only one using this ring/ether gateway configuration... How > did you solve it? We had the exact same problem. One of our ethernet machines was not using subnets. It could reach our gateways; our gateways could reach it; no other combo worked. When we set subnets up (all the other nodes already had them), everything was fine. John Thompson (jt) Honeywell, SSEC Plymouth, MN 55441 thompson@pan.ssec.honeywell.com As ever, my opinions do not necessarily agree with Honeywell's or reality's. (Honeywell's do not necessarily agree with mine or reality's, either) It doesn't matter whether you win or lose -- it's how you place the blame!
krowitz@RICHTER.MIT.EDU (David Krowitz) (09/01/90)
Your netmask should be the same for both sides of the gateway (ie. 255.255.255.0, since your network is 128.183 and your ringnet is subnet 30 and your ethernet is subnet 10). What bogus broadcasts do there seem to be? Routed junk? If so, you don't really even need to run routed ... just set the default route for all the non-gateway ringnet node to be the gateway node (after all, there's only one way to get off the ring) and then set the default route on the Apollo gateway to point at the gateway between your local ethernet (the one the Apollo gateway is attached to) and the your next hop to the Internet. That is how we do it here at MIT ... it's the official gospel local nodes do not run routed in any form (since even in quiet mode, some versions have the nasty habit of flushing routing tables of *everything*). Let *their* gateway worry about the routing! -- David Krowitz krowitz@richter.mit.edu (18.83.0.109) krowitz%richter.mit.edu@eddie.mit.edu krowitz%richter.mit.edu@mitvma.bitnet (in order of decreasing preference)
kerr@tron.UUCP (Dave Kerr) (09/02/90)
In article <9008311845.AA20019@pan.ssec.honeywell.com> thompson@PAN.SSEC.HONEYWELL.COM (John Thompson) writes: > >> I've got 4 SR10.2 machines on a ring, addresses 128.183.30.1, 2, 3, 4. >> I'd like to route through # .30.1 and out via its ether, address .10.55. >> I wasn't able to get it running under 9.7 and haven't been able to get it >> to work under 10.2. >> >> The gateway node gets to the rest of the ether world as well as the nodes >> on the ring. The ring nodes can telnet to the ring (.30.1) and ether >> (.10.155) side of the gate, but can't get past that to anywhere useful. >> >> What the solution, subnets? I'm using gateway masks: >> /etc/ifconfig dr0 128.183.30.1 netmask 255.255.255.0 >> /etc/ifconfig eth0 128.183.10.155 netmask 255.255.0.0 broadcast 0 >> and the other ring nodes are likewise masked 255.255.255.0 Make sure that the other hosts on your ethernet are also using subnets. Also, apollo's default to the broadcast addresss with the host part all 1's. In this case it would default to 128.183.10.255 for the eth0 side and 128.183.30.255 for the dr0 side and As far as I can tell, this is the correct form of the broadcast address. If you have Sun's on your ethernet, they default to a broadcast address with the host part equal to 0's. This will result in the gateway's not picking up each other's routing tables, and you'll have to change the broadcast on one or the other gateways. Also are aware of the routed patch for 10.2? Dave -- Dave Kerr (301) 765-4453 (WIN)765-4453 tron::kerr Internal WEC vax mail kerr@tron.bwi.wec.com from an Internet site kerr@tron.UUCP from a smart uucp mailer
chris@asylum.gsfc.nasa.gov (Chris Shenton) (09/04/90)
Thanks to all who answered my questions about TCP routing from ring to ether. The solution appears to be to set the net mask the same (255.255.255.0) on all ring and ether interfaces, on all nodes. Here's the kick in the groin, though: GSFC has a boatload of nodes, and the overlords have decided (for backward compatibility reasons, I suppose) that everyone on the ether must use 255.255.0.0 netmask, which basically says ``No netmasks''. Can anyone think of a workaround? Is there any good reason why I can't use that 2-byte mask on all my hardware? or will the gate get confused thinking that my gate addresses (128.183.10.155 and 128.183.30.1) are on the *same* network? Thanks in advance, again :-) -- ____________________________________________________________________________ INET: chris@asylum.gsfc.nasa.gov (128.183.10.155) NASA/GSFC: Code 735 UUCP: ...!uunet!asylum.gsfc.nasa.gov!chris Greenbelt, MD 20771 SPAN: PITCH::CHRIS Fax: 301-286-9214 Phone: 301-286-6093