RE: urgent TEQL problem

Linux Advanced Routing and Traffic Control

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

 



Thanks a lot, Grant!

I don't understand problem (1). Do you mean the routing between the two
subnets connected by the gateway? I think the gateway can route
correctly since I have 
"16.119.144.64   16.119.144.35   255.255.255.224 UG    0      0        0
eth1" in Node1's routing table, and a similar entry in Node2's routing
table. 

Problem (2) seems to be caused by "16.119.144.66   0.0.0.0
255.255.255.255 UH    0      0        0   teql0" in Node1's routing
table. There are two paths to 16.119.144.66, one with direct connection,
and one with the gateway. I don't know how to specify both cases at the
same time. As I said in my original post, It seems that the problem is
that when packets are forwarded by the teql0 interface to each local
Ethernet cards, the routing is still based on the routing entry for the
teql0_IP address in the routing table, instead of being based on the IP
of the local Ethernet interface that packets are passed to. Since I
can't both specify the gateway address and not specify at the same time,
there is only one link that teql finds working and pass packets to at
each time. But it sounds unreasonable for TEQL to be implemented that
way. Probably my configuration is wrong somewhere.

I think your idea of dummy interfaces will probably work, but I'll try
to change the gateway into a bridge first. Not sure which one is simpler
yet. 

Thanks a lot again!

-Ji

-----Original Message-----
From: lartc-bounces@xxxxxxxxxxxxxxx
[mailto:lartc-bounces@xxxxxxxxxxxxxxx] On Behalf Of Taylor, Grant
Sent: Thursday, June 23, 2005 5:03 PM
To: lartc@xxxxxxxxxxxxxxx
Subject: Re:  urgent TEQL problem

Based on your previous email and this email I see a couple problems with
your situation.

1)  Your ""Gateway will not be able to ""route correctly between two
interfaces on the same subnet.
2)  Node1 is ARPing out eth1 to try to find 16.119.114.66 on it's local
subnet as it should be there but it is not as it is on the other side of
a gateway.

If you do not want to do this on layer 2 and turn your gateway in to a
bridge I think you will have to do all of this on layer 3 and route
*BOTH* sides of teql0, how to do this I'm not sure of at the moment.  I
have a feeling you will need to ultimately try to reach an IP that is
not on any of your physical network cards but rather something on a
dummy interface that is accessible via a route using either interface to
get to.  I'll try to describe such a setup below.

Node 1:
- eth0:  <subnet 1>.1
- eth1:  <subnet 2>.1
- teql0: <subnet ?>.1
- dumy0: <subnet 3>.1

Node 2:
- eth0:  <subnet 1>.2
- eth1:  <subnet 4>.2
- teql0: <subnet ?>.2
- dump0: <subnet 5>.2

Gateway:
- eth0:  <subnet 2>.254
- eth1:  <subnet 4>.254

(Sitting her looking at this I'm not entirely sure that you even need
teql0 but rather ECMP routing.)

Node 1 routing table:
<subnet 3> via <subnet 1>.2   metric 2
<subnet 3> via <subnet 2>.254 metric 1

Node 2 routing table:
<subnet 5> via <subnet 1>.1   metric 2
<subnet 5> via <subnet 4>.254 metric 1

Gateway routing table:
<subnet 3> via <subnet 2>.1 metric 1
<subnet 5> via <subnet 4>.2 metric 1

This *SHOULD* (if I have things correct in my head) establish two routes
from <subnet 3> to <subnet 5> with the same overall metric of 2.
However to use ECMP you will need the metrics for both routes from
<subnet 3 or 5> to <subnet 5 or 3> to be the same on Node 1 and Node 2.
Thus I might modify the routing tables as such.

Node 1 routing table:
<subnet 3> via <subnet 1>.2   metric 2
<subnet 3> via <subnet 2>.254 metric 2

Node 2 routing table:
<subnet 5> via <subnet 1>.1   metric 2
<subnet 5> via <subnet 4>.254 metric 2

Gateway routing table:
<subnet 3> via <subnet 2>.1 metric 0
<subnet 5> via <subnet 4>.2 metric 0

As my pager goes off I realize that I have to submit a partial post back
to the list, but hopefully there is enough here to get a couple of
points across and enough for someone else to work with to help flesh out
this idea.



Grant. . . .

Li, Ji wrote:
> One more thing to add is that when I "ping 16.119.144.66" from Node1
> (16.119.144.66 is the teql0 of Node2), all messages I saw from
> Node1.eth1 are ARP messages "Who has 16.119.144.66? Tell
16.119.144.33".
> 
> Thanks,
> -Ji
_______________________________________________
LARTC mailing list
LARTC@xxxxxxxxxxxxxxx
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
_______________________________________________
LARTC mailing list
LARTC@xxxxxxxxxxxxxxx
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc


[Index of Archives]     [LARTC Home Page]     [Netfilter]     [Netfilter Development]     [Network Development]     [Bugtraq]     [GCC Help]     [Yosemite News]     [Linux Kernel]     [Fedora Users]
  Powered by Linux