Hello,
I have been searching in the archives of the lists for "success stories"
on the subject, but I could not find one.
The question is: has anyone run successfully linux mpls with a cisco box?
I couldn't.
I have successfully tested static label swapping, ldp and rsvpd before
between linux boxes (so as far as I can see it, my kernel, zebra, etc
are ok)
I have a cisco box connected to a linux machine. The goal is for now to
have a ldp session between the two.
They both run ospf, I have a dummy intf. on the linux and another
physical interface on the cisco, and ospf runs just fine (I can ping any
of the interfaces that I am interested in)
The mplsd config is the following:
[root@w02gva etc]# cat mplsd.conf
mpls ldp
!
interface eth2
mpls ldp
The cisco setup is like this:
ip cef distributed
no ip domain-lookup
mpls label protocol ldp
mpls ldp logging neighbor-changes
no mpls traffic-eng auto-bw timers frequency 0
interface FastEthernet2/0/0
ip address 192.168.0.2 255.255.255.0
no ip directed-broadcast
full-duplex
mpls ldp discovery transport-address 192.168.0.2
mpls label protocol ldp
tag-switching ip
no cdp enable
end
The cisco is a rsp7000 with ios 12.0.25S. I have tried a 12.2.14S1 with
the same results (lack of results, that is)
The output of show mpls ldp neighbor on cisco is null. The one on zebra
is like this:
w02gva# show ldp neighbor
Peer LDP Ident: 10.0.0.1:0; Local LDP Ident: 192.168.2.1:0
TCP connection: 0.0.0.0.646 - 192.168.0.2.646
State: NON-EXIST; Msgs sent/recv: 0/0; UNSOLICITED
Up time: 12194d 14:36:42
LDP discovery sources:
eth2
Addresses bound to peer:
As I remember from the linux setup, the Addresses bound to peer section
should not be null. Also, from the rfc, state: NON-EXIST doesn't sound
good (I don't remember what the linux setup said). I also did an
ethereal capture on the interface. It seems like the linux box tries to
initiate a tcp connection to the cisco box on tcp port 646, it fails,
and never tries again. I also noted the csn error message in mplsd, but
i get the same results if I try to comment out the various csn checks in
the source code.
I would be glad to receive at least a configuration dump of a succesful
setup. Of course, comments on my setup are also welcome.
Thank you,
Dan Nae
|