Re: [mpls-linux-devel] LDP Configuation

[<gsr...@al...>]

Hi!

Thanks for your information.

I decided to test the new configuration in the following scenario, only =
with
two routers to realized what LDP really does. =20

      13.0.0.0/24   .1   .1             .2   .1   14.0.0.0/24
               -------PE1-----------------PE3-------
                  eth2   eth1         eth0   eth2
                            12.0.0.0/24  =20

These are the .conf files I use for PE1:

ZEBRA
------------------------------------------------------------------
!
hostname PE1
password root
log file /var/log/quagga/zebra.log
debug zebra events
debug zebra kernel
!
!
interface lo
!
!
interface eth1
!
interface eth2
!
!

OSPFD
------------------------------------------------------------------
!
hostname PE1
password root
log file /var/log/quagga/ospfd.log
!
!
!
interface eth1
 description 12.0.0.1 netmask 255.255.255.0 broadcast 12.0.0.255 ! =
interface
eth2  description 13.0.0.1 netmask 255.255.255.0 broadcast 13.0.0.255 !
interface lo ! router ospf  ospf router-id 0.0.0.1  network 12.0.0.0/24 =
area
0  network 13.0.0.0/24 area 0 ! line vty !

LDPD
------------------------------------------------------------------
!
hostname PE1
password root
log file /var/log/quagga/ldpd.log
!
interface eth1
 mpls ip
!
interface eth2
!
interface lo
!
!
line vty
!

VTYSH
------------------------------------------------------------------
!
service integrated-vtysh-config
hostname PE1
!

As you suggested, after running all those daemons (zebra -d, ospfd -d,
ldpd-d) I type "vtysh" to finally set the proper LDPD commands: "mpls =
ldp",
and in each interface of 12.0.0.0/24 subnet, "mpls ip" and "mpls =
labelspace
0".

I got some problems with mpls labelspace, cause sometimes when I type =
the
command occurs an I/O error and the daemon zebra is killed. In the end, =
when
it works, I capture the traffic and this is what happens:

1.Both PEs pass the stage of discovering with HELLO messages.

2.I can see INITIALIZATION messages in both ways, but the ones from 12.2 =
to
12.1 have a TCP checksum error. However, the LDP message seems to be
correct. With correct LSRs identifiers and these session parameters:

        -Downstream unsolicited mode.
        -Loop detection disabled.
        -MPLS labelspace 0.

3.The KEEPALIVEs in the way PE1 also travel with checksum error as well =
as
all types of LDP messages but HELLOs. The ADDRESS messages are set with =
the
correct values of neighbor address.

4.The LABEL MAPPING messages carry the value of a label inside, which is =
the
same for both ways, 10000.

5.The problem is that after distributing the labels, 12.1 sends to 12.2 =
a
NOTIFICATION message with the status: fatal error, session rejected and =
then
the process starts again, now exchanging different values of labels =
(always
the same in both ways) 10001, 10002 and so on.

This is what I get in PE1 and the same in PE2 with other addresses:

PE1# sh ldp neig
Peer LDP Ident: 14.0.0.1:0; Local LDP Ident: 13.0.0.1:0
        TCP connection: 0.0.0.0.646 - 12.0.0.2.45099
        State: OPERATIONAL; Msgs sent/recv: 489/1459; UNSOLICITED
        Up time: 02:00:24
        LDP discovery sources:
          eth1
        Addresses bound to peer:
          12.0.0.2        14.0.0.1

I try to do: ping -c3 14.0.0.1 from PE1, but the ICMP packets don't =
carry
any LDP label. Besides, MPLS NHLFE and MPLS ILM tables are empty.

Is this the expected behavior?

Thanks in advance,

Gonzalo