Re: [mpls-linux-devel] MPLS-Linux-1.955 FC5

[James R. L. <jl...@mi...>]

I see some difference with my testing setup.  First, I create all
my routes via vtysh with LDP running at that time, in otherwords:

I start zebra, ospfd, and ldpd, then start my configuration via vtysh.
I configure EVERYTHING via vtysh.  The only interface on the box that is
not configure via vtysh is the managment interface which does not
participate in routing.

Is there a reason you do not use the system scripts to start zebra ospfd
and ldpd?  (ie service zebra start; service ospfd start; service ldpd start)

Have you tried turning on zebra debugging? (in vtysh: debug zebra kernel)
You will need to have logging configured for this to produce anything
useful.

I've run your exact configuration on my network at home at it works just fi=
ne.
There must be something with the order of events that is causing an issue.
Obviously I'd like to fix this issue, but I need to know what "mis-ordering"
is causing it and be able to reproduce it before I can fix it.

At this point I do not believe you have ever gotten LDP to work for you
so forgive me for being sceptical.

On Tue, Mar 13, 2007 at 09:27:53AM +0100, Sienes Rodr?guez Gonzalo wrote:
> Hi James:
> =20
> I?m trying two different experiments with the following scenario and thes=
e are the results.
> =20
>                        eth0     eth1               eth1     eth0
>         12.0.0.1-------PE1---------------------PE2-------13.0.0.1
>                                     .1  10.0.0.0/24    .2
> =20
>=20
> 1?    - I create static routes from PE1 to 13..0/24 and PE3 to 12..0/24.=
=20
> =20
>        - Then I launch zebra and ldpd typing zebra -d and ldpd -d by the =
command line.=20
>          (I load the mpls4 module since system boots).
> =20
>        - I configure PE1 and PE3 as LSRs by vtysh:
>  =20
>          > "mpls ldp"
>          > int eth1 (in each LSR): "mpls ip" and "mpls labelspace 0"
> =20
>        - Then I hang a host at 12..0/24 to send packets to 13.1 through 1=
0..0/24.
> =20
>  - RESULTS:
> =20
>   > Those packets can't even arrive to their destination, cause the stati=
c routes that
>     I configured DISAPEAR. From vtysh (sh ip route) I see the new routes =
with their LABEL,=20
>     but from the command line (ip route) I just see the default routes at=
 PE1 and PE3.
> =20
>   > The NHLFE and ILM entries for that routes are created properly:
>     =20
>       NHLFE entry key 0x00000002 mtu 1496 propagate_ttl
>            push gen 10000 set eth1 ipv4 10.0.0.2  (0 bytes, 0 pkts)
> =20
>       ILM entry label gen 10000 labelspace 0 proto ipv4
>               pop peek  (0 bytes, 0 pkts)
> =20
>   > So the problem is that with those entries the control component isn't=
 able to change=20
>     the kernel routes.
>    =20
> 2?    - I launch zebra and ospfd.=20
> =20
>        - Then I configure ospf by vtysh: "router ospf" -> "network 10.0.0=
.0/24 area 0" (PE1 and PE3)
>           -> "network 12.0.0.0/24 area 0" (PE3)
>           -> "network 13.0.0.0/24 area 0" (PE1)
> =20
>       - After new routes are learned I close vtysh, launch ldpd and confi=
gure PE1 and PE3 as LSRs by vtysh:
>  =20
>          > "mpls ldp"
>          > int eth1 (in each LSR): "mpls ip" and "mpls labelspace 0"
> =20
>       - Then I hang a host at 12..0/24 to send packets to 13.1 through 10=
..0/24.
> =20
>       - RESULTS:
> =20
>       > Those packets arrive to their destination, but without a label. >=
=46rom vtysh (sh ip route) I see the new ospf=20
>          routes with their hop count (not any label) and from the command=
 line (ip route) I see the same routes learned by ospf.
> =20
>       > The NHLFE and ILM entries for that routes are created properly:
> =20
>          NHLFE entry key 0x00000002 mtu 1496 propagate_ttl
>               push gen 10000 set eth1 ipv4 10.0.0.2  (0 bytes, 0 pkts)
> =20
>          ILM entry label gen 10000 labelspace 0 proto ipv4
>               pop peek  (0 bytes, 0 pkts)
> =20
>      > So the problem, again, is that with those entries the control comp=
onent isn't able to change=20
>        the kernel routes.
> =20
> Any idea of what's happening? As I told you, if you could provide me some=
 documentation for understanding how the code is organized, maybe I could h=
elp.
> =20
> Thanks for your time,
> Gonzalo.
>=20



--=20
James R. Leu
jl...@mi...