[e-nsp] MPLS not coming back after link failure

[Michał Margula]

W dniu 18.09.2012 11:56, Stephane Grosjean pisze:
> Hi Michal,
> 
> As I don't know your setup, I'm only speculating.
> Can you detail the layout of the network a bit? On what BGP is relying, etc.
> 
> 
> Regards,
> 
> Stef

OK!

I have two 1gig optical links between Cisco 6500-SUP720-3bxl and Extreme
X480-24x.

172.16.250.1/24 - Cisco, Link 1
172.16.249.1/24 - Cisco, Link 2

172.16.250.2/24 - Extreme, Link 1
172.16.249.2/24 - Extreme, Link 2

172.16.251.1/32 - Loopback on Cisco
172.16.251.2/32 - Loopback on Extreme

OSPF running on both links and advertising loopbacks. OSPF is UP,
everything is pingable and so on. MPLS LDP running on both links just
fine, defined three VPWS (Extreme side):

create l2vpn vpws MPLS-Jambox fec-id-type pseudo-wire 10
configure l2vpn vpws MPLS-Jambox add service vman Jambox
configure l2vpn vpws MPLS-Jambox mtu 1504
configure l2vpn vpws MPLS-Jambox add peer 172.16.251.1

And so on... Everything works fine, passing by traffic to Cisco. Then
after one of following things:

- physical disconnect of both links
- port shutdown on Cisco of both ports
- port disable on Extreme of both ports

And then recovering (reconnecting a cable, no shutdown or enable) LDP
fails to come back up, altough OSPF works just fine, everything is
pingable and so on. Extreme doesn't see its neighbors, Cisco doesn't see
Extreme. I have to do "disable mpls ldp vlan X" and "enable mpls ldp
vlan X" to make it work again. There is nothing I can do on Cisco to fix
it, so it is clearly problem on Extreme.

Conclusion: LDP doesn't recover after link failure. Why? Probably
software bug?

-- 
Michał Margula, alchemyx at uznam.net.pl, http://alchemyx.uznam.net.pl/
"W życiu piękne są tylko chwile" [Ryszard Riedel]