[j-nsp] aggregate-ethernet vs mpls ?

[Alexandre Snarskii]

On Thu, Jun 22, 2006 at 09:18:10AM +0300, Erdem Sener wrote:
> I'm just guessing, but how 'bout MTU settings on the Foundry side?

'jumbo' one. somewhere about 10000 bytes. 
MTU on foundry is not an likely source of problem - as I noted
we have no problems with transit traffic, but unreachability of
some hosts means that we have problem even with small (tcp syn/tcp syn+ack) 
packets to _some_ destinations..

> 
> On 6/21/06, Alexandre Snarskii <snar at paranoia.ru> wrote:
> >
> >Hi!
> >
> >We are facing strange problem with mpls over aggregate-ethernet link.
> >This aggregate-ethernet consists of 3x old (revision-a) ge-lx
> >cards, which connected to foundry switches and connecting
> >our routers over 10ge link. Topology is pretty simple:
> >
> >R1 <-3xGE-> S1 <-10GE-> S2 <-3xGE-> R2 <-stm-16-> R3
> >
> >Configuration is just copied from old one-ge link between routers R1 and R2
> >with only adjustments needed to ae to work.
> >
> >Problem is the next: our management station, connected to routing instance
> >on R1 (instance-type vrf, used for management) is not able to access _some_
> >addresses (l2 switches really) in the same routing instance on R2....
> >At the same time all l2circuits R1-R3 and R2-R3 works without problems,
> >and all addresses in the same VRF under R3 is accessible from management
> >station.. More than, all switches accessible in the same instance from R3..
> >So, for me it looks like we have a problem only with traffic, which enters
> >R2 only over aggregate-ethernet, and having only the top-label on the 
> >stack.
> >Traffic entering R2 over sonet with just top-label, or any traffic
> >entering R2 with more than one label seemed unaffected.
> >
> >Any clues on this issue ?
> >
> >The fact that the just _some_ of switches and only under R2 is inaccessible
> >is the most mystical for me... I'm able to telnet to x.x.x.66, but
> >telnet x.x.x.67 gets me "connect timeout..." (and at the same time i'm able
> >to connect them from R2 using 'telnet routing-instance XXXX', so, that's 
> >not
> >the connectivity problem).
> >(And, yes, we know a hint about setting static-routes inside vrf instance,
> >and they are present).
> >
> >PS: both routers in question is M20, JunOS 7.0R2.7, cards in aggregate
> >are ge-lx, rev. A, customer-facing cards is ge-lh, rev B.
> >Both switches are Foundry FESX424F+2XG, running 02.4.00dT3e1.
> >Configuration on juniper side is straightforward:
> >
> >snar at R2> show configuration interfaces ae0
> >description "GE-aggregate to R1 via Founfry";
> >vlan-tagging;
> >mtu 4488;
> >aggregated-ether-options {
> >   minimum-links 1;
> >   link-speed 1g;
> >}
> >unit 0 {
> >   description BackBone;
> >   vlan-id 10;
> >   family inet {
> >       mtu 1500;
> >       address A.B.C.133/30;
> >   }
> >   family mpls {
> >       mtu 4470;
> >   }
> >}
> >
> >configuration on foundry side is simple too:
> >
> >trunk ethe 21 to 24
> > config-trunk-ind
> > port-name "somename" ethernet 21
> > port-name "somename" ethernet 22
> > port-name "somename" ethernet 23
> > port-name "somename" ethernet 23
> >
> >
> >_______________________________________________
> >juniper-nsp mailing list juniper-nsp at puck.nether.net
> >http://puck.nether.net/mailman/listinfo/juniper-nsp
> >
> 
> 
> -- 
> Erdem