[j-nsp] MX480 L2/L3 with some packet loss for routed traffic on one interface only

Nicolaj Kamensek nsp at accelerated.de
Thu Aug 9 11:13:18 EDT 2012


Hello list,

I am a little lost here trying to find the cause for a little packetloss 
(up to 1%) I'm seeing in our network. Here is the situation:

a MX480 is acting as a router and layer2 device using irb interfaces for 
the routing. It connects two big switches with 10GE each. One is a Cisco 
6509 with sup32, the other one a Brocade SuperX. Rapid spanning-tree is 
active.
I am seeing 0% packet-loss for everything connected to the Cisco. There 
is also no packet-loss for switched traffic that passes both switches 
and the router as well. But there is packet loss for more or less all 
routed traffic coming from the SuperX which variies between 0% and 1%.

Since only routed traffic is affected, one might safely asume that this 
is not related to the switching hardware. The link itself can't be the 
cause as well. It's furthermore unlikely that the router itself is 
broken since everything coming from the cisco works fine.

Just to mention that: routed traffic might be externally or internally, 
it doesn't make a difference.

This is the configuration for the interface facing the SuperX:

nico at MX480> show configuration interfaces xe-1/1/0
flexible-vlan-tagging;
encapsulation flexible-ethernet-services;
unit 0 {
     family bridge {
         interface-mode trunk;
         vlan-id-list 0-4094;
     }
}


The other interface is actually an aggregated link consisting of 2x 10G 
going to the Cisco. It looks pretty much the same but since there are 
also a few routed-only units, the vlan-id-list looks a little different 
in order to be able to set up those routed ports.


I am now seeking for hints what to do. Might the currently unnecessary 
"flexible-vlan-tagging & flexible-ethernet-services" be the cause? This 
is just to keep options open but is right now not required though and 
will most likely consume a little more memory than required (shouldn't 
be a problem with the MX though).
Might the xe-1/1/0 interface be broken? Unlikely in my opinion since l2 
traffic works.


Any hint will be appreciated! Thanks in advance!


Nico


More information about the juniper-nsp mailing list