[c-nsp] Pseudowire and load-balancing - revisit

James Jun james at towardex.com
Tue Feb 12 13:18:03 EST 2019


Hey all,

I have a PW scenario that looks like this:

 Customer --- PE1 ---- P ----[4x10GE LAG]---- PE2 -- Customer

PE1, P and PE2 routers are all ASR9K.

EoMPLS PW is delivered to customer; both PE1 and PE2 configure FAT-PW as follows:

!
l2vpn
 load-balancing flow-src-ip
 !
 pw-class fat
  encapsulation mpls
   control-word
   load-balancing
    flow-label both static
   !
 !! xconnect p2p uses pw-class fat under neighbor

Majority of customer traffic direction is ingressing from PE1 and egressing at PE2.

So when customer is transporting IP traffic on the provided pseudowire, load balancing works
pretty nicely.  Customer passenger traffic appears to be evenly spread across the LAG between
P and PE2.

However, when customer is running their own MPLS/IP network over the provided PW, load balancing
completely breaks with the above fat-pw setup.  Imbalance occurs and customer traffic is lopsided
only on 1 member of the 4x10G LAG between P and PE2.

I then have customer configure multiple RSVP-TE LSPs between their own devices at each end of the
PW and load balance them (EMCP) via igp-shortcuts/auto-route announce. This ensures customer sent
passenger traffic entering the PW has many different MPLS labels injected by the customer, with
the hope that PE1 will be able to properly generate unique hash as it encapsulates and insert them
into fat-pw flow label.

But no -- all traffic on the PW is lopsided and sent down one member link on the LAG.  It seems
when PW passenger traffic is encapsulated inside customer's own MPLS, ASR9K is unable to hash and
can only load balance on VC label.  Even when customer sends their passenger traffic with multiple
different MPLS labels, ASR9K is not hashing, unless it is IP traffic.

Anyone else run into this issue? I haven't tried removing control-word along with FAT-PW, but I don't
believe that will make a difference in this case.  I think the issue is that A9K PE's are not looking
deeper into passenger traffic on L2CKT when customer themselves are running MPLS.


James


More information about the cisco-nsp mailing list