Load-balance problem using VPLS as transport

[Anatoly_Machekh]

Hi, everyone!

Here is lab network to test balancing on client ports on Extreme Network switches (X460 in lab, but we have the same problem with X670-G2 in production).
We are interested in sharing group 21 (ports 21, 22) on X460-1.

From Agilent (IP tester) IP traffic goes to Lab-M01 to 18 different VRFs (1001-1009 and 2001-2009). Lab-M01 sends out traffic in VRF 1001 through 1009 with one MPLS label, in VRF 2001 through 2009 with two MPLS labels. When we configure X460-1 and X460-2 as simple L2 switches with one vlan from sharing group 21 on X460-1 to sharing group 21 on X460-2, then balancing works.

1 MPLS label:

X460-1.8 # sh port 21,22,28 utilization packets
Port Link Rx Peak Rx Tx Peak Tx
State pkts/sec pkts/sec pkts/sec pkts/sec
================================================================================
21 A 0 1 6632 8034
22 A 0 0 13238 16037
28 A 19857 23909 0 1
================================================================================
2 MPLS labels:

X460-1.10 # sh port 21,22,28 utilization packets
Port Link Rx Peak Rx Tx Peak Tx
State pkts/sec pkts/sec pkts/sec pkts/sec
================================================================================
21 A 2 3 8725 10678
22 A 0 1 13118 13208
28 A 23657 23731 3 3
================================================================================
The problem is that when we configure VPLS on X460-1 and X460-2 to pass the same traffic balancing stops working. All traffic goes to one port (either 21 or 22 depending...).
What can we do to make balancing work with VPLS?

Thanks,
Anatoly

[Stephane_Grosj1]

Then configure LSP load sharing.

[Anatoly_Machekh]

Here is information on X460:
X460-1.1 # sh conf | i sharing enable sharing 21 grouping 21-22 algorithm address-based custom lacp
X460-1.2 # sho sharing
Load Sharing Monitor
Config Current Agg Min Ld Share Flags Ld Share Agg Link Link Up
Master Master Control Active Algorithm Group Mbr State Transitions
================================================================================
21 21 LACP 1 custom A 21 Y A 2
custom 22 Y A 0
================================================================================
Link State: A-Active, D-Disabled, R-Ready, NP-Port not present, L-Loopback
Minimum Active: (<) Group is down. # active links less than configured minimum
Load Sharing Algorithm: (L2) Layer 2 address based, (L3) Layer 3 address based
(L3_L4) Layer 3 address and Layer 4 port based
(custom) User-selected address-based configuration
Custom Algorithm Configuration: ipv4 L3-and-L4, xor
Distribution Mode Flags:
A - All: Distribute to all members
L - Local Slot: Distribute to members local to ingress slot
P - Port Lists: Distribute to per-slot configurable subset of members
Number of load sharing trunks: 1

[Stephane_Grosj1]

This is a LAG, correct? What config is on it (algo, hashing)? I'd recommend custom there. If not good enough, since 15.4 we have a feature - LSP Load Sharing - (for VPLS and RSVP-TE) to distribute the traffic more efficiently, using more LSPs (p.1123 of 22.1 User Guide).