Skip to content

weighted_ecmp_test

Jump to bottom
Github Action edited this page Apr 30, 2024 · 3 revisions

RT-2.13: Weighted-ECMP for IS-IS

Summary

This is to ensure that,

  • Implementations can be configured for weighted equal cost multipath (ECMP) routing for IS-IS neighbors that are one hop away.

  • When WECMP is enabled, traffic destined to an IS-IS route represented by a multipath set of next-hop interfaces will be unequally distributed across the interfaces based on their bandwidth.

Testbed type

TESTBED_DUT_ATE_8LINKS

Topolgy

Each LAG bundle below is made up of 2x100G ports.

graph LR;
A[ATE1:LAG1] <-- IBGP+IS-IS --> B[LAG1:DUT];
C[DUT:LAG2] <-- IBGP+IS-IS --> D[LAG1:ATE2];
E[DUT:LAG3] <-- IBGP+IS-IS --> F[LAG2:ATE2];
G[DUT:LAG4] <-- IBGP+IS-IS --> H[LAG3:ATE2];
Loading

Procedure

In the topology above,

  • Configure 1xLAG interface between ATE1<->DUT and 3xLAG interfaces between DUT and ATE2. Each LAG interface is expected to be of 2x100Gbps

  • Configure IPv4 and IPv6 L2 adjacencies between DUT and ATE LAG bundles. Therefore, DUT will have 1xIS-IS adjacency with ATE1 i.e. DUT:LAG1<->ATE1:LAG1, and 3xIS-IS adjacencies with ATE2 i.e. DUT:LAG2<->ATE2:LAG1, DUT:LAG3<->ATE2:LAG2 and DUT:LAG4<->ATE2:LAG3

    • /network-instances/network-instance/protocols/protocol/isis/global/afi-safi

    • /network-instances/network-instance/protocols/protocol/isis/global/config/level-capability, set to LEVEL_2

    • /network-instances/network-instance/protocols/protocol/isis/levels/level/config/metric-style set to WIDE_METRIC

  • Configure IPv4 and IPv6 IBGP peering between both ATEs and the DUT using their loopback addresses for both IPv4 and IPv6 address families.

    • /network-instances/network-instance/protocols/protocol/bgp/peer-groups/peer-group/afi-safis/afi-safi/config

  • Attach a network with an IPv4 and an IPv6 prefix to ATE2 and have it advertise these prefixes over its IBGP peering with the DUT. The DUT in turn should advertise these prefixes over its IBGP peering with ATE1

    • Please use IPv4 prefix = 100.0.1.0/24 and IPv6 prefix = 2001:db8:64:64::/64

  • Similarly, attach a different network to ATE1 with IPv4 and IPv6 prefixes and advertise the same over its IBGP peering with the DUT.

    • Please use IPv4 prefix = 100.0.2.0/24 and IPv6 prefix = 2001:db8:64:65::/64

  • On the DUT, enable WECMP loadbalancing for multipath IS-IS routes and set the load-balancing-weight to use LAG bandwidth.

    • /network-instances/network-instance/protocols/protocol/isis/global/config/weighted-ecmp set to Enabled

    • /network-instances/network-instance/protocols/protocol/isis/interfaces/interface/weighted-ecmp/config/load-balancing-weight set to Auto

RT-9.1: Equal distribution of traffic

  • Start 1024 flows from IPv4 addresses in 100.0.2.0/24 to 100.0.1.0/24

  • Start 1024 flows from IPv6 addresses in 2001:db8:64:65::/64 to 2001:db8:64:64::/64

Verification

  • Ensure that the DUT has learnt the routes for prefixes 100.0.1.0/24 and 2001:db8:64:64::/64 over IBGP. Following paths

    • /network-instances/network-instance/afts/next-hops/next-hop/state/ip-address

  • Ensure that the DUT has learnt routes to the IPv4 and IPv6 loopback addresses of ATE2. It is expected that these prefixes are reachable via 3 different Next-Hop addresses corresponding to the LAG1, LAG2 and LAG3 interfaces on ATE2.

  • It is expected that the IS-IS instance in DUT will equally distribute the traffic received on DUT:LAG1 over the LAG bundles corresponding to ATE2:LAG1, ATE2:LAG2 and ATE2:LAG3 when the 3 LAG bundles have the same bandwidth available.

    • Traffic distribution between DUT:LAG2, DUT:LAG3 and DUT:LAG4 is expected to be ~33% each of the total traffic received on DUT:LAG1.

    • Check for the following paths

      • /network-instances/network-instance/protocols/protocol/isis/global/state/weighted-ecmp, should be true

      • /network-instances/network-instance/protocols/protocol/isis/interfaces/interface/weighted-ecmp/state/load-balancing-weight, should be auto

      • /interfaces/interface/state/counters/out-pkts

      • /interfaces/interface/state/counters/in-pkts

RT-9.2: Unequal distribution of traffic

  • Stop traffic from RT-9.1 and introduce a failure by disabling one of the member interfaces in ATE2:LAG1.

  • Restart 1024 flows from IPv4 addresses in 100.0.2.0/24 to 100.0.1.0/24

  • Restart 1024 flows from IPv6 addresses in 2001:db8:64:65::/64 to 2001:db8:64:64::/64

Verification

  • It is expected that the IS-IS instance in DUT will unequally distribute the traffic received from ATE1:LAG1 over the LAG bundles corresponding to ATE2:LAG1, ATE2:LAG2 and ATE3:LAG3.

    • Traffic on DUT:LAG2 is expected to be ~20% while traffic on DUT:LAG3 and DUT:LAG4 is expected to be ~40% each of the total traffic received on DUT:LAG1. If the traffic is not unequally shared between the DUT LAG bundles towards ATE2 then this test is a failure.

    • Check for the following paths

      • /interfaces/interface/state/counters/out-pkts

      • /interfaces/interface/state/counters/in-pkts

Config paths

Telemetry Parameter Coverage

OpenConfig Path and RPC Coverage

The below yaml defines the OC paths intended to be covered by this test. OC paths used for test setup are not listed here.

TODO(OCPATH): Container path originally part of spec that needs to be separated into leaves: /routing-policy/defined-sets/prefix-sets/prefix-set:

paths:
  ## Config Paths ##
  /network-instances/network-instance/protocols/protocol/bgp/peer-groups/peer-group/config/peer-group-name:
  /network-instances/network-instance/protocols/protocol/bgp/peer-groups/peer-group/config/peer-as:
  /network-instances/network-instance/protocols/protocol/bgp/peer-groups/peer-group/afi-safis/afi-safi/config/afi-safi-name:
  /network-instances/network-instance/protocols/protocol/bgp/peer-groups/peer-group/afi-safis/afi-safi/config/enabled:
  /network-instances/network-instance/protocols/protocol/isis/global/afi-safi/af/config/afi-name:
  /network-instances/network-instance/protocols/protocol/isis/global/afi-safi/af/config/safi-name:
  /network-instances/network-instance/protocols/protocol/isis/global/afi-safi/af/config/enabled:
  /network-instances/network-instance/protocols/protocol/isis/global/config/level-capability:
  /network-instances/network-instance/protocols/protocol/isis/levels/level/config/metric-style:
  /network-instances/network-instance/protocols/protocol/isis/global/config/weighted-ecmp:
  /network-instances/network-instance/protocols/protocol/isis/interfaces/interface/weighted-ecmp/config/load-balancing-weight:
  /routing-policy/defined-sets/prefix-sets/prefix-set/prefixes/prefix/config/ip-prefix:
  /routing-policy/defined-sets/prefix-sets/prefix-set/prefixes/prefix/config/masklength-range:
    value: exact
  /routing-policy/policy-definitions/policy-definition/config/name:
  /routing-policy/policy-definitions/policy-definition/statements/statement/config/name:
  /routing-policy/policy-definitions/policy-definition/statements/statement/conditions/match-prefix-set/config/prefix-set:
  /routing-policy/policy-definitions/policy-definition/statements/statement/conditions/match-prefix-set/config/match-set-options:
  /routing-policy/policy-definitions/policy-definition/statements/statement/actions/config/policy-result:
    value: ACCEPT_ROUTE
  /network-instances/network-instance/protocols/protocol/bgp/peer-groups/peer-group/afi-safis/afi-safi/apply-policy/config/import-policy:
  /network-instances/network-instance/protocols/protocol/bgp/peer-groups/peer-group/afi-safis/afi-safi/apply-policy/config/export-policy:

  ## State Paths ##
  /network-instances/network-instance/protocols/protocol/isis/global/state/weighted-ecmp:
  /network-instances/network-instance/protocols/protocol/isis/interfaces/interface/weighted-ecmp/state/load-balancing-weight:
  /interfaces/interface/state/counters/out-pkts:
  /interfaces/interface/state/counters/in-pkts:

rpcs:
  gnmi:
    gNMI.Subscribe:
    gNMI.Set:
Clone this wiki locally