Velocloud Edge - Route auto-correction not happening correctly on cluster.

Products

• VMware VeloCloud SD-WAN

Issue/Introduction

Route auto correction is not happening when spoke moves from one cluster member to another and the new cluster member starts receiving the route from the spoke.

edge:Edge-1-1-2:~# debug.py --routes 61.0.0.1 0
Address          Netmask       Type   Gateway     Next Hop Name                           Next Hop ID  Destination Name                         Dst LogicalId  Reachable  Metric  Preference  Flags  Vlan         Intf  Sub IntfId   MTU  SEG
61.0.0.0   255.255.255.0  edge2edge       any  Spoke  c204775d-98ec-47bb-a54e-b0a50884bd84  Spoke            c204775d-98ec-47bb-a54e-b0a50884bd84       True       0         999   DSBR     1          any         N/A  1500    0
61.0.0.0   255.255.255.0        any  10.0.1.5               N/A                                   N/A               N/A                                   N/A       True       0     1000000     SB     1  br-network1         N/A   N/A    0
P - PG, B - BGP, D - DCE, L - LAN SR, C - Connected, O - External, W - WAN SR, S - SecureEligible, R - Remote, s - self, r - recursive, H - HA, m - Management, n - nonVelocloud, v - ViaVeloCloud, A - RouterAdvertisement, c - CWS, a - RAS, M - MTGRE, g - Global PG Static

Environment

Spoke edge connected to a cluster and same route is advertised via BGP to spoke and all cluster members.

Cause

Issue is tracked under issue #149885

Root cause analysis: With gateways running on 5.4, routes from spokes would be advertised to all the hubs irrespective of spoke-hub association exists or not. Due to this the route from the spoke would be advertised to all the members of the cluster and would be present in the remote routes table. When there is cluster member reassignment, gateway would not send the routes of spoke to the new cluster member again and since the trigger for auto correction happens either during BGP underlay route learning or overlay routes receiving path, this would end up not triggering the auto-correction.

Resolution

Fix will be available in 6.1.0.0

Workaround: Run route reinit on the cluster member connected to the spoke.