Technical Tip: ECMP neighbor flap causes all routes to refresh

[nwynn]

Description	This article describes the FortiOS route refresh logic that applies when you have BGP neighbors with ecmp routes, and one neighbor flaps. Causing all ecmp routes to refresh.
Scope	FortiOS 5.x, 6.x and 7.x.
Solution	The following cli output shows the routing table behavior of a BGP ecmp neighbor flap, and all routes on both neighbors refreshing. 2 BGP neighbors are configured: FGT-1 # show router bgp config router bgp     set as 65024         set ebgp-multipath enable             config neighbor                 edit "100.64.100.48"                     set advertisement-interval 1                     set soft-reconfiguration enable                     set remote-as 19855                     set send-community standard                     set send-community6 disable                     set keep-alive-timer 3                     set holdtime-timer 9                 next             edit "100.65.100.48"                 set advertisement-interval 1                 set soft-reconfiguration enable                 set remote-as 19855                 set send-community standard                 set send-community6 disable                 set keep-alive-timer 3                 set holdtime-timer 9             next         end Neighbor 100.64.100.48 recently went down at 00:34:54 uptime. Neighbor 100.65.100.48 remained stable at 02:02:27 uptime. FGT-1 # get router info bgp summary VRF 0 BGP router identifier 100.126.0.24, local AS number 65024 BGP table version is 134 23 BGP AS-PATH entries 1 BGP community entries Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd 100.64.100.48 4 19855 324888 362342 133 0 0 00:34:54 263 100.65.100.48 4 19855 324226 364067 132 0 0 02:02:27 263 Total number of neighbors 2 Note that all BGP routes reset uptime. Even though only one neighbor flapped. FGT-1 # get router info routing-table bgp Routing table for VRF=0 B 10.4.7.0/24 [20/0] via 100.65.100.48, wan1, 00:34:27 [20/0] via 100.64.100.48, wan2, 00:34:27 B 10.10.152.0/24 [20/0] via 100.65.100.48, wan1, 00:34:27 [20/0] via 100.64.100.48, wan2, 00:34:27 B 10.11.12.0/24 [20/0] via 100.65.100.48, wan1, 00:34:27 [20/0] via 100.64.100.48, wan2, 00:34:27 B 10.11.112.0/24 [20/0] via 100.65.100.48, wan1, 00:34:27 [20/0] via 100.64.100.48, wan2, 00:34:27 B 10.11.212.0/24 [20/0] via 100.65.100.48, wan1, 00:34:27 [20/0] via 100.64.100.48, wan2, 00:34:27 B 10.14.0.0/16 [20/0] via 100.65.100.48, wan1, 00:34:27 [20/0] via 100.64.100.48, wan2, 00:34:27 B 10.18.240.0/24 [20/0] via 100.65.100.48, wan1, 00:34:27 [20/0] via 100.64.100.48, wan2, 00:34:27 B 10.18.241.0/24 [20/0] via 100.65.100.48, wan1, 00:34:27 [20/0] via 100.64.100.48, wan2, 00:34:27 B 10.18.243.96/27 [20/0] via 100.65.100.48, wan1, 00:34:27 [20/0] via 100.64.100.48, wan2, 00:34:27 This is the expected behavior since they are ECMP route with 2 paths. A change of one path will make the whole route update, but there will not be any traffic disturbance on the other link.