Aller au contenu
Réseau multi-cloud

Routes are not being advertised or learned

For a site that is ONLINE — if it is not, start with registration.

  1. Is peering up at all? show-ip-bgp-summary. Read State/PfxRcd: a number means established, and Up/Down shows whether it has been stable. A small value on a site that has been up for days means the session is flapping.

  2. If the session is down, the cause is usually below BGP. Confirm the interface carrying the peering is up with ip-link-show, then confirm reachability to the peer address.

    • On Azure: Peers are Azure Route Server instances (az network routeserver peering list-learned-routes).
    • On AWS: Peers are the AWS VPC Router / gateway interface (10.150.0.1).
    • On-Premise KVM: Peers are the containerized FRR ToR BGP router (docker exec frr-router vtysh -c "show ip bgp summary").

  3. If the session is up but the VIP is unreachable, check what the node is actually advertising with show-ip-bgp-neighbors-advertised-route. If the VIP is absent there, the problem is on this node and looking at the peer will not help.

  4. If the node advertises correctly but traffic still does not arrive, the control plane and the data plane disagree. Check the forwarding table the data plane actually uses with rt, remembering that its VRF id is not the one BGP reports.

    Then resolve the next hop with nh. A prefix present in rt whose next hop is Drop is a blackhole.

  5. If the session flaps, show-ip-bgp-neighbors carries the last reset reason and the negotiated timers.