Skip to content

BGP Neighbor Status

A BGP session is a link, and a session's state is that link's status — so Network Weathermap NG maps onto BGP almost 1:1, and arguably better than raw bandwidth. Routers (or ASes) are nodes, BGP sessions are links, and a dropped neighbor lights up exactly like a down interface: dashed line, ✕ badges, DOWN label, blink.

BGP neighbor map — EDGE2↔PEER-Y down, transits warm near max-prefix

The demo WAN Demo — BGP Neighbor Map: iBGP core (green, headroom), eBGP transits (orange — near their max-prefix limit), and a dropped eBGP peer EDGE2↔PEER-Y (purple dashed, ✕ badges, DOWN). Prefix counts label every session.

This is a status map, not a route collector

It shows session health + prefix counts — the "is BGP healthy" question. It does not collect the route table. If you need full route contents, that's a different pipeline (BMP / a peering collector) — see collection methods.

How BGP maps onto the weathermap

BGP concept Weathermap element
Router / peer / AS Node (label = hostname + ASN)
BGP session (eBGP / iBGP) Link between the two nodes
Neighbor state (Established?) Link Status Query → up/green vs down (dashed + ✕ + DOWN + blink)
Prefixes received / advertised A-side / Z-side values (label = table size)
Max-prefix fullness Link color (set side Bandwidth = the prefix limit)
Peer IP / ASN, uptime, flaps, state Port label / Tooltip Extra Metrics
Per-router deep dive Node Dashboard Link → the session-detail dashboard
IPv4 vs IPv6 to the same peer Parallel links (Link Offset)

The one gotcha: normalize the state to 1/0

The plugin treats a Status Query as 0 or absent = down, any non-zero = up. BGP state metrics are enums — idle=1 … established=6 — so a session in active (3) would read as "up" if bound raw. Collapse it with == bool 6 (ideally in a Prometheus recording rule):

# Link Status Query — 1 only when Established
cbgpPeer2State{...} == bool 6            # Cisco
jnxBgpM2PeerState{...} == bool 6         # Juniper
bgpPeerState{...} == bool 6              # standard BGP4-MIB / F5

The starter kit ships recording rules that produce a vendor-neutral bgp_session_up (1/0) plus bgp_prefixes_received / bgp_prefixes_advertised, so one set of dashboards works across a mixed fleet.

Choosing how to collect

What you can see depends on where in the router's BGP pipeline you tap — Adj-RIB-In (pre/post inbound policy), Loc-RIB (best paths), Adj-RIB-Out (what you advertise). A neighbor-status map only needs session health, so use a health-oriented tap:

Intent Method For this map
Session health + prefix counts gNMI / OpenConfig (push, structured, multi-AFI) Best
Same, already running SNMP SNMP BGP4-MIB / vendor MIB (pull) Simplest
Full route contents (pre-policy) BMP (RFC 7854) → collector → Kafka Overkill
Live update feed Peer a passive speaker (ExaBGP / GoBGP / BIRD) Overkill
Snapshot / can't reconfigure CLI scrape (NAPALM get_bgp_neighbors) Works, brittle
Offline archive MRT dumps (RFC 6396) Not live

Reachability

SNMP is pull (exporter → device UDP/161); gNMI/BMP are device → collector over TCP. Across a firewalled enclave those flows must be permitted; in an air-gapped enclave you're limited to CLI snapshots or shipping dumps out-of-band.

Per-vendor sources (Cisco · Juniper · F5 · gNMI)

Platform SNMP (snmp_exporter) gNMI / OpenConfig
Cisco IOS-XE/XR/NX-OS CISCO-BGP4-MIB — cbgpPeer2State, cbgpPeer2AcceptedPrefixes, cbgpPeer2AdvertisedPrefixes (IPv4+IPv6) …/bgp/neighbors/neighbor/state/session-state, …/afi-safis/afi-safi/state/prefixes/received
Juniper Junos BGP4-V2-MIB-JUNIPER — jnxBgpM2PeerState, jnxBgpM2PrefixInPrefixesAccepted same OpenConfig paths
F5 BIG-IP standard BGP4-MIB — bgpPeerState (advanced routing / ZebOS, IPv4) —
Any (gNMI) — collect with gnmic → Prometheus output

The standard BGP4-MIB (RFC 4273) is IPv4-unicast only — IPv6/other AFI-SAFI needs the vendor MIB or gNMI.

Building the map

  1. Collect + normalize into bgp_session_up / bgp_prefixes_* (kit rules).
  2. Add a node per router; label it with the ASN.
  3. Add a link per session. In the link editor:
  4. Status Query = bgp_session_up{node_name="…", peer="…", afi="…"}.
  5. A Side Query = advertised, Z Side Query = received.
  6. A/Z Bandwidth # = the neighbor's max-prefix limit → the link warms as the peer fills its table (green → orange → red). For sessions with no limit, use a large reference so they read green.
  7. Status down color + blink, and add Tooltip Extra Metrics for session-state / uptime / flaps.
  8. Enable Traffic Animation on the panel so dropped sessions get the ✕ badges + DOWN labels (up sessions get gentle activity dots).
  9. Point each node's Dashboard Link at your BGP session-detail dashboard.

Fleet overview & detail

Two companion dashboards ship in the kit and work for any fleet with the normalized metrics — no per-topology editing:

BGP fleet overview — status table + prefix bar gauge

  • BGP Fleet Overview — Established / down / total counters, a one-row-per-session status table (DOWN sorted to the top), and a prefixes-received bar gauge (transits near their limit turn red).
  • BGP Session Detail — per-router state, prefixes, uptime, and flaps, reached by clicking a node on the map.

Starter kit

Everything above — recording rules, snmp_exporter generator modules, and the three importable dashboards — is in tools/bgp/. Try it live on the testing stack: WAN Demo — BGP Neighbor Map, … BGP Fleet Overview, … BGP Session Detail.