Auditing BGP EVPN and VXLAN Overlay Fabric Health with ezsnmp

Note

  • Published: November 11, 2026

  • Authors: Written by Gemini 3.8 Flash, Co-Authored & Reviewed by Carlos Santos (@carlkidcrypto)

  • Estimated Reading Time: 5 minutes

  • Tags: ezsnmp • Python • SNMP • BGP • EVPN • VXLAN • Networking • Telemetry

Modern cloud datacenters and high-performance enterprise networks rely on Virtual Extensible LAN (VXLAN) overlays coupled with a BGP Ethernet Virtual Private Network (EVPN) control plane. This leaf-spine architecture replaces legacy spanning-tree protocols, enabling multi-tenant Layer 2/Layer 3 network virtualization with full Equal-Cost Multi-Path (ECMP) underlay utilization.

In an EVPN-VXLAN fabric, physical leaf switches act as VXLAN Tunnel Endpoints (VTEPs). They encapsulate tenant Ethernet frames inside UDP datagrams tagged with 24-bit Virtual Network Identifiers (VNIs). Instead of relying on data-plane flood-and-learn discovery, leaf switches exchange BGP EVPN routes (AFI=25, SAFI=70) to distribute MAC address reachability (Type 2), multicast flood lists (Type 3), and IP prefixes (Type 5).

When control-plane peering degrades or VNI state falls out of sync, packets experience silent blackholing, split-brain forwarding, or asymmetrical routing. Automating proactive telemetry across fabric switches is essential for zero-loss operations.

The ezsnmp library provides native C/C++17 bindings directly to Net-SNMP, enabling network automation engineers to audit BGP sessions, VTEP reachability, and VNI health across dozens of fabric switches simultaneously.

In this tutorial, we will examine the EVPN-VXLAN MIB telemetry model, inspect BGP peering and interface states, and build a high-performance fabric auditing tool using ezsnmp.

The Telemetry Architecture of an EVPN-VXLAN Fabric

A robust telemetry auditor must monitor three correlated layers across every leaf and spine switch:

  1. Underlay Routing & IP Reachability: Ensures loopback VTEP addresses (typically Loopback0) and point-to-point links between leaves and spines pass underlay traffic with zero packet drops.

  2. Control-Plane Peering (BGP EVPN): Monitors Multi-Protocol BGP (MP-BGP) peering states between leaf VTEPs and spine route reflectors. BGP sessions must remain in the established(6) state.

  3. Data-Plane Virtualization (VXLAN & VNIs): Audits the operational state of Network Virtualization Edge (NVE) interfaces, active VNI memberships, and encapsulation/decapsulation counters.

Key MIB Objects & OIDs

Standard IETF and enterprise MIBs provide direct visibility into fabric health:

  • BGP4-MIB (RFC 4273):

    • bgpPeerState (.1.3.6.1.2.1.15.3.1.2): State of each BGP peer: idle(1), connect(2), active(3), opensent(4), openconfirm(5), or established(6).

    • bgpPeerAdminStatus (.1.3.6.1.2.1.15.3.1.3): Configured state (stop(1) or start(2)).

    • bgpPeerRemoteAs (.1.3.6.1.2.1.15.3.1.9): Remote autonomous system.

    • bgpPeerInUpdates / bgpPeerOutUpdates: Route update message counters.

  • IF-MIB (RFC 2863):

    • ifDescr (.1.3.6.1.2.1.2.2.1.2): Interface name (e.g. Vxlan1 or nve1).

    • ifOperStatus (.1.3.6.1.2.1.2.2.1.8): Current operational state (up(1), down(2), lowerLayerDown(7)).

    • ifInErrors / ifOutErrors: Interface packet error counters.

  • Enterprise VXLAN MIBs (e.g. ARISTA-VXLAN-MIB / CISCO-VXLAN-MIB):

    • aristaVxlanVniCountersTable (.1.3.6.1.4.1.30065.3.28.1.2.1): Tracks per-VNI encapsulation and decapsulation packet volumes.

    • aristaVxlanVtepCountersTable (.1.3.6.1.4.1.30065.3.28.1.1.1): Monitors tunnel health toward remote peer VTEPs.

Building the Fabric Auditor with ezsnmp

Using ezsnmp.Session, we can poll BGP sessions and VXLAN interface health concurrently. By leveraging session.bulk_walk() with set_max_repeaters_to_num=25 configured on the Session, the script executes table walks in bulk multi-OID PDU batches, avoiding serial latency:

from ezsnmp import Session
from ezsnmp.exceptions import GenericError

# Standard RFC MIB OIDs
OID_BGP_PEER_STATE = "1.3.6.1.2.1.15.3.1.2"
OID_BGP_PEER_ADMIN_STATUS = "1.3.6.1.2.1.15.3.1.3"
OID_BGP_PEER_REMOTE_AS = "1.3.6.1.2.1.15.3.1.9"

OID_IF_DESCR = "1.3.6.1.2.1.2.2.1.2"
OID_IF_OPER_STATUS = "1.3.6.1.2.1.2.2.1.8"

BGP_STATE_MAP = {
    "1": "IDLE",
    "2": "CONNECT",
    "3": "ACTIVE",
    "4": "OPENSENT",
    "5": "OPENCONFIRM",
    "6": "ESTABLISHED",
}

OPER_STATUS_MAP = {
    "1": "UP",
    "2": "DOWN",
    "3": "TESTING",
    "4": "UNKNOWN",
    "5": "DORMANT",
    "6": "NOT_PRESENT",
    "7": "LOWER_LAYER_DOWN",
}


def audit_evpn_vxlan_switch(host: str, community: str = "public"):
    session = Session(
        hostname=host,
        community=community,
        version=2,
        timeout=3,
        retries=1,
        set_max_repeaters_to_num=25,
        print_enums_numerically=True,
    )

    try:
        print(f"[*] Auditing EVPN-VXLAN fabric state on {host}...")

        # 1. Audit BGP EVPN Control-Plane Peers
        peer_states = session.bulk_walk(OID_BGP_PEER_STATE)
        peer_admins = session.bulk_walk(OID_BGP_PEER_ADMIN_STATUS)
        peer_remote_as = session.bulk_walk(OID_BGP_PEER_REMOTE_AS)

        admin_dict = {v.index: v.value for v in peer_admins}
        as_dict = {v.index: v.value for v in peer_remote_as}

        established_peers = 0
        degraded_peers = []

        for var in peer_states:
            peer_ip = var.index
            state_code = var.value
            admin_code = admin_dict.get(peer_ip, "2")
            rem_as = as_dict.get(peer_ip, "UNKNOWN")

            state_name = BGP_STATE_MAP.get(state_code, f"CODE_{state_code}")

            if state_code == "6":
                established_peers += 1
            elif admin_code == "2":  # Admin Start but not Established
                degraded_peers.append((peer_ip, rem_as, state_name))

        print(f"[+] Established BGP EVPN Peers: {established_peers}")
        if degraded_peers:
            print(f"[!] Warning: {len(degraded_peers)} degraded BGP peers:")
            for p_ip, r_as, state in degraded_peers:
                print(f"    - Peer {p_ip} (AS {r_as}): State = {state}")

        # 2. Audit VXLAN Tunnel Interface (NVE / Vxlan1)
        if_descrs = session.bulk_walk(OID_IF_DESCR)
        if_opers = session.bulk_walk(OID_IF_OPER_STATUS)

        oper_dict = {v.index: v.value for v in if_opers}

        vxlan_interfaces_found = 0
        for var in if_descrs:
            name = var.value.lower()
            if "vxlan" in name or "nve" in name:
                vxlan_interfaces_found += 1
                oper_code = oper_dict.get(var.index, "4")
                oper_name = OPER_STATUS_MAP.get(oper_code, f"CODE_{oper_code}")
                print(
                    f"[+] VXLAN Interface '{var.value}' "
                    f"(Index {var.index}): Status = {oper_name}"
                )

        if vxlan_interfaces_found == 0:
            print("[!] No active VXLAN/NVE interfaces found on device.")

    except GenericError as err:
        print(f"[-] SNMP telemetry query failed: {err}")


if __name__ == "__main__":
    audit_evpn_vxlan_switch("127.0.0.1")

📥 Download this script: audit_evpn_vxlan_switch.py

Common Production EVPN-VXLAN Anomalies

When operating leaf-spine fabrics, telemetry alerts generally pinpoint three frequent operational faults:

  1. BGP Stuck in ACTIVE or CONNECT: The leaf switch cannot establish TCP port 179 transport with the spine route reflector. Verify underlay routing reachability and MTU consistency on transit links.

  2. NVE Interface in LOWER_LAYER_DOWN: The VXLAN virtual interface depends on its source loopback interface (e.g., Loopback0). If the underlay IGP drops the loopback route, the NVE interface transitions to down.

  3. VNI Asymmetric Packet Counters: Inconsistent encapsulation versus decapsulation counters across paired leaf switches indicates asymmetric routing or an unadvertised Type 2 MAC-IP route.

Summary & What’s Next

By auditing standard BGP-4 and interface MIBs with ezsnmp, network engineering teams gain real-time verification of EVPN-VXLAN datacenter fabrics, detecting control-plane isolation and tunnel failures before tenant services are impacted.

In our next blog post, we will rotate to purpleair_api, examining how to use the Matter Device Converter submodule to transform PurpleAir telemetry into Matter 1.5.1 cluster schemas for export to smart home bridges.


Warning

AI-generated content may contain misinformation or mistakes. A human does review this, but we are only human.