SONiC
Module 4 · EVPN/VXLAN in SONiC

EVPN Route Types (SONiC View)

35 minLesson

A fast, CCNP+-level review of the route types that matter in a SONiC leaf-leaf fabric — type-2, type-3, and type-5 — focused entirely on how they appear in FRR, what SONiC does with them, and the failure signatures each one produces. No introductory definitions.

Prerequisites: Lesson 4.1 · CCNP+ EVPN route-type knowledge

Learning objectives

  • Recognize type-2, type-3, and type-5 routes in the FRR EVPN RIB and read their key fields.
  • State what SONiC installs (FDB entry, flood list, route) in response to each type.
  • Map each route type to the failure signature you see when it is missing or unprogrammed.
Why this matters

You do not need the definitions; you need to know which route type to look for when a specific symptom appears — unicast between known hosts vs BUM vs inter-subnet routing each implicate a different type.

The FRR EVPN RIB is where you confirm the control plane; reading it fluently is half of overlay troubleshooting.

Analogy — Kinds of directory entries
In the analogyIn SONiC
A parcel-level entry: this exact item is at this terminalType-2 (MAC/IP)
A terminal's membership card for a tenant's broadcast listType-3 (inclusive multicast)
A whole ZIP code reachable via this terminalType-5 (IP prefix)
What it actually is: Type-2 carries a specific MAC (and optionally its IP) — it builds remote unicast forwarding. Type-3 announces that a VTEP participates in a VNI so BUM traffic can be replicated to the right set of peers. Type-5 carries routed IP prefixes for inter-subnet / L3VNI forwarding where the destination MAC is not (and need not be) known end to end.
Where the analogy stops working

A directory is passive lookup; EVPN routes actively *drive* hardware programming, and the programming can fail even when the 'entry' is perfect. The analogy also flattens the type-2 MAC-only vs MAC+IP distinction, which matters for ARP suppression and IRB.

The three route types you will actually chase

TypeCarriesSONiC installsMissing/unprogrammed signature
2 (MAC)Remote MAC + VNI + VTEPRemote FDB entry via tunnelUnicast to a specific host fails; flooding may still 'work'
2 (MAC+IP)Remote MAC + IP + VNIFDB + ARP/ND (suppression), IRB host routeARP not suppressed; IRB host reachability gaps
3 (IMET)VTEP membership in a VNIHead-end replication / flood listBUM (ARP, DHCP discover) never reaches peer VTEP
5 (IP prefix)Routed prefix + L3VNI + RTRoute + tunnel next hopInter-subnet/routed reachability fails; same-subnet may work
Symptom → suspect type: 'ARP never arrives' → type-3; 'known host unreachable' → type-2; 'other subnet unreachable' → type-5.

Reading the EVPN RIB in FRR

FRR / vtyshIllustrative output
# Inside the bgp container / vtysh
show bgp l2vpn evpn summary          # are EVPN sessions established?
show bgp l2vpn evpn                   # full EVPN table (all types)
show bgp l2vpn evpn route type macip  # type-2 MAC/IP routes
show bgp l2vpn evpn route type multicast   # type-3
show bgp l2vpn evpn route type prefix      # type-5
show evpn vni                        # locally known VNIs and their kind (L2/L3)
show evpn mac vni <vni>              # MACs learned in a given VNI

You confirm the control plane in vtysh, not in SONiC's show layer. These are the workhorse commands; field layout varies slightly by FRR version.

Illustrative type-2 entry (annotated)

FRR / vtyshIllustrative output
Route Distinguisher: 10.0.0.1:5
*> [2]:[0]:[48]:[00:11:22:33:44:55]:[32]:[10.10.10.11]
      10.0.0.1 (VTEP / next hop = remote loopback)
      Extended Community: RT:65001:10010 ET:8
      Label1 10010 (L2VNI)  Label2 50010 (L3VNI, if IRB)
# Read: MAC 00:11:22:33:44:55 (+host 10.10.10.11) lives behind VTEP 10.0.0.1,
# in L2VNI 10010. The far leaf should install a remote FDB entry to that VTEP.

What a MAC+IP route looks like and the fields that matter when correlating to hardware.

Present in the RIB ≠ programmed in the ASIC

A perfectly formed type-2 route can sit in the EVPN RIB while the corresponding remote FDB entry never appears in ASIC_DB — SAI error, resource exhaustion, or a tunnel that is administratively present but not up. Lesson 4.6 turns this into a layered check; for now, treat RIB presence as step one, never the finish line.

Route targets and import

Whether a route even enters a leaf's EVPN RIB depends on route-target import. An RT mismatch means the advertising side is perfect and the receiving side simply never imports the route — a common, quietly self-inflicted failure. When a route is 'missing' on the receiver but present on the sender, suspect RT/import before suspecting the network.

Troubleshooting implications
  • Match the symptom to the route type before opening any capture.
  • Check both sides: advertised on the origin, and imported (RT match) on the receiver.
  • Confirm RIB presence, then confirm hardware programming — they are separate checks.

Key takeaways

  • type-2 = remote MAC/IP unicast; type-3 = BUM membership; type-5 = routed prefixes.
  • Symptom class points at the route type; use vtysh to confirm the control plane.
  • RT/import mismatch is a top cause of 'route present on sender, missing on receiver'.
  • RIB presence never proves ASIC programming.

Tested against: SONiC 202305 / 202311 lineage · Upstream (community) unless noted · SONiC-VS (virtual) baseline · last reviewed 2026-07. Exact commands and behavior can vary by SONiC release, image, hardware platform, and enterprise distribution.

Knowledge check

Q1

Host A cannot resolve ARP for Host B in the same subnet across the fabric, even though once you statically prime the ARP entry, unicast works. Which EVPN route type is most likely at fault?