Packet Path/

Native VLAN Mismatch Explained: When Untagged Traffic Lands in the Wrong VLAN

On an 802.1Q trunk, every VLAN is labelled with a tag — except one. The native VLAN crosses untagged, and the receiving switch files any untagged frame into its own native VLAN. If the two ends disagree about which VLAN is native, untagged traffic leaving one switch in VLAN 501 arrives at the other and lands in VLAN 1: two separate networks silently leak into each other, CDP logs %CDP-4-NATIVE_VLAN_MISMATCH, and spanning tree may block the port to stop it. This guide explains the leak frame by frame, the difference between the CDP warning and the PVST+ block, and the safe order for fixing it.

Switching · Intermediate · 11 min · November 20, 2026

Editorial illustration on a dark teal background: two network switches connected by a cable, teal packets travelling along it, while an orange packet collides with a purple shape at the right-hand switch with orange sparks.

The short answer

What is a native VLAN mismatch? A trunk is a link that carries many VLANs — separate virtual networks — between switches. To keep them separate, each frame crossing the trunk normally carries a small label, called an 802.1Q tag, saying which VLAN it belongs to. One VLAN is the exception: the native VLAN crosses with no tag at all. When a switch receives an untagged frame on a trunk, it has no label to read, so it files the frame into its own native VLAN. If switch A thinks the native VLAN is 501 and switch B thinks it is 1, an untagged frame that left A in VLAN 501 is filed by B into VLAN 1. Nothing fails loudly. Two networks that were supposed to be separate are now leaking into each other, one untagged frame at a time.

Picture two office buildings connected by a shared corridor, with labelled courier pouches moving between them. Every pouch wears a department badge — except the house pouch, which travels badgeless by agreement, and each building's front desk files any badgeless pouch into its own house department. That works only while both desks agree which department is the house department. Change it in one building and not the other, and every badgeless pouch is delivered to the wrong department — confidentially, repeatedly, and with no error on the pouch itself. A native VLAN mismatch is that disagreement, on a trunk.

In plain EnglishVLAN: a virtual network inside the physical one — devices in different VLANs cannot hear each other's broadcasts, as if they were on separate switches. Trunk: a link configured to carry many VLANs at once, usually between two switches. 802.1Q tag: a small label added to a frame on a trunk saying which VLAN it belongs to. Native VLAN: the one VLAN whose frames cross the trunk with no tag; untagged frames arriving are placed in it. PVID: port VLAN ID — the VLAN a port files untagged traffic into, which on a trunk is the native VLAN. CDP: Cisco Discovery Protocol, the neighbour-discovery protocol that first notices and logs this mismatch.

Why does untagged traffic exist at all?

The untagged exception is older than the tagging standard's dominance. When 802.1Q trunking was introduced, a switch needed a rule for frames that arrived with no tag — from older equipment, or from devices that do not tag — and the rule became: those frames belong to the port's native VLAN. The sending side mirrors it: frames in the native VLAN are sent across the trunk untagged, and frames in every other allowed VLAN are sent tagged. On Cisco switches the native VLAN is VLAN 1 by default, and you change it per trunk port with switchport trunk native vlan <id>. Cisco's own interface documentation states the rule plainly: native VLAN IDs must match on both ends of the trunk. The whole mechanism depends on that one agreement, because the frame itself carries no information to correct a disagreement — untagged means untagged, and only the receiving switch's configuration decides where it lands.

The leak, frame by frame

Take the exact case in the site's decoder entry: %CDP-4-NATIVE_VLAN_MISMATCH: Native VLAN mismatch discovered on GigabitEthernet1/0/49 (501), with Switch2 GigabitEthernet1/0/49 (1). Read it left to right: the local port is GigabitEthernet1/0/49, its native VLAN is the number in the first brackets — 501 — and the neighbour, Switch2, on its GigabitEthernet1/0/49, uses 1. Now follow one frame. A device in VLAN 501 sends a broadcast — an ARP request, a DHCP discover, any of the chatty traffic every VLAN generates. The first switch forwards it out of the trunk. Because VLAN 501 is that switch's native VLAN, the frame crosses untagged. Switch2 receives a frame with no label and applies the only rule it has: untagged means mine — VLAN 1. The broadcast is now delivered to devices in VLAN 1, and replies travel the same broken path in reverse. Tagged traffic is completely unaffected: a frame in VLAN 20 crosses with a tag saying 20 and lands in VLAN 20 on both sides, mismatch or not. That is the fingerprint of this fault — exactly two VLANs bleeding into each other across exactly one trunk, everything else normal, and the symptoms (strange DHCP answers, devices seeing neighbours from another network, one-way oddities) all confined to the two native VLANs.

No tag on the frame — only the receiving switch decides the VLANSwitch Anative VLAN 501Switch Bnative VLAN 1untagged frame from VLAN 501arrives — filed into VLAN 1tagged frame: VLAN 20 label intactlands in VLAN 20 on both sidesOnly the untagged path leaks — every tagged VLAN crosses correctly

The trunk is not broken and the frame is not corrupted. Each switch follows its own rule correctly — the rules simply disagree, and an untagged frame has no label with which to argue.

Two alarms, two different severities: CDP warns, PVST+ can block

CDP is the messenger. Switches running CDP exchange neighbour information across the link, including the native VLAN each side is using. When the two values differ, CDP logs %CDP-4-NATIVE_VLAN_MISMATCH with both numbers, both devices and both ports in a single line — one of the most self-diagnosing messages in the whole catalogue, and the one decoded in the site's entry for %CDP-4-NATIVE_VLAN_MISMATCH. But understand what CDP does not do: it changes nothing. Forwarding continues, the leak continues, and if nobody reads the log, the only symptom is the strange behaviour in the two affected VLANs. A network with CDP disabled gets no message at all — the leak is still there, just unannounced.

Spanning tree is the guard that can act. Cisco's PVST+ — per-VLAN spanning tree — puts the VLAN a BPDU (spanning tree's own control frame) belongs to inside the BPDU itself. Native VLAN BPDUs are sent untagged, so in a mismatch the receiving switch gets a BPDU that was originated in one VLAN but arrives, untagged, in a different one. Cisco's PVID troubleshooting document defines exactly that as a PVID inconsistency: a PVST+ BPDU received on a different VLAN than it was originated in. The switch responds by putting the port into an inconsistent state for the affected VLANs — in the document's example, when a BPDU originated in VLAN 10 is received in VLAN 1, both VLANs are blocked on that port. That block is deliberate protection, not a second fault: two VLANs bridged together can create the forwarding loops spanning tree exists to prevent. When the port stops receiving inconsistent BPDUs, the state clears and the switch logs that port consistency is restored. So the same mismatch can present two ways — a log line you could miss, or a blocked VLAN you cannot — depending on whether PVST+ catches the bridged BPDUs first. Check for the block explicitly with show spanning-tree inconsistentports: it lists every port spanning tree is currently holding in an inconsistent state.

How a mismatch happens: three causes, all half-finished changes

The decoder lists the causes engineers actually find, and all three are the same shape — one end changed, the other did not. Someone changed the native VLAN on one switch during a migration and not the other. A new switch was connected with default settings — native VLAN 1 out of the box — to a trunk configured for something else. Or the mismatch is part of a staged security change, moving the native VLAN off VLAN 1, that was only half finished. Nobody configures a mismatch on purpose; it is what a two-ended change looks like when only one end got done, which is also why the fix and the prevention below are both about treating a trunk as one object with two ends, never as two separate ports.

The fix, in order — align both ends in the same window

Step 1 — read both ends before touching either. Run show interfaces trunk on both switches and compare the native VLAN column for this trunk — do not trust the log line alone to tell you which side is wrong; the log tells you they differ, your network standard tells you which value is intended. In the same pass, run show spanning-tree inconsistentports on both switches so you know whether you are fixing a quiet leak or a leak plus an active block.

Step 2 — set the intended native VLAN identically on both ends. On the side that is wrong: switchport trunk native vlan <id> under the trunk interface, with the ID from step 1. Do both ends in the same change window and expect the obvious: between the first change and the second, the trunk is mismatched by definition, so this is a planned, brief window — not a change to make on one switch today and the other next week, which is how the fault was created in the first place. Production note: changing the native VLAN moves where untagged traffic lands the moment you type it, so if any real traffic rides the current native VLAN, schedule and announce it like the small traffic move it is.

Step 3 — verify the agreement, not just the command. show interfaces trunk on both switches now shows the same native VLAN; the CDP mismatch message stops appearing in the log; and show spanning-tree inconsistentports comes back empty — if PVST+ had blocked the port, it unblocks itself once inconsistent BPDUs stop arriving, and logs that consistency is restored. Then sanity-check the human symptom that brought you here: a device in the formerly leaking VLAN should now see only its own VLAN's broadcasts and DHCP servers.

Prevention: make the native VLAN a VLAN nobody uses

Alignment fixes today's trunk; the standing design removes the whole class of accident. Move the native VLAN off VLAN 1 to a dedicated, unused VLAN — one with no access ports assigned, no user devices and no management address living in it — and use that same unused VLAN as the native VLAN on every trunk. Then even if a mismatch does happen somewhere, the traffic that leaks is control traffic into an empty room, not user traffic into another user network, and the CDP message is a tidy-up instead of an incident. It also removes a security foothold: the classic double-tagging VLAN-hopping trick starts with an attacker sitting in the native VLAN, and a native VLAN with no user ports in it gives that trick nowhere to start. Optionally, tag the native VLAN as well with the global vlan dot1q tag native command — the decoder's own stronger fix — so no frame crosses untagged at all and there is nothing for a disagreement to re-file; if you use it, use it at both ends of every trunk, because a trunk where one side tags the native VLAN and the other expects it untagged is just the next mismatch wearing different clothes. Either way, the audit habit is cheap: show interfaces trunk across your switches, native column compared, belongs on the same checklist as the allowed-VLAN comparison in VLANs, Trunks, and Spanning Tree — and it pairs with the desk-edge guards covered in BPDU Guard and PortFast Explained and Port Security Explained, which watch the access edge while this one watches the links between switches.

Try it yourself

Catch the disagreement in configuration before CDP catches it in the log with the free Network Config Verifier — paste both switches' configurations and compare their trunk settings side by side — and see which switches a trunk actually joins with the free Config-to-Topology Mapper, so every trunk on the map gets its two ends checked as a pair. And when the log line is what you have, open the decoder entry for %CDP-4-NATIVE_VLAN_MISMATCH and work its checks in order — read the native VLAN on each end, check whether spanning tree has already blocked a port, then align both ends and verify the message stops.

Fact-check sources: the native VLAN carrying untagged traffic on a trunk, untagged traffic being assumed to belong to that VLAN, the rule that native VLAN IDs must match on both ends of the trunk, and the default native VLAN being the default VLAN when none is configured — Cisco Nexus 9000 Series NX-OS Interfaces Configuration Guide, Release 9.2(x), Configuring Layer 2 Interfaces (native VLAN section), via Cisco's published guide. The definition of a PVID inconsistency (a PVST+ BPDU received on a different VLAN than it was originated in), native VLAN BPDUs being sent untagged, both the receiving and originating VLANs being blocked on the port when an inconsistency is detected, the inconsistent state clearing when inconsistent BPDUs stop arriving (with a port consistency restored message), and show spanning-tree inconsistentports as the command that lists inconsistent ports — Cisco, Troubleshoot Spanning Tree PVID- and Type-Inconsistencies, opened and read in full this run. The %CDP-4-NATIVE_VLAN_MISMATCH sample message (native VLAN 501 on GigabitEthernet1/0/49 against Switch2's 1 on the same port), its meaning (untagged traffic leaving in one VLAN lands in the other, silently leaking the two networks together), its three causes (one-ended migration change, new switch on defaults, half-finished move off VLAN 1), its two checks (show interfaces trunk; show spanning-tree inconsistentports) and its fix (set switchport trunk native vlan identically on both ends; better still, tag the native VLAN with vlan dot1q tag native or move it to an unused VLAN with no devices in it) are quoted from the site's own decoder entry, sourced from Cisco 802.1Q trunking documentation. The per-interface command switchport trunk native vlan <id> and VLAN 1 as the Cisco default native VLAN are the standard IOS trunk configuration described in that same trunking documentation.

Related guides

Key takeaways

Keep reading