Why native vlans are bad design
When using a mixed vendor environment, the use of native vlans is a recipe for disaster. Use of this in a multiple uplink scenario can lead to loops and total loss of service to all switches.
Spanning Tree (802.1d) and Rapid Spanning tree (802.1w) are industry standard protocols which prevent loops on a layer2 basis. In the beginning, Cisco was ahead of everyone when it came to inventing protocols that would later go on to inspire an industry standards. Examples are HSRP (VRRP), Uplinkfast (Alternate ports in Rapid Spanning Tree), and ISL (802.1q).
Not all protocols were migrated to industry standard and Cisco still used proprietary extensions for defaults. Ciscos Per Vlan Spanning Tree (PVST) is such a protocol (Some vendors have paid for the rights to use it). Standard PVST or Rapid PVST run a spanning tree on each vlan. This allows rudimentary load balancing and traffic engineering on layer2 links.
When you introduce native vlans into a multi vendor enviornment, the 3rd party switch will expect to run 802.1 D or W only on the native vlan. The cisco will expect this to run on every vlan on that trunk - and will correctly form a STP relationship on the native vlan only! On the cisco, it will not hear any BPDU's (Because the 3rd party switch isn't sending them) and will unblock those vlans.
If there is any sort of secondary layer2 path between those two switches which involve multiple vlans, you will cause a loop, even if the secondary port isnt running a native vlan. Here is an example.
Both Cisco Switch SW1, Dell switch SW2 and SW3 have vlans 10-20 defined and are allowed on all trunks.
Cisco Trunk ports are configured as such:
Interface Range gi1/1 - 2 (This means the same config is applied to ports 1 and 2 of card/switch #1)
Description Link to Dell Uplink port
switchport
switchport trunk encapsulation dot1q
switchport mode trunk
switchport trunk native vlan 10
On the Dell(s) (Pardon if the syntax is wrong, its been years)
Interface g1
Description To Cisco Switch
switchport trunk allowed vlan 10-20
switchport mode trunk
Interface g2
description to Other Dell Switch
switchport trunk allowed vlan 10-20
switchport mode trunk
The Cisco will see a STP relationship on Vlan 10 and act normally. However, it is going to set vlan 11-20 to FWD on all ports. The Dells will act as if Vlan 10's spanning-tree is the global spanning tree (Vlan 1), and select to block its ports that way. If spanning-tree calculated that the Cisco is to block one of its ports, It will not as the Cisco has no idea that it is supposed to block on vlan 11-20... Causing the dressed layer-2 loop.
I learned this lesson The Hard Way (c) while integrating a on board switch of an IBM Blade center into a Cisco environment. the management interface had to be on the native vlan, and use of vlan 1 anywhere is unadvised.
The only use of a native vlan I have ever seen was on old Cisco 3524xl POE switches which could not do an access-port with a voice vlan. In this case you had to use a native vlan for PC connectivity.
Servers and workstations belong on (generally) access ports, switches on trunks. Keep things simple, and please stay away from native vlans.
