STP

[hs08]

Hello,

If i have topology like the picture where on SW2 there are direct connection from g0/4 to g0/5 in same switch.

G0/4 and G0/5 on SW 2 is on access vlan 10 mode, and i have some question :

1. Is there l2 loop on SW2? If yes then below question is relevant.
2. The L2 loop will occur for entire SW2 or only occur to any port which access to vlan 10 in SW1?
3. The L2 Loop also will have effect to entire SW1 or only any port which access to vlan 10 in SW1 or SW1 will not impacted?

[paul driver]

Hello
You will incur no loop as both ports will be in administrative mode of access.
These ports will be classed as edge ports so spanning-tree on these ports would not expect to receive BPDUs unlike ports that are as set as trunks that would expect BPDUs such as when connected as other switches.

[hs08]

Hi @paul driver 

With same question, how if the g0/4 and g0/5 in trunk port?

[MHM Cisco World]

NOTE:- below for both Access and Trunk Ports
1- these port you assume you use for host not to connect to other SW

If you not config bpduguard one this port then L2 Loop (temporally) will happened

2- these port assume to connect to other SW' 

Here STP must block one port (in most cases) 
l2 Loop (temporally) will happened 

above temporally what if L2 loop permanent (in case you mis-config bpdu filter i.e. disable STP)
L2 Loop permanent this will effect entire SW!! Yes entire SW' the cpu must handle all multicast and broadcast and if there is l2 loop then the cpu will full of process of handling multicast and broadcast strom' this make cpu high utilize and in end the SW reboot.

MHM

[Joseph W. Doherty]

---

@hs08 wrote:

With same question, how if the g0/4 and g0/5 in trunk port?

---

Basically pretty much the same results, although looping would also be flooding any of the VLANs on the trunk.  I.e. impact likely "magnified" if more ports involved.

Understand the problem, on a switch, with a loop, is unknown destination unicast, broadcast, and (non-IGMP suppressed) multicast can continue to circulate "forever".  Besides likely consuming much if not all bandwidth, recirculated broadcast packets causes receiving NICs to examine them to determine if they are of interest.  So all the hosts in the looping L2 domain are usually much impacted too.

[Joseph W. Doherty]

---

@paul driver wrote:

You will incur no loop as both ports will be in administrative mode of access.
These ports will be classed as edge ports so spanning-tree on these ports would not expect to receive BPDUs unlike ports that are as set as trunks that would expect BPDUs such as when connected as other switches.

---

@paul driver , (specifically Cisco's STP) eh/huh?

[Ruben Cocheno]

@hs08 

Loop right there, but STP blocks it

[Giuseppe Larosa]

Hello @hs08 ,

STP uses the port ID as a discriminator, the port ID is formed by port priority and a port identifier that is actually the SNMP ifindex.

port gi0/4 having a lesser SNMP ifindex will be preferred and left in forwarding state as Designated port on the link

port gi0/5 will go to STP blocking state.

if the ports are STP edge ports and you enable STP BPDU filter then you can build  a loop because you are preventing the ports from sending or processing STP BPDUs.

if the ports are STP edge ports ( port fast in Cisco terms) and you have enabled BPDU guard one of the two ports will go to errordisabled state.

Use of STP edge ports with BPDU guard is recommended

use of STP edge ports with BPDU filter should be limited to specific scenarios.

Edit:

>> With same question, how if the g0/4 and g0/5 in trunk port?

Again the STP mode PVST or Rapid PVST or MST will cause a comparison of sent and received BPDUs for each STP instance but the end result will be gi0/4 winner as designated port in each STP instance and gi0/5 in blocking mode for gi0/4 lower Port ID.

Hope to help

Giuseppe

[Joseph W. Doherty]

#1 Yes.

#2 & 3 Loop will only be in VLAN 10, but switch impact not limited to just VLAN 10, the switch's control plane often impacted.

Regarding other switches, not directly involved in loop, they might only be impacted traffic being flooded to them, which may also have a very negative impact.

[Joseph W. Doherty]

BTW, my understanding is, on ports enabled with port fast, Cisco still has STP running, so it's possible when STP "sees" BPDUs looping, it will block a port to break the loop, but this also assumes the switch's control plane is NOT quickly so overwhelmed, where it cannot perform this STP function.