Does loopgaurd always put a non-designated port in loop-inconsistent

[bluesteel]

Hi,

I have a question with regards to loopgaurd. If you have a block of 6 access switches daisy chained together and you uplink switches 1 and 6 back to the distribution layer. This means spanning tree blocks a port in the centre of the block so the top half uses switch 1's uplink to the distribution layer and the bottom half of the block will use switch 6s uplink. If an uplink fails spanning tree should unblock the blocked port in the centre of the block so all switches in the block can use the active link back the distribution switch.

My question is would loopgaurd be counter productive in this topology as it is reliant on the port being unblocked with the loss of BPDU's. If loopgaurd is active will this not cause the port to remain blocked essentialy issolating half of the block from the rest of the network, if the uplink of the side of the blocked port fails.

Kind Regards

Daniel

[Yogesh Ramdoss]

Daniel,,

In PVST+ or Rapid-PVST+, the loopguard helps to stabilize the network, and not becomes counter-productive.

From your example:

ROOT---SW1--SW2--SW3--SW4--SW5--SW6---ROOT

(here, ROOT is the same switch).

If the link between ROOT and SW1 goes down, then the root-port is going down, and STP will reconverge, by unblocking the link between Sw3 and Sw4, without any issue.

If the root-port stays up and SW1 stop receiving BPDUs from ROOT on that port but still getting BPDUs from SAME root device but on another port (connected to SW2) then it indicates the ROOT device is up and running but something went wrong between ROOT and SW1 e.g., UniDirectional link). In this scenario, the uplink port on SW1 goes loop-inconsistent - because, on loopguard enabled root port, it CANNOT go from ROOT-Port to Designated. In summary, if link stays up and BPDU stop, loopguard kicks in. If the link completely fails, STP will reconverge with no impact.

Hope this helps.

- Yogesh

[bluesteel]

Hi Yogesh,

Thanks, so that means it would be counter productive as STP will only reconverge from a direct link failure. If one of the switch uplinks in the chain fail you would need the blocked port to become DP as all the swiches along the chain towards the failed link transition their DP to RP and RP to DP. Loopgaurd would not allow the port to pass traffic from an indirect loss of BPDUs.

Kind Regards

Daniel

[Yogesh Ramdoss]

Daniel,

STP Loopguard should NOT be counter productive, even in the case link failure (either directly connected or not).

Assume the same example:

ROOT---SW1---SW2---SW3---SW4---SW5---SW6---ROOT

ROOT is the same device, and SW3 has a blocked port (connected to SW4)

If the link between ROOT and SW1 goes down, then SW3 stops getting BPDU from SW2 (which in turn gets from SW1 and ROOT). Thats when the STP Proposal/Agreement comes into the picture. SW3 moves its BLK port to FWD (after sending proposal to SW2, and SW1/2/3 has inferior STP parameters). Please review the Proposal/Agreement sequence for more info.

http://www.cisco.com/en/US/tech/tk389/tk621/technologies_white_paper09186a0080094cfa.shtml#topic5

Regards,

Yogesh

[bluesteel]

Hi Yogesh,

Yes it would work with RSTP, but my client is using PVST. Does PVST not conform to traditional STP? I am aware of how RSTP uses synchronisation to converge.

Kind Regards

Daniel

[Yogesh Ramdoss]

Daniel,

I setup a quick lab, so that it will be clear. I have the following topology:

SW1(fas1/1) ---(fas1/1)SW2(fas1/2)---(gig1/2)SW3

All links are trunk, and test vlan is 504. All running PVST.

SW1 is the primary ROOT for vlan 504 and SW3 is the secondary ROOT.

As a test, I removed vlan 504 from SW1 port 1/1 (simulating uni-directional link failure for vlan 504)

SW2 reports:

May 15 16:01:42:019: %SPANTREE-2-LOOPGUARD_BLOCK: Loop guard blocking port FastEthernet1/1 on VLAN0504.

SW3 reports:

May 15 16:01:42.102: %SPANTREE-2-LOOPGUARD_BLOCK: Loop guard blocking port GigabitEthernet1/2 on VLAN0504
May 15 16:01:57.103: %SPANTREE-2-LOOPGUARD_UNBLOCK: Loop guard unblocking port GigabitEthernet1/2 on VLAN0504

(15 seconds convergence time for legacy per-vlan STP)

STP status:

In SW2:  1/2 moved from DESG FWD to ROOT FWD, 1/1 moved to Desg BKN *LOOP_Inc

In SW3:  1/2 moved from ROOT FWD to DESG FWD

I added vlan 504 back in SW1 port 1/1:

SW2 reports:

May 15 16:05:44:387: %SPANTREE-2-LOOPGUARD_UNBLOCK: Loop guard unblocking port FastEthernet1/1 on VLAN0504.

All STP status restored to what it was earlier.

Hope this helps. I do not see STP loopguard as counter-productive.

Regards,

Yogesh

[bluesteel]

Hi Yogesh,

Thanks for the reply, So let me get this right. If any link fails along the chain on the blocked port side of the central switch then the port will initially be put into loop inconsistent? The switch with the direct failure will transition its DP to RP, same will happen for all switches leading up to the port which is in loopinconsistent. When the neighbour on that segment transitions its DP to RP the switch realises it needs to allow the blocked port to be DP and releases the port from loop inconsistent?.

This would make sense but the documentation states that a port in loop inconsistent can only be recovered once it recieves BPDUs again.

Regards

Daniel

[Yogesh Ramdoss]

Daniel,

>>  This would make sense but the documentation states that a port in  loop inconsistent can only be recovered once it recieves BPDUs again.

The port in Loop Inconsistent state will recover only when it start to receive the BPDU. In the example I discussed above, SW2 Port 1/1 was in Loop_Inc state till I re-add vlan back in SW1 port 1/1 (during 16:01:42 - 16:05:44). So, documentis right.

I am confused with first set of questions. I would request you to draw a sample topology and ask questions.

Yogesh