11 Replies Latest reply: Jun 14, 2018 8:59 AM by jh RSS

    STP load balancing

    isaac

      Hi Guys

       

      I came across this little section when reading the CISCO LAN switching book(a very good book). And I tried to Lab the below scenario with no luck. Now I ask if you could assist me in clearing my head with this one, I just want to know if it's even possible. To me I don't think SW-D can have both ports forwarding, doesn't make any sense. Diag on the right in focus

       

        • 1. Re: STP load balancing
          jh

          It is possible and required for C<-->D segment.

          • 2. Re: STP load balancing
            Steven Davidson

            The problem that you're going to run into with this is the PVST simulation.  The MST region (represented by Sw-D) only interacts with the non-MST region (Cat-A, Cat-B, and Cat-C) on VLAN 1.  According to the diagram, Cat-A is root bridge for VLAN 1.  Sw-D will listen to all PVST BPDUs arriving on the boundary ports and if Sw-D doesn't reach the same decision for ALL VLANs 2-4094 that it does for VLAN 1 it will trigger a simulation failure and put ports in a BRK state (broken).  I don't think it's possible, with the CIST root in the non-MST region, to operate multiple VLANs and balance them (say odd over Cat-B <-> Cat-A and even over Cat-C <-> Cat-A) with out tripping the PVST simulation failure on Sw-D.  The best bet is to set VLAN 1 to a priority of 8192 and all other VLANs to 4096 on Cat-A and let the blocking port sit on one of the MST region's (Sw-D) boundary ports.

             

            When you say you tried to lab this scenario with no luck what do you mean?  What happened when you tried this in a lab?

            • 3. Re: STP load balancing
              isaac

              Hi JH

              how can it be accomplished?

               

              Hi Steven

              I 1st used the switches with just VLAN 1 running and no MST configuration as I thought VLAN 1 STP instance would be treated the same (MAC 0180.c200.0000).

               

              I then used an Alcatel switch with just VLAN 1 running 1 instance of STP (original 802.1D) as SW-D

               

              I made CAT-A the root bridge for VLAN 1 and all other VLANs.

              I then continued to play around with the cost value as shown in the book and it did not work. but I was in doubt from the beginning. I thought if the root bridge is CAT-A, then from SW-D's the are 2 path to the root bridge, so how could you possibly have both forwarding(1 RP and the other DP). I don't know if my rational is misguided.

              • 4. Re: STP load balancing
                jh

                hmm... tbh i'm still not sure if the OP question is about PVST+ and MST interaction or basic spanning-tree port states? i jumped to the conclusion it was the latter (i continue to think that from your second port isaac) and originally ignored the left diagram completely "Diag on the right in focus", but Steven has probably read it correctly.

                • 5. Re: STP load balancing
                  Steven Davidson

                  If the Alcatel is allowing the tunneling of PVST BPDUs then you should be able to manipulate the costs such that Sw-D concludes that one port is root and that the other port is designated for VLAN 1.  In the reference topology you should be able to manipulate path costs to put any of the links into a blocking state on one side (which, by definition, would make the other side designated on p2p links).  If you're trying to get this to work in a lab using the costs indicated on the diagram then pay special attention to what the Alcatel might be doing to the cost of the BPDU.  On Cat-C issue this command:

                   

                  show span interface 2/1 detail (of course your port will be different)

                   

                  If Sw-D were a Cisco switch running MST then it would add 20000 to the cost of the VLAN 1 BPDU and simply manipulating the metrics to the values suggested in the diagram would do nothing to make the path across the top the better path to root for VLAN 1.

                  • 6. Re: STP load balancing
                    isaac

                    I'm starting to get the picture but still a little confused, cause SW-D(alcatel) is supposed to process VLAN 1 BPDUs. So how can it then have 2 paths to the root both forwarding?

                     

                    It doesn't look like the Alcatel is letting PVST BPDUs through.

                     

                    SW-D

                     

                    SW-C

                    SW-B

                     

                    • 7. Re: STP load balancing
                      Steven Davidson

                      isaac wrote:

                       

                      I'm starting to get the picture but still a little confused, cause SW-D(alcatel) is supposed to process VLAN 1 BPDUs. So how can it then have 2 paths to the root both forwarding?

                       

                       

                       

                      There's no prohibition for having more than one forwarding port on a switch.  You cannot have two ROOT ports on a switch but Sw-D doesn't have two root ports.  It has just the one RP pointing at Cat-B.  The other port is designated.  Consider that port to be downstream from the root bridge.  Because Cat-C is blocking the loop on Cat-C's side it technically is downstream from the root from a reachability standpoint.  Any frames sent by Sw-D toward Cat-C will never reach the root bridge.

                      • 8. Re: STP load balancing
                        isaac

                        Thanks

                        what got me really intrigued is how would changing cost on CAT-C 2/2 and 2/3 turn SW-D port from blk to forwarding especially for VLAN 1 since SW-D will process VLAN 1 STP BPDUs.

                        I understand from the diagram on the left how would manipulating cost cause 2/1 to block. but for the one on the right, I don't get it. 

                         

                        If you can Steven please break it down for me, I'm wrecking my head and I'm just stuck.

                        • 9. Re: STP load balancing
                          Steven Davidson

                          Sw-A sends a BPDU for VLAN1 toward Cat-B and Cat-C.  Cat-B forwards that toward Sw-D and Cat-C.  Because Cat-B is closer to Cat-A then Sw-D, Cat-B will be designated for that segment and Sw-D will accept the superior BPDU for VLAN 1 and become root forwarding.  Sw-D will forward that BPDU on to Cat-C.  Cat-C will receive three BPDUs for VLAN 1.  One will come directly from Cat-A.  Another will come from Cat-B via the segment between Cat-B and Cat-C.  If you tune the cost on the links so that the BPDU forwarded by Sw-D onto the Sw-D <-> Cat-C segment is superior to the BPDU forwarded onto that segment by Cat-C then Sw-D will become desginated forwarding for that segment.  Cat-C will block because it already has a root port on the segment it shares with Cat-A.

                          • 10. Re: STP load balancing
                            isaac

                            Thanks Steven

                            I got it to work after your explanation.

                             

                            I'm still in the dark(I think I will need to restudy load balancing using cost)

                             

                            the cost via SW-D is 57(19+19+19), on the segment between D&C. SW-D also has cost =57(19+38)

                            and SW-D has worse BID(2cfa.xxxx.xxxx) compared to CAT-C(0016.xxxx.xxxx),y is the DP on SW-D if the cost is equal?

                             

                            the cost via CAT-D is 39(20+19)

                            and cost is 38 on port 2/3

                             

                            what am I missing here

                             

                            only costs on CAT-C have been changed, the other ports are at 19

                            • 11. Re: STP load balancing
                              jh

                              The snippet from the book you've given may be out of context, but I'm not sure that you are missing anything; the text does say increase them "enough".