Spanning-Tree UplinkFast

Lessons Discussion
21

Dear Andrew ,

We also see that Alternate port exist on PVST …

Switch#show spanning-tree 
VLAN0001
  Spanning tree enabled protocol ieee
  Root ID    Priority    32769
             Address     0000.0CD9.D544
             Cost        19
             Port        1(FastEthernet0/1)
             Hello Time  2 sec  Max Age 20 sec  Forward Delay 15 sec

  Bridge ID  Priority    32769  (priority 32768 sys-id-ext 1)
             Address     00D0.974A.B1E9
             Hello Time  2 sec  Max Age 20 sec  Forward Delay 15 sec
             Aging Time  20

Interface        Role Sts Cost      Prio.Nbr Type
---------------- ---- --- --------- -------- --------------------------------
Fa0/1            Root FWD 19        128.1    P2p
Fa0/2            Altn BLK 19        128.2    P2p

Switch#

br/
zaman

22

PVST is a Cisco proprietary implementation of 802.1 D (so it doesn’t follow the standards exactly). One of the differences is the Alternate port.

There is a discussion on this topic here:

https://supportforums.cisco.com/discussion/10402686/stp-port-roles-8021d-aka-pvst

23

Hlw Rene,

Nice explanation in right WAY, GREAT !

So , if we use Uplinkfast then no downtime occure ?? So how SWITCHC will update its MAC table as SWITCHB updated its MAC table using Multiicast Frame sending by SWITCHC ? I think there is 15 sec downtime needed due MAC table flush in SWITCHC.

Another Questions …After STP convergence , Only Root Bridge will send BPDU(Hello time interval) and other SWITCH will receive BPDU [if no Topology Chage].

br//
zaman

24

Hi Rene, i noticed that the port cost of the uplinkfast port jumped from 38 to 3038, what’s up with that, there was no explanation of this, could you elaborate? Thanks,

VLAN0001 new root port Fa0/16, cost 3038

25

Hi Kam,

When you enable uplinkfast, the switch does two things:

  • Increase the cost to 3000 so that it's unlikely that this interface will be elected as designated port.

Hope this helps!

Rene

26

Hello Rene,

I have been going through your material & found it one of the best!

Have a small query WRT to the below paragraph.
----
Once SwitchC switches over to use its non-designated port it will create a dummy multicast frame. The source MAC address of this Ethernet Frame will be all the MAC addresses that it has in its MAC address table. In my example above this is only the MAC address of ComputerB. The destination multicast address is a proprietary Cisco MAC address.
----
let’s say if there are 100 mac in the mac table, how it’ll specify all 100 entries in source mac.

27

Udaya,
In your example, you would have 100 separate frames being sent - one sourced from each MAC in the table!

28

Thank you, Andrew.

If that’s the case, if there are huge entries in the mac-table, will it not burden the link?

29

I thought the same thing!

30

Hi all,

in the initial setup the port fa0/16 on SW3 is in ND/blocking state, why does it only need 30sec to move to forwarding state?
I would guess that it would stay 20sec in blocking before moving through the other states, no?

Thanks

Florian

31

Hello Florian.

As soon as SW3 realizes that the current root port fa0/14 is in a down state, then the fa0/16 port immediately goes into the Listening state for 15 seconds, and then to the Learning state. During those two states, it should recieve BPDUs from the root bridge on fa0/16 and once the 30 seconds are up, it will become the new root port and it goes into the forwarding state.

The key here is that the port will not remain in the blocking state, but will immediately change to the Listening and learning states once the current root port fails.

I hope this has been helpful for you!

Laz

32

Hi Laz,

thanks for your answer!
But this is only the case if uplink-fast is configured, right?
If it is not configured the port would stay in block state for 20 sec and then move through the other states!?

Thanks

Florian

33

Hi Laz,

just read the article again and now iam a bit confused…

When does a port actually stay in blocking state before it moves to listening and so on?

If the root port is down the blocking interface goes directly to listening?

And Uplink-fast just helps to move the port directly to forwarding, right?

Thanks

Florian

34

Hello florian.

To answer your first question:

But this is only the case if uplink-fast is configured, right?
If it is not configured the port would stay in block state for 20 sec and then move through the other states!?

Yes, my description in the above post is for situations where uplinkfast is not implemented. In case uplinkfast is configured, this is reduced from 30 seconds to almost 0 seconds. Cisco explains it like so:

The UplinkFast feature is based on the definition of an uplink group. On a given switch, the uplink group consists of the root port and all the ports that provide an alternate connection to the root bridge. If the root port fails, which means if the primary uplink fails, a port with next lowest cost from the uplink group is selected to immediately replace it.

This is found at http://www.cisco.com/c/en/us/support/docs/lan-switching/spanning-tree-protocol/10575-51.html#uplink_fast_theory

Concerning your second question:
When does a port actually stay in blocking state before it moves to listening and so on?
If the root port is down the blocking interface goes directly to listening?
And Uplink-fast just helps to move the port directly to forwarding, right?

A port remains in blocking state as long as there is another active path to the root bridge. To say it another way, a port remains in blocking state when STP has converged. If there is a topology change, ports do not remain in a blocked state for any period of time before moving to listening and learning. So yes, “if the root port is down, the blocking interface goes directly to listening” in normal STP operation. If you use uplinkfast, the listening and learning states are skipped.

It is only during root bridge elections that a port remains in a blocking state for 20 seconds. If a topology change occurs but the root bridge does not change, then this extra blocking state for 20 seconds is skipped.

A very clear step by step explanation can be found here:

Uplink failure without uplinkfast enabled: http://www.cisco.com/c/en/us/support/docs/lan-switching/spanning-tree-protocol/10575-51.html#uplink_fail

Uplink failure with uplinkfast enabled: http://www.cisco.com/c/en/us/support/docs/lan-switching/spanning-tree-protocol/10575-51.html#uplink_fast_fail

I hope this has been helpful!

Laz

35

Hi Laz,

thanks a lot.

Regards

Florian

36

19 posts were merged into an existing topic: Spanning-Tree UplinkFast

37

So I have a question on this. it is stated that dummy multicast are flooded out all ports that contain the switch with the broken links Mac address table. in addition this mechanism will cause the 20 second timer enacted by the topology table to stop and the other switches can go to receiving the new mac address information immediately. quote from web lesson:

Take a look again at the MAC address table for SW2. The MAC address (000c.29e2.03ba) that I highlighted belongs to H2. When SW2 receives an Ethernet Frame for H2 it will be forwarded to SW1 and it will be dropped! (Well at least for 15 seconds until the topology change mechanism kicks in…).

I tend to really learn when I have to explain this back to myself or another person as if I was teaching. When I did so the I saw an inquisitive explorative question that might arise if we just did not take something as fact because it was said.

==========

why does the dummy multicast cause the switches to stop holding those MAC addresses contained within the dummy multicast??? why does it not drop the dummy multicast like it does in the quote above from the web lesson??

Now if asked this I would simply say because the dummy multicast are a special address that the switches recognize so they simply drop those from its aging table and then they are able to receive.

however, I am just making the most logical guess that I can make as I did not see that posted anywhere. the true answer would be I don’t know for certain.

Anyway diving deep into questions like this helps me to remember the content better as its all embedded behind the reasoning and must be known to get to this point so I figured why not ask =)

I actually am being naughty, I have not read my book or tried to watch the videos to see if they state this detail. so I am being lazy. What I will do is this; I will go back read my book and try to go through and find the answer they give in my video and post here. I think its worth having the question on the forums as others can benefit as well. If someone provides an answer here before I can get back and post that’s great to as I a:) may not be able to find my answer or b:) will confirm what I find when I come back to post my findings!

Ok first information non related to my question but I will post as its useful:

Configuring Uplink Fast Uplink Fast is enabled on Access layer switches and keeps track of possible paths to the Root Bridge. Once the Uplink Fast feature is enabled globally, it is enabled for the entire switch and all VLANs. By default, when Uplink Fast is enabled, Cisco IOS software performs the following actions on the local switch:

  • The Bridge Priority of the switch is raised to 49,152
  • The Port Cost of all VLANs is increased by 3,000

These two actions ensure that the switch will never be elected Root Bridge, and it makes the path through this switch as undesirable as possible for any downstream switches. For this reason, Uplink Fast should never be enabled on the Root Bridge because it will lose its Root status or lose switches that have other downstream switches connected to them.

Tafa, Farai. Cisco CCNP SWITCH Simplified (Kindle Locations 2588-2596). Reality Press Ltd. Kindle Edition.

===========================
Next is the answer to my question. was not in the book basically the same as the web lesson it did have a extra piece of information or two:

By transitioning the port to a Forwarding state almost immediately, the Uplink Fast feature presents the potential problem of incorrect entries in the CAM tables of the other switches because they have not had an opportunity to re-learn the new path for the MAC addresses of the devices connected to the Access switch.

To prevent this, the Access layer switch on which the Uplink Fast feature is enabled floods dummy frames with the different MAC addresses that it has in its CAM as a source. The frames are sent to the Multicast address 01-00.0C-CD-CD-CD and appear to originate from the hosts connected to the switch so all the upstream switches can learn of these addresses through the new port.

By default, the switch sends out these Multicast frames at a rate of 150 packets per second (pps). However, this value can be adjusted by using the spanning-tree uplinkfast max-update-rate [rate] global configuration command.

Tafa, Farai. Cisco CCNP SWITCH Simplified (Kindle Locations 2618-2620). Reality Press Ltd. Kindle Edition.

===============================
so next I will try to scan through my INE CCNP video and see if they mention the answer to my question!

===================

They did not give an answer on INE as well they just said it sent dummy multicast and that fixed it. so I am only left to guess that it fixes it and is not dropped because the address is special and the software code says hey when you see this address drop the MAC addresses and add these Mac addresses from this location or maybe it does not drop but just changes but the effect is the same.

if someone is able to find this answer feel free to post at least I now did my due diligence in trying to find the answer!

note: I did find the following it does not spell out the answer of how either, it does once again logically suggest that the “how” is because its a special address and strengthens it with terms such as “ensures” ect…

http://www.cisco.com/c/en/us/support/docs/lan-switching/spanning-tree-protocol/10575-51.html

In order to solve this problem, switch A begins to flood dummy packets with the different MAC addresses that it has in its CAM table as a source. In this case, a packet with C as a source address is generated by A. Its destination is a Cisco proprietary multicast MAC address that ensures that the packet is flooded on the whole network and updates the necessary CAM tables on the other switches.

38

Interesting lesson ; but, some potential confusing points to be clarified.

* Point-1

In one of the previous lessons (TCN), it was said about 50s delay for moving from Blocking to Forwarding state. This lesson does only talk about 30s delay. Something important that this lesson has to say is : SW3 does not need to wait for the 20s of “Max Age” (parameter setting on the switch) ; because it detects immediately that a port (belonging to the switch) is not working.

* Point-2

The lesson is saying that the Listening and Learning delay could be as lower as 14s. It would be important to recall that this delay is always twice (2 x 7s) the “Forward Delay” parameter (setting on the switch).

* Point-3

The end of the lesson is very confusing. Once the link between SW3 and SW1 is back ; nothing explains by which process and how long it takes for the ports of those two switches to get back to their former states.

39

Hello Maodo

In the TCN lesson, it mentions that there will be a maximum of 50 seconds delay when there is a change in topology. When a TCN is received, a port will go into the blocking state for the Max Age interval which is an interval of 20 seconds by default before beginning the STP recalculation. The maximum of 50 seconds is ONLY for topology changes.

In this lesson, it is mentioned that it takes a maximum of 30 seconds for STP to converge when STP function occurs from scratch, that is, after the switches are rebooted or turned on. This includes ONLY the Listening (15s) and Learning (15s) states. In this case, 30 seconds is only needed.

How the result of approximately 14 seconds has been achieved depends on several factors including network diameter. For further information, you can check this Cisco documentation that further explains STP timers in detail.

As is stated in the lesson, essentially nothing happens. The root port remains the same and Fa0/14 remains blocked. Since the network is currently working, no changes have been made, BPDUs are successfully exchanged, the network remains in this state until there is a topology change.

I hope this has been helpful!

Laz

40

About my Point-3, I’m still confused about the status of SW1-fa0/17 and SW3-fa0/14 (while and after the link failure).

- Why the link failure is detected only on SW3 side and not on SW1 side ?
- How come SW1 did nothing with fa0/17 when it gets TCN from SW3 (fa0/17 stays in “D” status) ?
- Why did SW3 not send a TCN to SW1 (ROOT) after the link is repaired ?

It is also hard for me to understand the consistency between the last schema and the last debug informations. The schema says nothing about ports fa0/14 and fa0/16 on SW3 ; but, the debug information says : fa0/14 -> blocking, new root port fa0/14 -> , fa0/16 -> blocking. Does SW3 tries to return immediately to the state before failure ?