Unanswered Question
Aug 23rd, 2010


After Uplinkfast is enabled, when Root Port fails, access switch quickly sends dummy multicast frames to upstream bridges over the new Root Port on behalf of the stations in CAM table.

Currently convergence time from direct link failure takes 30 seconds (2 Forward Delay), shy not use Uplinkfast to reduce convergence time on non-access switches?


  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
Giuseppe Larosa Mon, 08/23/2010 - 14:02

Hello Jingyi,

unless you are already using a form of Rapid STP (MST or Rapid PVST) enabling uplinkfast on access switches is a good move

your understanding is correct

note: to be enabled on devices that haven't other switches connected downstream or you will need to look for backbone fast that is a different but related matter.

generally speaking migrating to a Rapid STP version will give you fast convergence if edge ports are correctly defined (using spanning-tree portfast).

Hope to help


shelley_wang Mon, 08/23/2010 - 16:24

Hi Giuseppe,

Thanks for your response.

Backbonefast is designed to reduce the convergence time of indirect link failures. Uplinkfast is designed to reduce the convergence time of direct link failures, but only on access switch. I don’t understand why Uplinkfast cannot be used on non-access switches.



narainarun Mon, 08/23/2010 - 22:32

Hi Jingyi,

I will try my best to explain why uplink is not recommended (its not that cannot be used) in non-access switches. Non-access switches (distribution) will be usually in the transit path. When configuring uplinkfast below things happen in the switch:

1. the switch prioirty will be increased. this will make the switch to be least preferred for root bridge election.

2. the STP port cost of all ports are increased in that switch which will make the switch less preferred as transit switches.

For the above two reasons, if uplinkfast is enabled in non-access switches i.e. distribution which acts as transit for the traffic, then it will have very undesirable impact on traffic flow. This could be the reasons why Cisco does not recommend to enable it on non-access switches.

As Guiseppe said, you can look for enabling either RSTP or MSTP as these protocols have in-built rapid convergence mechanisms integrated into it.



Giuseppe Larosa Tue, 08/24/2010 - 03:59

Hello Jingyi,

>> esigned to reduce the convergence time of direct link failures, but only on access switch. I don’t understand why Uplinkfast cannot be used on non-access switches.

because it would be dangerous as it bypasses all STP state transitions using the alternate link, what if multiple switches react to a topology change?

The risk is to get a loop.

STP target is to prevent bridging loops by handling link redundancy.

backbone fast as 802.1W and others use Root link query messages to ask for an alternate path to the root to the neighbors so it is totally different then uplink fast where the single device reacts on its own without consulting its neighbors.

with backbone fast you can save the max-age time in case of an indirect failure but the selected port has to move on STP listening and learning states before going to forwarding state.

Hope to help


shelley_wang Tue, 08/24/2010 - 10:11

Hi Giuseppe,

It will makes more sense if you can give example and diagram.




This Discussion