Can we use spanning-tree portfast port to connect switches if ...?

[interfacedy]

Can we use spanning-tree portfast port to connect switches if these switches do not have physical loop topology?
For example, sw1---sw2----sw3---sw4

[balaji.bandi]

spanning-tree portfast  - suggest to config always where end device connected (not to connect any switch) - even its not STP loop you know or aware not there.

instead, you can use other methods available :

https://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst3850/software/release/16-8/configuration_guide/lyr2_lyr3/b_168_lyr2_lyr3_3850_cg/b_168_lyr2_lyr3_3850_cg_chapter_010.html

[interfacedy]

Thank you for your reply. portfast should always be used for end device. if there is no any physical loop structure, the real loop at switch would not happen anyway, can i say it like that? 

[balaji.bandi]

Ok nice, personally i would not advise if you using are Trunk or port-channel here.

if this is access port and configured as yes you can (but int eh future if you like to extend any VLAN to other switches, you may have to bringdown the switch to config) - to minimise this we need to set port as Trunk.

Hope this make sense ?

[Giuseppe Larosa]

@interfacedy ,

STP is used to prevent bridging loops. once a bridging loop is formed in some cases unplugging cables or switching off a core switch ( before VSS / SVL era)  is/was the only way to recover.

From where it comes the recommendation to avoid to use portfast between switches ?

They have many ports and it is very easy to build a loop it is just enough in your topology to add a cable between SW2 and SW4 for example.

Avoiding to use portfast when a switch is detected on the link helps to prevent a loop.

I can report that a customer years ago was able to tear down the whole site with simple:

adding unmanaged switches downstream access layer switches

having someone connecting with one cable two ports of the unmanaged switch together.

We had to go on site and it took time to recover

So yes deploy STP portfast on access ports but use also STP BPDU guard.

The switches need the time to detect each other and to revert to normal STP behaviour.

Hope to help

Giuseppe

[MHM Cisco World]

"Rapid STP (RSTP) — Detects network topologies to provide faster convergence of the spanning tree. This is most effective when the network topology is naturally tree-structured, and therefore faster convergence might be possible. RSTP is enabled by default."
from cisco

[Joseph W. Doherty]

"RSTP is enabled by default."

I may be mistaken, or defaults have changed with more recent switch IOS versions, but I recall (?) Cisco default STP was not rapid variant.

Perhaps you meant to write PVSTP was enabled by default?

[Richard Burts]

This discussion has evolved into discussion of various forms of spanning tree. But the original question was asking about configuring port fast. Port fast is advised for ports connecting hosts in the network. port fast is not advised for ports connecting to other switches (but is not necessarily a problem).

The original post suggests a topology like sw1---sw2----sw3---sw4. Humans can evaluate this topology and determine that no loop could be present. But switches can not make this determination without using spanning tree. 

[Mohsin Alam]

i agree to Richard's point , not recommended to have it on trunk.

You never know when you will expand the topology or add more devices. 

## Make sure to mark post as helpful, If it resolved your issue. ##

[paul driver]

Hello

Yes you can, As long as they are access ports and not trunks