STP / RSTP query with Etherchannel

[mvsheik123]

Hello,

Correct me if Iam wrong...

Etherchannel between 2switches considered as single link where as Primary trunk & backup trunk considered 2 seperate links.

With STP/RSTP the convergence will be on the link incase primary trunk failure (for dual trunk).

In case of Etherchannel the traffic will be converged to working link incase of link failure.

The question is - incase of Etherchannel , does it make any difference with STP/RSTP use?

Thanks in advance.

MS

[krahmani323 · Accepted solution]

Hi Mvsheik,

If I understood correctly, the question is about to know how the Etherchannel spanning-tree parameters reacts in case of a bundle link failure..

If yes, and as you stated the Etherchannel is considered to be a single logical link in regards to the STP with its own cost following the same STP rules as normal link…

The reactions is all about the Po STP cost evolution in case of a link failure.

By default =>

Interface	STP Port Cost Short Mode (16-Bit)	STP Port Cost Long Mode (32-Bit)
100-Mbps Ethernet	19	200,000
GE	4	20,000
Two-port Gigabit EtherChannel (GEC)	3	10,000
Three-port GEC	2	6666
Four-port GEC	2	5000
Five-port GEC	2	4000
Six-port GEC	2	3333
Seven-port GEC	2	2857
Eight-port GEC	1	2500
10-GE	2	2000
Two-port 10-Gigabit EtherChannel	1	1000

(An internal formula is used to determine the default cost for an etherchannel which leads to the above default values)

Now in case of an link member failure =>

-          In Cisco IOS, the port cost for the EtherChannel is updated immediately in order to reflect the new available bandwidth.

-          In CatOS, the spanning-tree port cost for an EtherChannel stays the same after the port channel member link failure.

* http://www.cisco.com/en/US/products/hw/switches/ps708/products_white_paper09186a00801b49a4.shtml#cg6

(Spanning Tree Port Cost Calculation)

-          In NXOS, for vPC links, all vPCs (except vpc peer-link) have the same cost of 200. This cost is hard-coded and does not depend on the number of links in the PortChannel.

* http://www.cisco.com/en/US/prod/collateral/switches/ps9441/ps9670/C07-572829-01_Design_N5K_N2K_vPC_DG.pdf p12 ( Spanning-Tree Cost for vPC)

The recommended spanning-tree configuration in preparation for a vPC deployment is =>

Enable spanning-tree pathcost method long to account for the use of 10 Gigabit Ethernet links (and whenever multi(Ten)gigabit links are involved for the differentiation of such STP costs) 

Otherwise :

=> if we need more granularity and keep a dynamic cost update then the command spanning-tree pathcost method long could be configured

=> If we want to keep a stability, ie no possible reconvergence due to Po path calculation (many does not want a stp recalculation due to an Etherchannel link failure!), spanning-cost under the interface is the command

Best regards.

Karim

Message was edited by: Karim RAHMANI => "spanning-tree pathcost method long" clarifications

[Peter Paluch]

Karim,

Very nice answer indeed! One question, though:

In order to avoid any possible recalculation and possible link failover in IOS spanning-tree pathcost method long is recommended.

As far as I know, this command merely changes the size of the root path cost field in the BPDUs (or better said, it changes the used size of the fixed 4B field) resulting in larger range of available port costs and resulting root path costs. Are you sure it prevents EtherChannels from changing the cost if their bandwidth changes as the result of link failures? Should not the spanning-tree cost interface-level command be sufficient for that?

Best regards,

Peter

[krahmani323]

Hello Peter,

Thank you very much!

Yes Indeed I think I was not very clear on this last point; was thinking in fact the contrary => if we need more granularity and keep a dynamic cost update then the command could be configured. 

And as you said if we want to keep a stability, ie no possible reconvergence due to Po path calculation (many does not want a stp recalculation due to an Etherchannel link failure!), spanning-cost under the interface is the command.

Thanks for pointing that.

Kind regards.
Karim  

[Peter Paluch · Accepted solution]

Hi,

STP/RSTP/MSTP sees the entire EtherChannel bundle as a single interface. A failure in one of the links in an EtherChannel bundle will not result in a port-down event to STP but it will cause a decrease of the bandwidth of the Port-channel interface and as a result, the STP cost of the Port-channel interface will change as well. This may trigger a change to the active topology but the extent of the change depends strongly on the particular physical topology and the STP parameters of the network. Setting the STP cost of the Port-channel interface manually using the spanning-tree cost command would prevent the STP from seeing the change altogether untill the entire EtherChannel goes down (all ports get disconnected).

Best regards,

Peter

[mvsheik123]

Thank you Both.

MS