Posted on April 21, 2017 5:54 pm
 |  Asked by Ajay Chauhan
Print Friendly, PDF & Email

what is the command on different arista platforms to see how many ECMP group objects are supported ? in a Spine ,leaf architecture let’s consider 10 routes are being learned from 16 fixed spines ,does that form one ECMP group or its going to be per prefix.



Posted by Alexandru
Answered on April 22, 2017 12:16 pm

Hi Ajay,

The knob you are looking for is maximum paths (BGP):

• paths maximum number of parallel routes. Default value is 1.
• ecmp_paths maximum number of ECMP paths for each route. Default is maximum value.

Value for each parameter ranges from 1 to the number of interfaces available per ECMP group, which is platform dependent.
• 7500E: Value ranges from 1 to 128. Default value is 128.
• 7150: Value ranges from 1 to 32. Default value is 32.
• 7050: Value ranges from 1 to 32. Default value is 32.
• 7050X: Value ranges from 1 to 128. Default value is 128.
• This command configures the maximum number of BGP parallel paths to 2.
The ECMP value for each route is 4.

switch(config)#router bgp 1
switch(config-router-bgp)#maximum-paths 2 ecmp 4

Posted by Ajay Chauhan
Answered on April 25, 2017 8:45 am

Thanks Alexandru.

I am aware about what you explained.

I got this definition somewhere in cisco forum and was wondering if this is something which can be checked on box.

ECMP object is used by an IPv4 route/IPv6 route/MPLS entry when there are multiple paths corresponding to that prefix. ECMP object consists of an ECMP group object which points to an array of member paths. Each member path corresponds to a specific path through which the packet can be forwarded.
When there is only a single path to forward the packet, ECMP object will not be used.
Each ECMP object uses entries from the member path table. The number of entries used is equal to the number of paths in ECMP. If either of the resources run out, we will hit OOR condition and this will result in not programming rest of the route updates.


Posted by Aesha Parikh
Answered on April 27, 2017 11:49 pm

Hi Ajay,

Each ECMP group gets programmed in the ECMP next-hop table with a unique ID. If 10 prefixes are reachable via same ECMP group, all these routes are programmed with the same ECMP next-hop ID. In other words, there will be only 1 ECMP next-hop ID in the next-hop table and not 10. If there are multiple different ECMP groups for different routes, you are limited by the next-hop table size but its rare to reach these limits as multiple routes will be reachable via same next-hop (single/ECMP). Each platform have different L3 table sizes. Are you looking for scale limits for a particular platform? 



Posted by Ajay Chauhan
Answered on April 28, 2017 6:00 am

Thanks Aesha,

That does make sense . Let me give you scenario what I was thinking .

Let’s say all routes are not coming from same set of next hop for example 10/8 coming from 2 set of next hop and 20/8 coming from different set of next hop , meaning 2 ECMP group . I was trying to explore that limit if we have any and supported by CLI .




Post your Answer

You must be logged in to post an answer.