09-20-2015 09:57 AM - edited 03-08-2019 01:52 AM
Hi
I have a network setup as follows:
Router
|
|
Core Switch 1----------Core Switch 2
The switches are running HSRP for all the access layer switches. Now I understand this is not the best setup but its all we have. The router has no spare interfaces.
Currently there is static routing but I want to configure EIGRP. In this scenario, would the core switches be EIGRP neighbours with each other or can the router form and EIGRP relationship with both switches.
To rephrase would The core switches act as a single entity running layer 2 between them and form a single relationship with the router and do we have to separate the 3 devices into 3 entities.
Thanks
Solved! Go to Solution.
09-20-2015 12:58 PM
It depends on how the router is connected to the switch.
If you use a vlan and have an SVI on each switch and the router interface in the same vlan then the router will peer with both switches.
If you connect the router to the switch with a L3 routed link then it will only peer with the switch it is connected to.
To be honest I am not sure what benefit you get by peering with both switches as long as the switches peer to each other so the switch that is not connected to the router gets the routes.
Any traffic to and from the router has to go via the switch it is connected to anyway.
In fact you may be better off just peering with the router from the connected switch otherwise you may get suboptimal traffic flows eg.
you have an access switch with a client vlan on it and the uplink from the access layer switch to core switch 2 is blocking due to STP. If the router peers with both switches then it may send traffic to a client in that vlan to the next hop IP on core switch 2 but then core switch 2 has to send it back across the interconnect to the other switch because it has the active link to the access layer switch.
If you only peer with core switch 1 then traffic would only be sent at most once across the interconnect if it happened that core switch 2 has the active link to the access layer switch.
Hope that makes sense :-)
Obviously though if you do only peer with the switch the router is connected to then the core switches must also peer with each other.
Jon
09-20-2015 09:07 PM
If EIGRP is the path you want to go, you will have to remember how EIGRP discovers adjacencies/neighbours, which is by means of a multicast Hello on 224.0.0.10. This means that both your switches and your router, will somehow need to be in a shared subnet for multicast to happen and successful discovery can take place. If this is not the case. or if you cant make that happen, then your switch2 will not be able to establish an EIGRP adjacency with your router. That would mean that deploying EIGRP would not be very use full. Having said that, it should also be possible to configure "static" unicast eigrp nighbours (although I have never tested this myself).
To be honest, because you still have the single point of failure between router and switch 1, you will not add more redundancy anyway, unless I am missing your point.
HTH
09-20-2015 12:58 PM
It depends on how the router is connected to the switch.
If you use a vlan and have an SVI on each switch and the router interface in the same vlan then the router will peer with both switches.
If you connect the router to the switch with a L3 routed link then it will only peer with the switch it is connected to.
To be honest I am not sure what benefit you get by peering with both switches as long as the switches peer to each other so the switch that is not connected to the router gets the routes.
Any traffic to and from the router has to go via the switch it is connected to anyway.
In fact you may be better off just peering with the router from the connected switch otherwise you may get suboptimal traffic flows eg.
you have an access switch with a client vlan on it and the uplink from the access layer switch to core switch 2 is blocking due to STP. If the router peers with both switches then it may send traffic to a client in that vlan to the next hop IP on core switch 2 but then core switch 2 has to send it back across the interconnect to the other switch because it has the active link to the access layer switch.
If you only peer with core switch 1 then traffic would only be sent at most once across the interconnect if it happened that core switch 2 has the active link to the access layer switch.
Hope that makes sense :-)
Obviously though if you do only peer with the switch the router is connected to then the core switches must also peer with each other.
Jon
09-21-2015 01:03 AM
Hello
Just like to add to Jons comments
I understand your post is querying about eigrp implementation, however even if you connect via an IGP I dont think would mater, As there isn’t any physical resiliency
The physical aspect is still the same, if the connection from router to 1st switch or stack goes down Connection is lost to the router even its connected to a routed port or vlan via access port.?
To make it be a viable dyanmic routing solution an addtion physcal port is required on the router.which you state isnt availble
res
Paul
09-21-2015 01:47 PM
Thanks for the advice, Its definitely not the ideal solution whether peering with both switches or the one. I will work with peering with just switch 1, and then switch 1 and 2 will peer with each other.
Ideally I will try to add another interface first and that will then provide the redundancy required
09-20-2015 09:07 PM
If EIGRP is the path you want to go, you will have to remember how EIGRP discovers adjacencies/neighbours, which is by means of a multicast Hello on 224.0.0.10. This means that both your switches and your router, will somehow need to be in a shared subnet for multicast to happen and successful discovery can take place. If this is not the case. or if you cant make that happen, then your switch2 will not be able to establish an EIGRP adjacency with your router. That would mean that deploying EIGRP would not be very use full. Having said that, it should also be possible to configure "static" unicast eigrp nighbours (although I have never tested this myself).
To be honest, because you still have the single point of failure between router and switch 1, you will not add more redundancy anyway, unless I am missing your point.
HTH
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide