Core resiliency HSRP Etherchannel

[josegarcia123]

I have an issue in my hands, we made a project and the switches were bought before lab tests were done the issue is with core redundancy and resiliency, the structure is as follow

Core: 2 x C4948 - Uplinked in Etherchannel HSRP (Gateway to the servers / Teaming)

Dist: 2 x 3750X-24S-S - Stacked and Uplinked to the C4948

Access: 10 Switches uplinked in etherchannel to the distribution

My question is what happens to the core if the Etherchannel fails? Is there a way to setup redundancy through the distribution for this scenario without too much resiliency?

[Reza Sharifi]

My question is what happens to the core if the Etherchannel fails? Is  there a way to setup redundancy through the distribution for this  scenario without too much resiliency?

Well, that is one of the purpose of Etherchannel, so you have multiple physical links in it and in case one link fails you have another one or more to use.  With Etherchannel from access switches to distro and from distro to core, you have enough links and redundancy.

HTH

[josegarcia123]

Reza thank you for the reply, but what happens when you lose one of the cores in such architecture, from lab tests the client is claiming that in case a core fails the distribution will simply go crazy since we won´t have HSRP working anymore between the cores. Does this make sense?

[Reza Sharifi]

Carlos,

It depends on the design.

If the 3750 stack is doing the routing, you have multiple (two in this case) so you should be fine and there is no need for HSRP or VRRP since the stack is logically one switch.  If the 4948s are routing then you need to run HSRP, or VRRP on them so in case one fails, the other one will take over.

HTH

[josegarcia123]

Reza

the 3750 would do Intervlan and C4948 L3, but just so I understand if I use etherchannel

HSRP between the C4948 and one of the switches fails the distribution/core will see this and foward traffic only to the active core? also when the switch is available again will the HSRP restore automatically?

So basically with this setup I would have redundancy if one C4948 failed and my access switches would still have connection to the servers?

[Reza Sharifi]

Ok, if the 3750s are doing intervlan routing, then they are doing layer-3.  In this case no HSRP or VRRP need it, but you need to make your portchannles connecting the 3750s to the 4948s layer-3 with /30s.  You also need a layer-3 link between the 4948s.  You also need to tun a routing protocol between the 3750s and the 4948s. So, lets say you are using OSPF, that will take care of any link failure or device failure.  So, you have redundancy.

HTH

[josegarcia123]

Reza,

another issue is C4948 is running IP Base so no support for OSPF but C3750X IP Base has limited support for OSPF, what are my options, how would you set this up?

C4948 IPBase:

IP Base image

Standard Layer 3 image, including Routing Information Protocol Version 1 (RIPv1), RIPv2, static routes, and Enhanced Interior Gateway Routing Protocol (EIGRP) stub

C3750X IP Base:

OSPF  for Routed Access supports only one OSPFv2 and one OSPFv3 instance with  a combined total of 200 dynamically learned routes. The IP Base image  provides OSPF for routed access.

[Reza Sharifi]

Carlos,

I would spend the money and get IP Services license for the 4900 and run OSPF, unless you want to do RIP

HTH