Redundancy in two Data centres

[q-le]

Hi All,

We have 2 x DCs. One is newer with all mod cons and 2 x 7K Cores plus 200 x3750 access switches and a smaller DC with 2 x  5K cores plus 50x 3750 access switches for surrounding areas.

The 7K and 5K are link by VPC.

Recently we deploy a new 3850 switch dual link (etherchannel) to 2 x 7K Cores.

However a user whom have  important devices connected to this switch - want the switch connected to both 7K and 5K for redundancy ie one link to 7K and one link to 5K.

is this recommended and can it potentially causing looping ?

Any advice are much appreciated.

Thanks

Ql.

[Francesco Molino]

Hi

I believe this switch is doing L2 only and L3 is connected to N7K right now.
If so, for sure, connecting redundant links on each DC will build a loop that will be contained by your STP. If STP is configured correctly, there’s nothing to be worried about as 1 link will disabled and come back UP when the active will go down. And that’s OK.

If you want to avoid L2 STP issues, you can see if you can bring L3 down to your 3850 for these specifics devices.

[q-le]

Hi Francesco,
what do you mean by bring L3 down to the 3850 if I want to avoid L2 STP issues?
Thanks
QL

[Francesco Molino · Accepted solution]

What i meant is if you want to avoid layer 2 stp (due to loop creation), you can create the SVI (Layer3 interfaces for vlans in where all critical devices are sitting in) could be created on 3850 and interconnect it with both Nexus 7k and 5k using routing.
In this situation, you just need to prefer a path over the other and no need to deal with stp.
I said that because by experience a lot of people don't want to deal with stp.

In your specific case, it's not mandatory as soon as you make sure stp is configured correctly and loop will be contained.