03-24-2025 02:28 PM
Greetings,
We have a typical ACI data center fresh out of the box with two spines and 4 leafs. Our requirements will evolve in the long term to split the data center in half to another site near by for failover. In order to save money, I'm being asked if we can just split it in half between the two pods. That is 1 spine and 2 leafs per pod. The APIC of course will be 2+1 so no further purchase there. I see no supported configuration in the multi-pod white paper that supports a single leaf but to him this makes sense as it's active/active and will failover to the other pod in a failover event anyways. Can someone explain why this isn't a supported configuration or point to documentation that can explain it?
Solved! Go to Solution.
03-24-2025 10:49 PM
Hello @KVS7
I can understand your concern about running a single spine and two leaf switches in each pod of your ACI fabric. While it might seem cost-effective initially, it deviates significantly from the supported architecture and introduces several critical limitations and risks.
The idea of splitting the existing infrastructure might seem appealing from a cost perspective, the risks associated with a single spine configuration in each pod are significant and outweigh the potential savings. It's crucial to adhere to Cisco's recommended architecture to ensure the stability, reliability, and scalability of your ACI fabric.
Hence, I won't suggest you to split such a small Pod in the initial phase. You anyhow can add another Pod in the future. Please do create a provision charter prepared for future Pod.
Hope This Helps!!!
AshSe
Community Etiquette:
03-25-2025 07:21 AM - edited 03-26-2025 03:00 AM
Cisco got back and said we can stretch the fabric but not in the deprecated "stretched pod" configuration but just the current pod split in half with longer cables reaching to the other datacenter across the street. I don't think it's a widely published architecture but ACI does not know of the difference. I think they called it multi-tiering or some other variation of it but in any case that's what we're going with. Of course all we would need to purchase is another APIC for standby.
03-24-2025 10:49 PM
Hello @KVS7
I can understand your concern about running a single spine and two leaf switches in each pod of your ACI fabric. While it might seem cost-effective initially, it deviates significantly from the supported architecture and introduces several critical limitations and risks.
The idea of splitting the existing infrastructure might seem appealing from a cost perspective, the risks associated with a single spine configuration in each pod are significant and outweigh the potential savings. It's crucial to adhere to Cisco's recommended architecture to ensure the stability, reliability, and scalability of your ACI fabric.
Hence, I won't suggest you to split such a small Pod in the initial phase. You anyhow can add another Pod in the future. Please do create a provision charter prepared for future Pod.
Hope This Helps!!!
AshSe
Community Etiquette:
03-25-2025 02:12 AM - edited 03-25-2025 02:35 AM
I looked at remote leaf but it's not intended for resiliency so we may just stick with our current single datacenter pod and just use the legacy network in the other datacenter to house redundant services i.e domain controllers or anything redundant.
03-25-2025 07:21 AM - edited 03-26-2025 03:00 AM
Cisco got back and said we can stretch the fabric but not in the deprecated "stretched pod" configuration but just the current pod split in half with longer cables reaching to the other datacenter across the street. I don't think it's a widely published architecture but ACI does not know of the difference. I think they called it multi-tiering or some other variation of it but in any case that's what we're going with. Of course all we would need to purchase is another APIC for standby.
03-26-2025 03:28 AM
@KVS7 it looks Cisco has suggested you to geographically stretch Single Pod in two different data centers using long distance (Single Mode) optic cables. Please do ensure the latency to be <50ms in this case.
Keep Cisco Tac in loop.
Good wishes & vibes!
03-26-2025 07:18 AM
@KVS7 , can you describe your current topology and what is expected from the desired solution? This may give us a better understanding and provide a suitable answer to your needs.
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