Multi-Pod IPN failure scenarios?

[tuanquangnguyen]

What would be the behaviour of each pod as IPN breaks in a two-pod design? The White Paper seems to mention only for APIC cluster quorum not being achieved for one pod. However most of the time it is forwarding behaviour that we care about:

- Between endpoints in EPG/BD that are local to each pod?

- Between endpoints that are in the same EPG (or different EPGs in same BD) that are stretched across pods?

[Sergiu.Daniluk · Accepted solution]

Hi @tuanquangnguyen 

Since we can decouple the management plane (APICs) from the control and data plane (on spines/leaves), you can look at the "IPN failure" as simply isolation of two CLOS fabrics:

- communication between local EPs in same or different EPGs/BDs (regardless if the EPG is present in one or both pods) will function without issues

- communication between EPs located in different pods, part of the same or different EPGs/BDs, will not functions.

Cheers,

Sergiu