Upon viewing the various VCF on VxRAIL deployment options, I noticed that there are changes in version 4.3 when it comes to the NSX-T Edges. In our current setup, we have 4 edge nodes that run in the VI Workload Domain. 2 on each site (stretched cluster).
But it appears that in VCF 4 and onwards, they deploy 2 edge nodes that are “active” on the preferred site. This could be why they changed the name from “stretched clusters” to “Availability Zones”.
So basically site traffic that is destined for the edge cluster goes over to the preferred site, and when that site goes down the edges do an HA restart and the secondary site go active, as seen in these diagrams:
I am curious of why VMware made this change to the design, currently we have 4 edge nodes, 2 in each cluster that are active-active. In the 4.X configuration, the overlay traffic will basically go across the intra-site links and then out through that cluster. This seems like it will have some networking performance impacts. We have applications which require that we run some workloads on both AZ`s.
According to the offical NSX-T documentation, the active-active deployment is supported. But does VCF 4.X support this topology?
Hi William, long time no see.
Active/Active across both AZs are supported in VCF 4.x. The reason VMware changed their recommended design is to avoid possible risk for asymmetric routing and to limit traffic between the AZs when you have Edge Nodes handling n/s traffic in two different data centers. If you have that under control, this design should be fine and we have many customers using it already, also with VCF on VxRail.