tier-0 gateway number per edge cluster

[vmb01]

On the manual I read:

An Edge node can support only one tier-0 gateway or logical router. When you create a tier-0 gateway or logical router, make sure you do not create more tier-0 gateways or logical routers than the number of Edge nodes in the NSX Edge cluster..

What happens if I create 2 T0 GW on a cluter composed by two  nodes and than one of the two nodes fails?

[ZibiM]

Hello

Couple of things to consider:

1. NSX-T 3.0 brought VRF-lite -> You can create one main T0 and then subsequent number of subordinated T0s for each VRF (tenant). Those subordinated T0s should not count toward the limit of 1 T0 per node.

2. You can create Edge based on VMs -> You just need to create as many VM Edges as you need for your T0s

3. You can create multiple Edge clusters -> you can for example have dedicated Bare Metal Edge cluster for your VRF enabled T0s and Edge VM cluster for those T0s that are independent.

IMHO limit 1 T0 per Node is kinda valid for Enterprise workloads -> for the cases where T0 is really the sole big pipe from Enterprise DC networks toward the north.

As such it's crucial for whole environment availability and performance.

For the cloud providers you have really small T0s - majority of the tenants can be small, and not many of them will be in any meaningful size

Back to your question - i'd guess it depends really 🙂

Size of the edges, amount of the function inside, T0 topology (active-active or active-standby ?)

[mauricioamorim]

Why do you want 2 T0 gateways with only 2 edge nodes? With this configuration you don't have high availability. If an edge goes down you lose the T0.

What you should do is work with a single T0 and have it leverage both edges in the single T0, either active-active or active-standby. A single T0 can span up to 8 edge nodes in an active-active configuration.

[vmb01]

Thanks for the answers.

My question was just releated to what is oficially supported and what not.

I do not want 2 T0 GW on a cluter composed by two  nodes but I need to know if this is a configuration supported...

Similar questions could be:

-is it supported an edge cluster composed by 8 nodes with 8 T0 inside?

-is it supported an edge cluster composed by 8 nodes with more than 8 T0 inside?

[simonre]

Hi vmb01,

Just to add to what others have said here...

In a NSX-T 3.0 lab I added four edges to one edge cluster.

I added an Active/Active T0 with interfaces on sa-nsxedge-01 and sa-nsxedge-02.

I found:

1) If I tried to add a second Active/Active T0 with an interface on either sa-nsxedge-01 or sa-nsx-edge-02, then the operation failed with an error (see screenshot).

2) If I tried to add a second Active/Active T0 with interfaces on sa-nsxedge-03 and sa-nsxedge-04 (ie on edges with no existing T0 interface) then it was fine (see screenshot).

So it seems:

o I can have two T0s in one edge cluster,

o but each edge node can have interfaces from only one T0.

Cheers

Simon

[mauricioamorim]

What would be the use case? Configuration works, you can create as many T0s as edge nodes you have in a cluster, but what is the use of this type of configuration. If you have 8 nodes in an edge cluster you will have no more than 8 T0.

Although having multiple T0 in the same edge cluster works, I suggest that you dedicate clusters to T0. It will make things much simpler.

[ZibiM]

Cloud provider environment with strict network separation for customers (tenants)

The requirements for that seems to be either having T0 with VRFlite subT0s or separated T0s for the tenants

Imagine having 100+ pair of T0s

Anyway I'm really looking forward for the NSX-T Implementation Guidelines for vCloud Director

[daphnissov]

I can guarantee you that in a service provider setting where they're providing networking services to tenants, they are going to absolutely need redundancy and resiliency, so no one is going to deploy a single T0 to a single edge node.

[ZibiM]

Yes, that's right

We are looking at plenty of nodes in different formats

We are looking at setting up different Edge Clusters for different T0 roles

On the last Vmworld there was Nimish Desai session concerning Large Scale and Providers.

Landing page

Around 34 minute the fun part begins

MultiTier T0

VRF lite

This is really a maze that will be hard to get right 😕

[RaymundoEC]

I guess you have seen documentation on version 2.5.x so if that is the case you can only support a routing context on an EDGE cluster, so if you want to have more than one then you need another cluster to support it , being said having 2 T0's backend different clusters.

I f you consider version 3.X this is not a problem since you can have more than one context in same T0 GW.

[TronAr]

Raymundo,

the 3.1 docs still say only 1 T0 per edge node...

[gfirth77]

If you check the note its for NSX Cloud only which is a deployment to the cloud provider, not on-prem. I too was trying to find the limit for on-prem, but cannot find any numbers. I need to do stretched segments for front-end segments and stretch them between 2 sites which you cannot do. Global Manager does not support VRF yet either so my left over option is multi-T0 to the same edge. I am playing with deploying multiple T0 from GM to the same Edge cluster. The system hasn't complained yet and I've deployed 5, but not connected them yet to an uplink VLAN for BGP peering.

[TronAr]

https://configmax.esp.vmware.com

NSX-T Data Center 3.2.0 Configuration Limits

Logical Routing

Tier-0 Gateways per Edge Node              1

Tier-0 Logical Routers per Edge Node     1

[gfirth77]

Thanks for that @TronAr. I didn't see that someone else further in the chain had tried with screenshots. I ran into the same problem when I tried to assign interfaces. I hope future releases they can do Global T0 with VRF support.