DLR, transport zones and clusters

[rogerscual]

I think that I'm missing some concept in the relationship between DLR, transport zone and clusters. Here is my example:

I have a DLR that has interfaces in some logical switches, all those logical switches belong to the same transport zone, let say for instance TZ-A. TZ-A expand across two cluster Cluster-A and Cluster-B, now I create another transport zone this time the name is TZ-B, but it only expand across Cluster-A, after that I add a logical switch that belongs to TZ-B if I want to assign an interface from the DLR to the logical switch in the TZ-B that is not an option in other words I can't, but if I only add Cluster-B to switch in TZ-B now I can. My question is idea behind this? Why is this a restriction?

Thanks.

[smitmartijn]

Hi,

As per the manual: A given logical router instance cannot be connected to logical switches that exist in different transport zones. This is to ensure that all logical switches and logical router instances are aligned.

VMware NSX 6.2 for vSphere Documentation Center

This kind of makes sense, because a LDR spans all the hosts that a logical switch spans..If you have three hosts, with one in TZ1 and TZ2, one in TZ1 and one in TZ2...the LDR on the host in TZ2 cannot reach the logical switch in TZ1.

Makes sense?

[rogerscual]

Thanks smitmartijn‌

That is what I thought but does not look right, take a look at my implementation:

The logical switch test-test belongs to a Transport Zone Test and the logical switch APP belongs to Default-TransportZone. But...

The DLR has interfaces in both, no problem at all. It only seems to be a problem if both Transport Zone do not include the same clusters, in that case when I try to assign an interface to the new transport zone that does not include the same set of cluster that transport zone where the DLR is connected already, then that is not possible, but you can assign interfaces in the DLR to logical switches in different transport zone.

Thanks for taking the time to help me out here.

[smitmartijn]

For some reason this post was in draft...could've sworn I hit 'Add reply' 😞

-----

Ahah..well that kind of makes sense because if you have 2 TZs that are actually on the same clusters, the LDR will actually work. It'll stop working if you use 2 TZs that are on different clusters. It kind of looks like either an oversight in documentation or the code only disallowing when it'll actually not work and allowing when it'll work. I wonder what'll happen if you add the different TZs to the LDR and remove one of the TZs from a cluster later on...Either way it's a discrepancy, and in that case you should go with the documentation.

Stick to 1 TZ per LDR, otherwise you might get issues in the future.