NSX Logical Switches and VGT Mode

[rikherlaar]

Hi,

I reckon the answer is "no" but just to play it safe, can I assume that VGT is not supported for virtualwire Port-Groups ?

The configurations allows to defune vlan trunking alright but my thinking is we need a one-to-one relationship between vlan-id and vxlan vnid - (if at all a vlan-id is used e.g. for L2GW purposes)

The questions came about while testing a nested environment...

cheers

/Rik

[Matthew Larson]

Not some thing I have actually done, but I have read the following:

This one specifically references using VGT:

NSX Minimum MTU – VMware Professional Services

The release notes for NSX 6.2 mention Ability to keep VLAN tags over VXLAN as a new capability:

VMware NSX for vSphere 6.2.0 Release Notes

[rikherlaar]

Thx for the pointer. Indeed the RN specifies it as supported - but little detail is given alas and cannot track anything very  useful back in the configuration guide. Guess I'll give it a shot by adding a second address under the same logical interface at the ESG level , at that point however you cannot tell which vlan-id = affiliated with what subnet ...so guessing load here that it's sort it out w/o any vlan tags at the ESG level...

Still curious how that would pan out on L2GW but I'll give that a shot too just to be sure. Thx again.

/Rik

[rikherlaar]

I can already confirm that L2GW (VLAN Briding) won't be possible in this setup - provided there's indeed a 1-to-1 mapping required -

Leaving either Distributed PG or the Logical Switch in "vlan trunking" throws up an error - as expected.

regards

rik

[Pooja Patel]

Enabling Guest VLAN tagging is supported with VXLAN Overlays. The method to enable it is a REST API call and not VGT mode on the virtualwire Port-groups.

REST API call to do the same is documented in the API guide. An example is as follows:

PUT https://<nsxmgr-ip>/api/2.0/vdn/scopes/vdnscope-1/virtualwires

<virtualWireCreateSpec>

<name>LS_vlan_tagging</name>

<description>For guest VLAN tagging</description>

<tenantId>virtual wire tenant</tenantId>

<controlPlaneMode>UNICAST_MODE</controlPlaneMode>    

<guestVlanAllowed>true</guestVlanAllowed>

</virtualWireCreateSpec>

[rikherlaar]

Thanks muchfor clarifying that part - but how to terminate these vlan-id's again at some end-point ? - i.e. i can do that with vlan backed port-groups , it's possible  to carry multiple vlan tags and keep isolation at the ESG level which invokes dot1q proper through "vlan-to-subnet mapping" , but how does an ESG with LIF=pseudowire with multiple vlan-tags - make the distinction between these vlans and keep aligned with the intended IP subnet.

The only way I see this (potentially) working is VM to VM (provided the VM runs dot1q and can filter based on vlan-id) - unless it's again an API only function?

Cheers

Rik

[Pooja Patel]

We do not support crossing DLR with Guest VLAN tagging. It has to be L2 only...VM to VM(within a VXLAN) or VM-to-VM(Software L2 bridging).

[rikherlaar]

Clear...thx