VLAN tagging not working properly

[HendersonD]

We have a situation where External Switch Tagging, which we prefer, will not work on some new HPE server while Virtual Switch Tagging works fine. The two diagrams below tell the story. Is there something wrong with our setup? Something else to check?

[daphnissov]

On the first example (External Switch Tagging), if you're prefering VLAN 123 as the native VLAN, try to add that VLAN ID as a tag to the vmkernel port. Do you have connectivity? On some hardware, you still need to specify the VLAN ID even if it is in access mode. I don't know if that's the case here, but try it out and see if that changes the situation.

[HendersonD]

I have been looking at this article

VMware Knowledge Base

It states: "The portgroups connected to the virtual switch must have their VLAN ID set to 0."

If the vlan is set on the access port on the switch, it sound like I should not be setting a vlan on the ESXi side

I believe the vlan on the ESXi side is set on the portgroup and not on the vSwitch or kernel

[daphnissov]

Yes, that's true. If you're in trunk mode and you're passing multiple VLANs, you normally tag those on the port group side, which then strips the 802.1Q tag off the packets before they're forwarded. If you're in access mode you do not set a VLAN ID on the port group (by setting it to 0). According to your switch port config, if in access mode it should work, but if it isn't, try tagging the vmkernel port group to see if it passes the traffic. If so, it may need to be removed as the native VLAN.

[HendersonD]

As soon as I set the vlan ID on the portgroup to 123, the management interface stopped working. This is with the two switchports on my Juniper switch set to access with vlan of 123

[daphnissov]

Then clearly ID 0 needs to be set when in access mode. As for why some new HPE servers are working and others aren't, there may be some hardware settings influencing this. From a vSphere perspective, you have it right, though.

[HendersonD]

That is what I thought. I have ticket opened with Juniper about this as well as HPE