Provide raw trunk to an ESXi guest VM

[gsmitheidw]

I have Proxmox 7.4-4 on baremetal server - it is fed with a trunk which I have working tagged on vmbr0. Within proxmox I have an ESXi guest VM to which I wish to provide the raw trunk vmbr0. I've successfully got it working by creating entries in my config eg:

auto vmbr0.172
iface vmbr0.172 inet static
        address 172.16.2.103/12
        gateway 172.16.1.250

auto vmbr0.192                                                                                                         
iface vmbr0.192 inet static
       address 192.168.1.3/16

But I've to specify each vlan and create a nic on each guest. So I can do this for other tagged networks but I'd prefer to just pass on the raw trunk in it's entirety to ESXi (or any other hypervisor) and then break out the networks there in the VM with tags. Is this possible? I've only 1 physical trunk connection going into the host, so management vlan still needs to be accessible to the host regardless. In short I want the trunk to be as if Proxmox wasn't there at all and ESXi just saw the trunk as if it was directly into the switch.

Any suggestions?

The backstory is this is an educational cluster, we've provided nested hypervisors for segregation - Tech staff (me) looks after the bare metal, lecturers control the next layer of hypervisors and then the students are their "tenants" underneath on the next nested layer. We've been using ESXi on the baremetal layer, but we're keen to migrate that to Proxmox. The overhead of nesting is worth sacrificing some performance for logical segregation. The purpose is for students to learn the constructs of cloud - virtualisation and hands-on experience with different hypervisors etc. This is one node of a large cluster.

 

[gsmitheidw]

Just to finish this off - the solution was looking me in the face and I just didn't spot it.

The network config as per above post is fine on the proxmox layer, you still need the vlans broken out for the host. Particularly one for managenent of proxmox itself. But the simple solution is not to create a nic for each vlan and tag it in the guest.

Instead all I had to do in the guest was have a single "net0" attached to the vmbr0 bridge and not tag anything there - that is then is carrying the raw trunk connected to the host. Then I can just tag the management interface in the guest (ESXi) and tag other networks as required within the ESXi vswitch thereafter.

I hope this post helps others trying to do similar things.