nsx_edge_transport_node fails to create edge with two standard_host_switch entries

[martinrohrbach]

Describe the bug

We've eagerly awaited the release of the new provider to test nsx_edge_transport_node to deploy new edges. The challenge for us is that we use two switches per edge and that did not work with the nsx_transport_node resource.

This is what we tested with:

resource "nsxt_edge_transport_node" "test2” {
  description  = "edge-test"
  display_name = "edge-test"

  standard_host_switch {
    host_switch_id = "xxx"
    ip_assignment {
      static_ip_pool = nsxt_policy_ip_pool.ip-pool-for-test-edge.realized_id
    }
    transport_zone_endpoint {
      transport_zone          = "xxx“
      transport_zone_profiles = ["xxx"]
    }
    host_switch_profile = ["xxx"]
    pnic {
      device_name = "fp-eth0"
      uplink_name = "uplink"
    }
  }

  standard_host_switch {
    host_switch_id = "yyy"
    ip_assignment {
      assigned_by_dhcp = true
    }
    transport_zone_endpoint {
      transport_zone          = "yyy“
      transport_zone_profiles = ["yyy“]
    }
    host_switch_profile = ["yyy"]
    pnic {
      device_name = "fp-eth0"
      uplink_name = "uplink"
    }
  }

  deployment_config {
    form_factor = "SMALL"
    node_user_settings {
      cli_password  = "VMware1!VMware1!"
      root_password = "VMware1!VMware1!"
    }
    vm_deployment_config {
        <redacted>
    }
  }
 
  node_settings {
    hostname             = "edge-test"
    allow_ssh_root_login = true
    enable_ssh           = true
  }
}

Note that those two switches have completey different configs, i.e. they are on a different transport zone, have different host_switch_ids and everything.

Deploying the test edge with just either one of the standard_host_switch configs works find and we end up with a new edge and a switch called "nsxDefaultHostSwitch".

However, if we use both standard_host_switch at the same time, we get the following API error:

Failed to create Transport node edge-test: Multiple HostSwitches were found with the same HostSwitch nsxDefaultHostSwitch in TransportNode.

If we do this through the API or in the GUI, we can specify a distinct name for each switch, obviously some default is used here which leads to the clash.

Can you expose a property to actually set a distinct name for each switch (while keeping the default as it is)?

Reproduction steps

1. Create a nsxt_edge_transport_node with multiple standard_host_switch blocks

Expected behavior

Edge node is created with two switches and no error.

Additional context

No response

[annakhm]

Hi @martinrohrbach, thanks for filing this, indeed there is host_switch_name attribute which we didn't expose since its deprecated on nsx.
Can you please provide your nsx version?

[martinrohrbach]

Hi @annakhm, good point, I meant to include that info, sorry.

We're testing against NSX 4.1.1.0 in our test environment.

[liftconfig]

Also facing this issue particularly when deploying v2t edge nodes which require a switch assigned to an overlay TZ, and a second switch assigned to a VLAN TZ. Another concern is IP assignment is always a required field in the host switch block. If the switch is only assigned to a VLAN TZ, there are no IPs required as there aren't any TEPs.