Expose trunk details in metadata and configdrive

[Baodong (Robert) Li]

Vlan trunking has been supported since Mitaka. Refer to https://wiki.openstack.org/wiki/Neutron/TrunkPort. However, for guest instances with trunk ports, vlan subinterfaces are not automatically configured. This bug attempts to address this by exposing trunk details in metadata and configdrive, with which cloud-init will configure vlan subinterfaces.

[Jay Pipes (jaypipes)]

Hi Robert,

How about simply extending the 'nic' device type in the existing device metadata?

http://specs.openstack.org/openstack/nova-specs/specs/ocata/approved/virt-device-tagged-attach-detach.html

Best,

-jay

[sean mooney (sean-k-mooney)]

while this would be nice to have, im not really sure its a bug since automatically creating subports in the guess was considered out of scope of this feature as it is not always wanted.

having the info available in the metadata is a good idea but creating subports automatically will break vnf that expect to process the vlans themselves and not expect vlan subports to be created.

as such i think the auto creation should not be enabled by default and only be done if the tenant expressly asks for it.

[Matt Riedemann (mriedem)]

This is not a bug. If you want to add this it will require a blueprint.

[Baodong (Robert) Li (baoli)]

@jaypipes:

Thanks for the link. I took a look at it, and it doesn't seem to me that it's intended for this purpose. To configure trunk port and its sub interfaces require clear identification of parent-subport relationship that tagging won't be able to adequately support.

A simple idea is to populate the network_data.json file in the configdrive based on the trunk_details in trunk ports. For example, the following network_data.json file contains one trunk port with three sub interfaces.

network_data.json file

services": [],
"networks": [
    {
        "network_id": "9f1c0636-ba1f-42fd-a741-a2d0194704b5",
        "link": "tapad7e2436-35",
        "type": "ipv4_dhcp",
        "id": "network0"
    },
    {
          "network_id": "ea077fad-e52a-4fa1-952f-06110e436d7d",
          "link": "tap143d8632-e3",
          "type": "ipv4_dhcp",
          "id": "network1"
    },
    {
        "network_id": "bbd6f034-a600-46c1-b22f-0502697e179e",
        "link": "tap4bc16b9c-e2",
        "type": "ipv4_dhcp",
        "id": "network2"
    },
    {
        "network_id": "0597cc43-7a2f-4ac0-847e-d9639d83fb6e",
        "link": "tap9189504d-42",
        "type": "ipv4_dhcp",
        "id": "network3"
    }
],
"links": [
    { // trunk
        "ethernet_mac_address": "fa:16:3e:54:2e:51",
        "mtu": 1500,
        "type": "ovs",
        "id": "tapad7e2436-35",
        "vif_id": "ad7e2436-3583-4ae3-844b-06e6b0adaf04"
    },
    { // sub interface
        "vif_id": "143d8632-e320-4fdb-aa07-bf2767085efd",
        "mtu": 1500,
        "vlan_link": "tapad7e2436-35",
        "vlan_id": 3279,
        "type": "vlan",
        "id": "tap143d8632-e3",
        "vlan_mac_address": "fa:16:3e:8e:b4:99"
    },
    { // sub interface
        "vif_id": "4bc16b9c-e2b6-4c09-ba98-e94160cce12d",
        "mtu": 1500,
        "vlan_link":"tapad7e2436-35",
        "vlan_id": 3274,
        "type": "vlan",
        "id": "tap4bc16b9c-e2",
        "vlan_mac_address": "fa:16:3e:a4:93:98"
    },
    { // sub interface
        "vif_id": "9189504d-4251-46ea-a676-2573b4e14580",
        "mtu": 1500,
        "vlan_link": "tapad7e2436-35",
        "vlan_id": 3272,
        "type": "vlan",
        "id": "tap9189504d-42",
        "vlan_mac_address": "fa:16:3e:72:d5:33"
    }
]
}

The network_data.json file is currently supported by cloud-init (and glean).

[Baodong (Robert) Li (baoli)]

@sean-k-mooney:

Thanks for the comment.

I was thinking about that. Shall a config option be introduced to enable/disable the auto generation of trunk sub interfaces in the config drive?

[Baodong (Robert) Li (baoli)]

@mriedem:

sure. I can add a blueprint.

[Baodong (Robert) Li (baoli)]

Added a blueprint and spec: https://blueprints.launchpad.net/nova/+spec/expose-vlan-trunking