Multi-tenancy in the Bare Metal service

Multi-tenancy in the Bare Metal service


It is possible to use dedicated tenant networks for provisioned nodes, which extends the current Bare Metal service capabilities of providing flat networks. This works in conjunction with the Networking service to allow provisioning of nodes in a separate provisioning network. The result of this is that multiple tenants can use nodes in an isolated fashion. However, this configuration does not support trunk ports belonging to multiple networks.


Network interfaces

Network interface is one of the driver interfaces that manages network switching for nodes. There are 3 network interfaces available in the Bare Metal service:

  • noop interface is used for standalone deployments, and does not perform any network switching;
  • flat interface places all nodes into a single provider network that is pre-configured on the Networking service and physical equipment. Nodes remain physically connected to this network during their entire life cycle.
  • neutron interface provides tenant-defined networking through the Networking service, separating tenant networks from each other and from the provisioning and cleaning provider networks. Nodes will move between these networks during their life cycle. This interface requires Networking service support for the switches attached to the baremetal servers so they can be programmed.

Physical networks

A Bare Metal port may be associated with a physical network using its physical_network field. The Bare Metal service uses this information when mapping between virtual ports in the Networking service and physical ports and port groups in the Bare Metal service. A port’s physical network field is optional, and if not set then any virtual port may be mapped to that port, provided that no free Bare Metal port with a suitable physical network assignment exists.

The physical network of a port group is defined by the physical network of its constituent ports. The Bare Metal service ensures that all ports in a port group have the same value in their physical network field.

When attaching a virtual interface (VIF) to a node, the following ordered criteria are used to select a suitable unattached port or port group:

  • Require ports or port groups to not have a physical network or to have a physical network that matches one of the VIF’s allowed physical networks.
  • Prefer ports and port groups that have a physical network to ports and port groups that do not have a physical network.
  • Prefer port groups to ports. Prefer ports with PXE enabled.

Configuring the Bare Metal service

See the Configure tenant networks section in the installation guide for the Bare Metal service.

Configuring nodes

  1. Ensure that your python-ironicclient version and requested API version are sufficient for your requirements.

    • Multi-tenancy support was added in API version 1.20, and is supported by python-ironicclient version 1.5.0 or higher.
    • Physical network support for ironic ports was added in API version 1.34, and is supported by python-ironicclient version 1.15.0 or higher.

    The following examples assume you are using python-ironicclient version 1.15.0 or higher.

    Export the following variable:

    export OS_BAREMETAL_API_VERSION=<API version>
  2. The node’s network_interface field should be set to a valid network interface. Valid interfaces are listed in the [DEFAULT]/enabled_network_interfaces configuration option in the ironic-conductor’s configuration file. Set it to neutron to use the Networking service’s ML2 driver:

    openstack baremetal node create --network-interface neutron --driver ipmi


    If the [DEFAULT]/default_network_interface configuration option is set, the --network-interface option does not need to be specified when creating the node.

  3. To update an existing node’s network interface to neutron, use the following commands:

    openstack baremetal node set $NODE_UUID_OR_NAME \
        --network-interface neutron
  4. Create a port as follows:

    openstack baremetal port create $HW_MAC_ADDRESS --node $NODE_UUID \
        --local-link-connection switch_id=$SWITCH_MAC_ADDRESS \
        --local-link-connection switch_info=$SWITCH_HOSTNAME \
        --local-link-connection port_id=$SWITCH_PORT \
        --pxe-enabled true \
        --physical-network physnet1
  5. Check the port configuration:

    openstack baremetal port show $PORT_UUID

After these steps, the provisioning of the created node will happen in the provisioning network, and then the node will be moved to the tenant network that was requested.

Configuring the Networking service

In addition to configuring the Bare Metal service some additional configuration of the Networking service is required to ensure ports for bare metal servers are correctly programmed. This configuration will be determined by the Bare Metal service network interfaces you have enabled and which top of rack switches you have in your environment.

flat network interface

In order for Networking service ports to correctly operate with the Bare Metal service flat network interface the baremetal ML2 mechanism driver from networking-baremetal needs to be loaded into the Networking service configuration. This driver understands that the switch should be already configured by the admin, and will mark the networking service ports as successfully bound as nothing else needs to be done.

  1. Install the networking-baremetal library

    $ pip install networking-baremetal
  2. Enable the baremetal driver in the Networking service ML2 configuration file

    mechanism_drivers = ovs,baremetal

neutron network interface

The neutron network interface allows the Networking service to program the physical top of rack switches for the bare metal servers. To do this an ML2 mechanism driver which supports the baremetal VNIC type for the make and model of top of rack switch in the environment must be installed and enabled.

This is a list of known top of rack ML2 mechanism drivers which work with the neutron network interface:

Cisco Nexus 9000 series
To install and configure this ML2 mechanism driver see Nexus Mechanism Driver Installation Guide.
networking-fujitsu ML2 driver supports this switch. The documentation is available here.
Networking Generic Switch
This is an ML2 mechanism driver built for testing against virtual bare metal environments and some switches that are not covered by hardware specific ML2 mechanism drivers. More information is available in the project’s README.
Creative Commons Attribution 3.0 License

Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.