Scenario - Open Virtual Network (OVN)

Overview

Operators can choose to utilize the Open Virtual Network (OVN) mechanism driver (ML2/OVN) instead of ML2/LXB or ML2/OVS. This offers the possibility of deploying virtual networks and routers using OVN with Open vSwitch, which replaces the agent-based models used by the legacy ML2/LXB and ML2/OVS architectures. This document outlines how to set it up in your environment.

Prerequisites

  • Open vSwitch >= 2.17.0

Inventory Architecture

While OVN itself supports many different configurations, Neutron and networking-ovn leverage specific functionality to provide virtual routing capabilities to an OpenStack-based Cloud.

OpenStack-Ansible separates OVN-related services and functions into three groups:

  • neutron_ovn_northd

  • neutron_ovn_controller

  • neutron_ovn_gateway

The neutron_ovn_northd group is used to specify which host(s) will contain the OVN northd daemon responsible for translating the high-level OVN configuration into logical configuration consumable by daemons such as ovn-controller. In addition, these nodes host the OVN Northbound and OVN Southbound databases the ovsdb-server services. Members of this group are typically the controller nodes hosting the Neutron APIs (neutron-server).

The neutron_ovn_controller group is used to specify which host(s) will run the local ovn-controller daemon for OVN, which registers the local chassis and VIFs to the OVN Southbound database and converts logical flows to physical flows on the local hypervisor node. Members of this group are typically the compute nodes hosting virtual machine instances (nova-compute).

The neutron_ovn_gateway group is used to specify which hosts are eligible to act as an OVN Gateway Chassis, which is a node running ovn-controller that is capable of providing external (north/south) connectivity to the tenant traffic. This is essentially a node(s) capable of hosting the logical router performing SNAT and DNAT (Floating IP) translations. East/West traffic flow is not limited to a gateway chassis and is performed between an OVN chassis nodes.

When planning out your architecture, it is important to determine early if you want to centralize OVN gateway chassis functions to a subset of nodes or across all compute nodes. Centralizing north/south routing to a set of dedicated network or gateway nodes is reminiscent of the legacy network node model. Enabling all compute nodes as gateway chassis will narrow the failure domain and potential bottlenecks at the cost of ensuring the computes can connect to the provider networks.

The following section will describe how to configure your inventory to meet certain deployment scenarios.

Deployment Scenarios

OpenStack-Ansible supports the following common deployment scenarios:

  • Collapsed Network/Gateway Nodes

  • Collapsed Compute/Gateway Nodes

  • Standalone Gateway Nodes

In an OpenStack-Ansible deployment, infrastructure hosts are intended to run OVN northd-related services, while compute hosts are intended to run OVN controller-related services.

In openstack_user_config.yml, specify the hosts or aliases that will run the ovn-northd service(s), like so:

network-northd_hosts:
  infra01:
    ip: 172.25.1.11
  infra02:
    ip: 172.25.1.12
  infra03:
    ip: 172.25.1.13

Alternatively, an alias can be specified, like so:

network-northd_hosts: *infrastructure_hosts

It is up to the deployer to dictate which nodes are considered “OVN Gateway Chassis” nodes by using the network-gateway_hosts inventory group in openstack_user_config.yml.

In openstack_user_config.yml, specify the hosts or aliases that will run the ovn-controller service and act as an OVN Gateway Chassis, like so:

network-gateway_hosts:
  network-node1:
    ip: 172.25.1.21
  network-node2:
    ip: 172.25.1.22
  network-node3:
    ip: 172.25.1.23

Existing inventory aliases can also be used. In the following example, members of the infrastructure_hosts group are also network hosts and will serve as OVN Gateway Chassis nodes:

network-gateway_hosts: *infrastructure_hosts

In the following example, members of the compute_hosts group running the ovn-controller service will also serve as OVN Gateway Chassis nodes:

network-gateway_hosts: *compute_hosts

Lastly, specific hosts can also be targeted:

network-gateway_hosts:
  compute5:
    ip: 172.25.1.55
  compute10:
    ip: 172.25.1.60
  compute15:
    ip: 172.25.1.65
  compute20:
    ip: 172.25.1.70
  compute25:
    ip: 172.25.1.75

OpenStack-Ansible user variables

To deploy OpenStack-Ansible using the ML2/OVN mechanism driver, set the following user variables in the``/etc/openstack_deploy/user_variables.yml`` file:

neutron_plugin_type: ml2.ovn

neutron_plugin_base:
  - ovn-router

neutron_ml2_drivers_type: "vlan,local,geneve,flat"

The overrides are instructing Ansible to deploy the OVN mechanism driver and associated OVN components. This is done by setting neutron_plugin_type to ml2.ovn.

The neutron_plugin_base override enables Neutron to use OVN for routing functions.

The neutron_ml2_drivers_type override provides support for all type drivers supported by OVN.

Provider network overrides can be specified on a global or per-host basis, and the following format can be used in user_variables.yml or per-host in openstack_user_config.yml or host vars.

Note

When network_interface_mappings are defined, the playbooks will attempt to connect the mapped interface to the respective OVS bridge. Omitting network_interface_mappings will require the operator to connect the interface to the bridge manually using the ovs-vsctl add-port command.

# When configuring Neutron to support geneve tenant networks and
# vlan provider networks the configuration may resemble the following:
neutron_provider_networks:
  network_types: "geneve"
  network_geneve_ranges: "1:1000"
  network_vlan_ranges: "public"
  network_mappings: "public:br-publicnet"
  network_interface_mappings: "br-publicnet:bond1"

# When configuring Neutron to support only vlan tenant networks and
# vlan provider networks the configuration may resemble the following:
neutron_provider_networks:
  network_types: "vlan"
  network_vlan_ranges: "public:203:203,467:500"
  network_mappings: "public:br-publicnet"
  network_interface_mappings: "br-publicnet:bond1"

# When configuring Neutron to support multiple vlan provider networks
# the configuration may resemble the following:
neutron_provider_networks:
  network_types: "vlan"
  network_vlan_ranges: "public:203:203,467:500,private:101:200,301:400"
  network_mappings: "public:br-publicnet,private:br-privatenet"
  network_interface_mappings: "br-publicnet:bond1,br-privatenet:bond2"

(Optional) DVR or Distributed L3 routing

DVR will be used for floating IPs if the ovn / enable_distributed_floating_ip flag is configured to True in the neutron server configuration.

Create a group var file for neutron server /etc/openstack_deploy/group_vars/neutron_server.yml. It has to include:

# DVR/Distributed L3 routing support
neutron_neutron_conf_overrides:
  ovn:
    enable_distributed_floating_ip: True

Useful Open Virtual Network (OVN) Commands

The following commands can be used to provide useful information about the state of Open vSwitch networking and configurations.

The following ad-hoc command can be executed to find the current state and the leader of the NB/SB database:

ansible neutron_ovn_northd -m command -a "ovs-appctl -t /var/run/ovn/ovnnb_db.ctl cluster/status OVN_Northbound"
ansible neutron_ovn_northd -m command -a "ovs-appctl -t /var/run/ovn/ovnsb_db.ctl cluster/status OVN_Southbound"

The ovs-vsctl list open_vswitch command provides information about the open_vswitch table in the local Open vSwitch database and can be run from any network or compute host:

root@mnaio-controller1:~# ovs-vsctl list open_vswitch
_uuid               : 7f96baf2-d75e-4a99-bb19-ca7138fc14c2
bridges             : []
cur_cfg             : 1
datapath_types      : [netdev, system]
datapaths           : {}
db_version          : "8.3.0"
dpdk_initialized    : false
dpdk_version        : none
external_ids        : {hostname=mnaio-controller1, rundir="/var/run/openvswitch", system-id="a67926f2-9543-419a-903d-23e2aa308368"}
iface_types         : [bareudp, erspan, geneve, gre, gtpu, internal, ip6erspan, ip6gre, lisp, patch, stt, system, tap, vxlan]
manager_options     : []
next_cfg            : 1
other_config        : {}
ovs_version         : "2.17.2"
ssl                 : []
statistics          : {}
system_type         : ubuntu
system_version      : "20.04"

If you want to check only for only a specific field from the ovs-vsctl output, like applied interface mappings, you can select it in the following way:

root@mnaio-controller1:~# ovs-vsctl get open . external_ids:ovn-bridge-mappings
"vlan:br-provider"

You can also get information about the agent UUID which will be stated in openstack network agent list output via similar command:

root@mnaio-controller1:~# ovs-vsctl get open . external_ids:system-id
"a67926f2-9543-419a-903d-23e2aa308368"

Note

Commands towards OVN Southbound and Northbound databases are expected to be run from neutron_ovn_northd hosts. OpenStack-Ansible places an openrc file named /root/ovnctl.rc on these hosts. Once you source that file, required environment variables will be set to connect to the database. Alternatively, you can use --no-leader-only flag to connect to the local database only instead of the leader one (which is default).

The ovn-sbctl show command provides information related to southbound connections. If used outside the ovn_northd container, specify the connection details:

root@mnaio-controller1:~# ovn-sbctl show
Chassis "5335c34d-9233-47bd-92f1-fc7503270783"
    hostname: mnaio-compute1
    Encap geneve
        ip: "172.25.1.31"
        options: {csum="true"}
    Encap vxlan
        ip: "172.25.1.31"
        options: {csum="true"}
    Port_Binding "852530b5-1247-4ec2-9c39-8ae0752d2144"
Chassis "ff66288c-5a7c-41fb-ba54-6c781f95a81e"
    hostname: mnaio-compute2
    Encap vxlan
        ip: "172.25.1.32"
        options: {csum="true"}
    Encap geneve
        ip: "172.25.1.32"
        options: {csum="true"}
Chassis "cb6761f4-c14c-41f8-9654-16f3fc7cc7e6"
    hostname: mnaio-compute3
    Encap geneve
        ip: "172.25.1.33"
        options: {csum="true"}
    Encap vxlan
        ip: "172.25.1.33"
        options: {csum="true"}
    Port_Binding cr-lrp-022933b6-fb12-4f40-897f-745761f03186

You can get specific information about chassis by providing either it’s name, where name is UUID of the agent (external_ids:system-id from the ovs-vsctl output), for example:

root@mnaio-controller1:~# ovn-sbctl list Chassis ff66288c-5a7c-41fb-ba54-6c781f95a81e
_uuid               : b0b6ebec-1c64-417a-adb7-d383632a4c5e
encaps              : [a3ba78c3-df14-4144-81e0-e6379541bc89]
external_ids        : {}
hostname            : mnaio-compute2
name                : "ff66288c-5a7c-41fb-ba54-6c781f95a81e"
nb_cfg              : 0
other_config        : {ct-no-masked-label="true", datapath-type=system, fdb-timestamp="true", iface-types="afxdp,afxdp-nonpmd,bareudp,erspan,geneve,gre,gtpu,internal,ip6erspan,ip6gre,lisp,patch,srv6,stt,system,tap,vxlan", is-interconn="false", mac-binding-timestamp="true", ovn-bridge-mappings="vlan:br-provider", ovn-chassis-mac-mappings="", ovn-cms-options="", ovn-ct-lb-related="true", ovn-enable-lflow-cache="true", ovn-limit-lflow-cache="", ovn-memlimit-lflow-cache-kb="", ovn-monitor-all="false", ovn-trim-limit-lflow-cache="", ovn-trim-timeout-ms="", ovn-trim-wmark-perc-lflow-cache="", port-up-notif="true"}
transport_zones     : []
vtep_logical_switches: []

As you might see, other_config row also contains bridge-mapping, which can be fetched from the table similarly to the ovs-vsctl way:

root@mnaio-controller1:~# ovn-sbctl get Chassis ff66288c-5a7c-41fb-ba54-6c781f95a81e other_config:ovn-bridge-mappings
"vlan:br-provider"

The ovn-nbctl show command provides information about networks, ports, and other objects known to OVN and demonstrates connectivity between the northbound database and neutron-server.

root@mnaio-controller1:~# ovn-nbctl show
switch 03dc4558-f83e-4531-b854-156292f1dbad (neutron-a6e65821-93e2-4521-9e31-37c35d52d953) (aka project-tenant-network)
    port 852530b5-1247-4ec2-9c39-8ae0752d2144
        addresses: ["fa:16:3e:d2:af:bf 10.3.3.49"]
    port 624de478-7e75-472f-b867-e6f514790a81
        addresses: ["fa:16:3e:bf:c0:c3 10.3.3.3", "unknown"]
    port 1cca8ef3-d3c9-4307-a779-13348db5e647
        addresses: ["fa:16:3e:4a:67:ed 10.3.3.4", "unknown"]
    port 05e20b32-2933-414a-ba31-eac683d09ac2
        addresses: ["fa:16:3e:bd:5d:e8 10.3.3.5", "unknown"]
    port 5a2e35cb-178b-443b-9f15-4c6ec4db4ac7
        type: router
        router-port: lrp-5a2e35cb-178b-443b-9f15-4c6ec4db4ac7
    port 2d52a2bf-ab37-4a18-87bd-8808a99c67d3
        type: localport
        addresses: ["fa:16:3e:30:b4:a0 10.3.3.2"]
switch 3e03d5f1-4cfe-4c61-bd4c-8a661634d77b (neutron-b0b4017f-a9d1-4923-af35-944b88b7a393) (aka flat-external-provider-network)
    port 022933b6-fb12-4f40-897f-745761f03186
        type: router
        router-port: lrp-022933b6-fb12-4f40-897f-745761f03186
    port 347a7d8d-fd0f-48be-be02-d603258f0a08
        addresses: ["fa:16:3e:f4:6a:17 192.168.25.5", "unknown"]
    port 29c83838-329d-4839-bddb-818c7e2e9bc7
        addresses: ["fa:16:3e:a3:48:a8 192.168.25.3", "unknown"]
    port 173c9ceb-4dd3-4268-aaa3-c7b0f693a557
        type: localport
        addresses: ["fa:16:3e:0c:37:ed 192.168.25.2"]
    port 7a0175fd-ac09-4466-b3d0-26f696e3769c
        addresses: ["fa:16:3e:ad:19:c2 192.168.25.4", "unknown"]
    port provnet-525d3402-d582-49b4-b946-f28de8bbc615
        type: localnet
        addresses: ["unknown"]
router 5ebb0cdb-2026-4454-a32e-eb5425ae7296 (neutron-b0d6ca32-fda3-4fdc-b648-82c8bee303dc) (aka project-router)
    port lrp-5a2e35cb-178b-443b-9f15-4c6ec4db4ac7
        mac: "fa:16:3e:3a:1c:bb"
        networks: ["10.3.3.1/24"]
    port lrp-022933b6-fb12-4f40-897f-745761f03186
        mac: "fa:16:3e:1f:cd:e9"
        networks: ["192.168.25.242/24"]
        gateway chassis: [cb6761f4-c14c-41f8-9654-16f3fc7cc7e6 ff66288c-5a7c-41fb-ba54-6c781f95a81e 5335c34d-9233-47bd-92f1-fc7503270783]
    nat 79d8486c-8b5e-4d6c-a56f-9f0df115f77f
        external ip: "192.168.25.242"
        logical ip: "10.3.3.0/24"
        type: "snat"
    nat d338ccdf-d3c4-404e-b2a5-938d0c212e0d
        external ip: "192.168.25.246"
        logical ip: "10.3.3.49"
        type: "dnat_and_snat"

Floating IPs and Router SNAT are represented via NAT rules in the NB database, where FIP has type dnat_and_snat. You can fetch the list of NAT rules assigned to a specific router using the router name in the OVN database, which is formatted like neutron-<UUID>, where UUID is the UUID of the router in Neutron database. Command will look like this:

Note

Keep in mind, that GATEWAY_PORT will not be defined for dnat_and_snat rule when the external network of the router is on a geneve network and the router is bound to the chassis instead of it’s external port.

root@mnaio-controller1:~# ovn-nbctl lr-nat-list neutron-b0d6ca32-fda3-4fdc-b648-82c8bee303dc
TYPE             GATEWAY_PORT          EXTERNAL_IP        EXTERNAL_PORT    LOGICAL_IP          EXTERNAL_MAC         LOGICAL_PORT
dnat_and_snat    lrp-16555e74-fbef-    192.168.25.246                      10.3.3.49
snat                                   192.168.25.242                      10.3.3.0/24

The mapping/location of the router to the gateway node can be established via logical ports of the router when the external network to which router is connected happens to have either VLAN or FLAT type. For that you need to know the UUID of the external port attached to the router. The port name in the OVN database is constructed as lrp-<UUID>, where UUID is the Neutron port UUID. Given that an external network in the topic is named public, you can determine the gateway node in a following way:

root@mnaio-controller1:~# openstack port list --router b0d6ca32-fda3-4fdc-b648-82c8bee303dc --network public -c ID
+--------------------------------------+
| ID                                   |
+--------------------------------------+
| 16555e74-fbef-4ecb-918c-2fb76bf5d42d |
+--------------------------------------+
root@mnaio-controller1:~# ovn-nbctl get Logical_Router_Port lrp-16555e74-fbef-4ecb-918c-2fb76bf5d42d status:hosting-chassis
"5335c34d-9233-47bd-92f1-fc7503270783"
root@mnaio-controller1:~# ovn-sbctl get Chassis 5335c34d-9233-47bd-92f1-fc7503270783 hostname
mnaio-compute1
root@mnaio-controller1:~# openstack network agent show 5335c34d-9233-47bd-92f1-fc7503270783 -c host
+-------+-----------------------------+
| Field | Value                       |
+-------+-----------------------------+
| host  | mnaio-compute1              |
+-------+-----------------------------+

To list all gateway chassis on which the logical port is scheduled with their priorities, you can use:

root@mnaio-controller1:~# ovn-nbctl lrp-get-gateway-chassis lrp-16555e74-fbef-4ecb-918c-2fb76bf5d42d | cut -d '_' -f 2
5335c34d-9233-47bd-92f1-fc7503270783     2
cb6761f4-c14c-41f8-9654-16f3fc7cc7e6     1

In order to migrate active router logical port to another node, you can execute the following command:

root@mnaio-controller1:~# ovn-nbctl lrp-set-gateway-chassis lrp-16555e74-fbef-4ecb-918c-2fb76bf5d42d ff66288c-5a7c-41fb-ba54-6c781f95a81e 10

Additional commands can be found in upstream OVN documentation and other resources listed on this page.

In cases when a Geneve network acts as the external network for the router, Logical Router will be pinned to the chassis instead of it’s LRP:

# ovn-nbctl --no-leader-only get Logical_Router neutron-b0d6ca32-fda3-4fdc-b648-82c8bee303dc options
{always_learn_from_arp_request="false", chassis="5335c34d-9233-47bd-92f1-fc7503270783", dynamic_neigh_routers="true", mac_binding_age_threshold="0"}

All LRPs of such router will remain unbinded.

OVN database population

In case of OVN DB clustering failure and data loss as a result, you can always re-populate data in OVN SB/NB from stored state in Neutron database.

For that, you can execute the following:

root@mnaio-controller1:~# lxc-attach -n $(lxc-ls -1 | grep neutron-server)
root@mnaio-controller1-neutron-server-container-7510c8bf:/# source /etc/openstack-release
root@mnaio-controller1-neutron-server-container-7510c8bf:/# /openstack/venvs/neutron-${DISTRIB_RELEASE}/bin/neutron-ovn-db-sync-util --config-file /etc/neutron/neutron.conf --config-file /etc/neutron/plugins/ml2/ml2_conf.ini --ovn-neutron_sync_mode repair

Command neutron-ovn-db-sync-util is also used during migration from OVS to OVN. For that, you need to supply --ovn-neutron_sync_mode migrate instead of repair as shown in example above.

Notes

The ovn-controller service will check in as an agent and can be observed using the openstack network agent list command:

+--------------------------------------+------------------------------+-------------------+-------------------+-------+-------+----------------------------+
| ID                                   | Agent Type                   | Host              | Availability Zone | Alive | State | Binary                     |
+--------------------------------------+------------------------------+-------------------+-------------------+-------+-------+----------------------------+
| 5335c34d-9233-47bd-92f1-fc7503270783 | OVN Controller Gateway agent | mnaio-compute1    |                   | :-)   | UP    | ovn-controller             |
| ff66288c-5a7c-41fb-ba54-6c781f95a81e | OVN Controller Gateway agent | mnaio-compute2    |                   | :-)   | UP    | ovn-controller             |
| cb6761f4-c14c-41f8-9654-16f3fc7cc7e6 | OVN Controller Gateway agent | mnaio-compute3    |                   | :-)   | UP    | ovn-controller             |
| 38206799-af64-589b-81b2-405f0cfcd198 | OVN Metadata agent           | mnaio-compute1    |                   | :-)   | UP    | neutron-ovn-metadata-agent |
| 9e9b49c7-dd00-5f58-a3f5-22dd01f562c4 | OVN Metadata agent           | mnaio-compute2    |                   | :-)   | UP    | neutron-ovn-metadata-agent |
| 72b1a6e2-4cca-570f-83a4-c05dcbbcc11f | OVN Metadata agent           | mnaio-compute3    |                   | :-)   | UP    | neutron-ovn-metadata-agent |
+--------------------------------------+------------------------------+-------------------+-------------------+-------+-------+----------------------------+