Skip to content

Latest commit

 

History

History
118 lines (79 loc) · 5.12 KB

provider_networks.md

File metadata and controls

118 lines (79 loc) · 5.12 KB
Raw

Provider Networks

Table of Contents

Introduction

Provider networks map directly to an existing physical network in a data center. Example of network types include flat (untagged), VLAN (802.1Q tagged) and VXLAN. OpenShift clusters that are installed on provider networks do not require tenant networks or floating IP addresses (FIPs). Therefore, the OpenShift installer does not create these resources during installation. More information can be found about provider networks terminology here.

Here is a basic architecture of one OCP cluster running on a provider network and another one on a tenant network:

OCP on a provider network

Prerequisites

Considerations when creating provider networks

  • The provider network has to be shared with other tenants, otherwise Nova won't be able to request ports on that external network. For more details, see BZ#1933047.

      openstack network create --share (...)

    To secure that network, it is advised to create RBAC rules so the network can be only usable by a specific project.

  • The project that you use to install OpenShift must own the provider network.

    The provider network and the subnet must be owned by the project that is used to install OpenShift instead of admin. If they are not, you will have to run the installer from the admin user to create ports on the network.

    It is important that the provider network and the subnet are owned by the same project that will be used to install OpenShift (from the clouds.yaml) and we don't want them to be owned by admin otherwise it'll cause Terraform to fail creating the ports.

    Example commands to create a network and subnet for a project that is named openshift:

      openstack network create --project openshift (...)
  openstack subnet create --project openshift (...)

    More information can be found about how to create provider networks here.

  • You'll have to make sure that the provider network can reach the Metadata IP (169.254.169.254) which, depending on the OpenStack SDN and how Neutron is configured (e.g. DHCP servers provide metadata network routes) might involve to provide the route when creating the subnet:

    openstack subnet create --dhcp --host-route destination=169.254.169.254/32,gateway=$ROUTER_IP" (...)

Note We're working on removing the nova-metadata requirement but for now it is mandatory and must be reachable from the provider network.

Deploying cluster with primary interface on a provider network with IPI

  • Considerations: make sure all prerequisites documented previously have been met.

  • Create install-config.yaml:

    • Set platform.openstack.apiVIP to the IP address for the API VIP.
    • Set platform.openstack.ingressVIP to the IP address for the Ingress VIP.
    • Set platform.openstack.controlPlanePort.fixedIPs.subnet.id to the subnet ID of the provider network subnet and/or platform.openstack.controlPlanePort.fixedIPs.subnet.name to the name of the provider network.
    • Set networking.machineNetwork.cidr to the CIDR of the provider network subnet.

    Note platform.openstack.apiVIP and platform.openstack.ingressVIP both need to be an unassigned IP address on the networking.machineNetwork.cidr.

    Example:

      (...)
  platform:
    openstack:
      apiVIP: <IP address in the provider network reserved for the API VIP>
      ingressVIP: <IP address in the provider network reserved for the Ingress VIP>
      controlPlanePort:
        fixedIPs:
          - subnet:
            id: <provider network subnet ID>
      (...)
  networking:
    machineNetwork:
    - cidr: <provider network subnet CIDR>

  • Run the OpenShift installer:

    ./openshift-install create cluster --log-level debug

  • Wait for the installer to complete.