Skip to content

Latest commit

 

History

History
281 lines (220 loc) · 15 KB

customization.md

File metadata and controls

281 lines (220 loc) · 15 KB
Raw

OpenStack Platform Customization

Beyond the platform-agnostic install-config.yaml properties, the installer supports additional, OpenStack-specific properties.

Table of Contents

Cluster-scoped properties

  • cloud (required string): The name of the OpenStack cloud to use from clouds.yaml.
  • computeFlavor (deprecated string): The OpenStack flavor to use for compute and control-plane machines.
  • externalDNS (optional list of strings): The IP addresses of DNS servers to be used for the DNS resolution of all instances in the cluster. The total number of dns servers supported by an instance is three. That total includes any dns server provided by the underlying openstack infrastructure.
  • externalNetwork (optional string): Name of external network the installer will use to provide access to the cluster. If defined, a floating IP from this network will be created and associated with the bootstrap node to facilitate debugging and connection to the bootstrap node during installation. The apiFloatingIP property is a floating IP address selected from this network.
  • apiFloatingIP (optional string): Address of existing Floating IP from externalNetwork the installer will associate with the OpenShift API. This property is only valid if externalNetwork is defined. If externalNetwork is not defined, the installer will throw an error.
  • ingressFloatingIP (optional string): Address of an existing Floating IP from externalNetwork the installer will associate with the ingress port. This property is only valid if externalNetwork is defined. If externalNetwork is not defined, the installer will throw an error.
  • octaviaSupport (deprecated string): Whether OpenStack supports Octavia (1 for true or 0 for false)
  • region (deprecated string): The OpenStack region where the cluster will be created. Currently this value is not used by the installer.
  • trunkSupport (deprecated string): Whether OpenStack ports can be trunked (1 for true or 0 for false)
  • clusterOSImage (optional string): Either a URL with http(s) or file scheme to override the default OS image for cluster nodes or an existing Glance image name.
  • clusterOSImageProperties (optional list of strings): a list of properties to be added to the installer-uploaded ClusterOSImage in Glance. The default is to not set any properties. clusterOSImageProperties is ignored when clusterOSImage points to an existing image in Glance.
  • apiVIP (optional string): An IP address on the machineNetwork that will be assigned to the API VIP. Be aware that the 10 and 11 of the machineNetwork will be taken by neutron dhcp by default, and wont be available.
  • ingressVIP (optional string): An IP address on the machineNetwork that will be assigned to the ingress VIP. Be aware that the 10 and 11 of the machineNetwork will be taken by neutron dhcp by default, and wont be available.
  • controlPlanePort (optional object): the UUID and/or Name of an OpenStack Network and its Subnets where to install the nodes of the cluster onto. For more information on how to install with a custom subnet, see the custom subnets section of the docs.
  • defaultMachinePlatform (optional object): Default OpenStack-specific machine pool properties which apply to machine pools that do not define their own OpenStack-specific properties.

Machine pools

  • additionalNetworkIDs (optional list of strings): IDs of additional networks for machines.
  • additionalSecurityGroupIDs (optional list of strings): IDs of additional security groups for machines.
  • serverGroupPolicy (optional string): Server group policy to apply to the group that will contain the machines in the pool. Defaults to "soft-anti-affinity". Allowed values are "affinity", "soft-affinity", "anti-affinity", "soft-anti-affinity".
    • It is not possible to change a server group policy or a server's affiliation to a group after creation
    • A strict "affinity" policy prevents migrations, and therefore affects OpenStack upgrades
    • An additional OpenStack host is needed when migrating instances with a strict "anti-affinity" policy
  • type (optional string): The OpenStack flavor name for machines in the pool.
  • rootVolume (optional object): Defines the root volume for instances in the machine pool. The instances use ephemeral disks if not set.
    • size (required integer): Size of the root volume in GB. Must be set to at least 25. For production clusters, this must be at least 100.
    • type (deprectated string): The volume pool to create the volume from. It was replaced by types.
    • types (required list of strings): The volume pool to create the volume from. If compute zones are defined with more than one type, the number of zones must match the number of types.
    • zones (optional list of strings): The names of the availability zones you want to install your root volumes on. If unset, the installer will use your default volume zone. If compute zones contains at least one value, rootVolume.zones must also contain at least one value. Indeed, when a machine is created with a compute availability zone and a storage root volume with no specified rootVolume.availabilityZone, CAPO will use the compute AZ for the volume AZ. This can be problematic if the AZ doesn't exist in Cinder, therefore we enforce that rootVolume.zones to be set if zones is set.
  • zones (optional list of strings): The names of the availability zones you want to install your nodes on. If unset, the installer will use your default compute zone.

Note The bootstrap node follows the type, rootVolume, additionalNetworkIDs, and additionalSecurityGroupIDs parameters from the controlPlane machine pool.

Note Note when deploying the control-plane machines with rootVolume, it is highly suggested to use an additional ephemeral disk dedicated to etcd.

Examples

Some example install-config.yaml are shown below. For examples of platform-agnostic configuration fragments, see here.

Minimal

An example minimal OpenStack install config is:

apiVersion: v1
baseDomain: example.com
metadata:
  name: test-cluster
platform:
  openstack:
    apiFloatingIP: 128.0.0.1
    cloud: mycloud
    defaultMachinePlatform:
      type: m1.s2.xlarge
    externalNetwork: external
    externalDNS:
      - "8.8.8.8"
      - "192.168.1.12"
pullSecret: '{"auths": ...}'
sshKey: ssh-ed25519 AAAA...

Custom machine pools

An example OpenStack install config with custom machine pools:

apiVersion: v1
baseDomain: example.com
controlPlane:
  name: master
  replicas: 3
compute:
- name: worker
  platform:
    openstack:
      type: ml.large
      rootVolume:
        size: 30
        types:
        - performance
  replicas: 3
metadata:
  name: test-cluster
platform:
  openstack:
    apiFloatingIP: 128.0.0.1
    cloud: mycloud
    defaultMachinePlatform:
      type: m1.s2.xlarge
    externalNetwork: external
pullSecret: '{"auths": ...}'
sshKey: ssh-ed25519 AAAA...

Image Overrides

The OpenShift installer pins the version of RHEL CoreOS and normally handles uploading the image to the target OpenStack instance.

If you want to download the image manually, see CoreOS bootimages for more information about bootimages. This is useful, for example, to perform a disconnected installation. To do this, download the qcow2 and host it at a custom location. Then set the openstack.clusterOSImage parameter field in the install config to point to that location. The install process will then use that mirrored image. In all other respects the process will be consistent with the default.

Note For this to work, the parameter value must be a valid http(s) URL.

Note The optional sha256 query parameter can be attached to the URL. This will force the installer to check the uncompressed image file checksum before uploading it into Glance.

Example:

platform:
  openstack:
      clusterOSImage: http://mirror.example.com/images/rhcos-43.81.201912131630.0-openstack.x86_64.qcow2.gz?sha256=ffebbd68e8a1f2a245ca19522c16c86f67f9ac8e4e0c1f0a812b068b16f7265d

If the user wants to upload the image from the local file system, he can set clusterOSImage as file:///path/to/file. In this case the installer will take this file and automatically create an image in Glance.

Example:

platform:
  openstack:
      clusterOSImage: file:///home/user/rhcos.qcow2

If the user wants to reuse an existing Glance image without any uploading of binary data, then it is possible to set clusterOSImage install config parameter that specifies the Glance image name. In this case no new Glance images will be created, and the image will stay when the cluster is destroyed. In other words, if clusterOSImage is not an "http(s)" or "file" URL, then the installer will look into Glance for an image with that name.

Example:

platform:
  openstack:
      clusterOSImage: my-rhcos

Custom Subnets

In the install-config.yaml file, the value of the controlPlanePort property contains the Name and/or UUID of the network and subnet(s) where the Kubernetes endpoints of the nodes in your cluster are published. The Ingress and API ports are created on the subnets, too. By default, the installer creates a network and subnet for these endpoints and ports. Alternatively, you can use a OpenStack network containing one subnet or two, in case of dual-stack, of your own by specifying their and/or in the controlPlanePort property. To use this feature, you need to meet these requirements:

  • Any subnet used by controlPlanePort have DHCP enabled.
  • The CIDR of any subnet listed in controlPlanePort.fixedIPs matches the CIDRs listed on networks.machineNetwork.
  • When using dual-stack Network the api and ingress Ports needs to be pre-created by the user. Also, the installer user must have permission to add tags and security groups to those pre-created Ports. The value of the fixed IPs of the Ports needs to be specified at the apiVIPs and ingressVIPs options in the install-config.yaml.
  • If not using dual-stack, the installer user must have permission to create ports on this network, including ports with fixed IP addresses.

You should also be aware of the following limitations:

  • If you plan to install a cluster that uses floating IPs, the controlPlanePort must be attached to a router that is connected to the externalNetwork.
  • The installer will not create a private network or subnet for your OpenShift machines if the controlPlanePort is set in the install-config.yaml.
  • By default when not using dual-stack, the API and Ingress VIPs use the .5 and .7 of your network CIDR. To prevent other services from taking the ports that are assigned to the API and Ingress VIPs, set the apiVIP and ingressVIP options in the install-config.yaml to addresses that are outside of the DHCP allocation pool.
  • You cannot use the externalDNS property at the same time as a custom controlPlanePort. If you want to add a DNS to your cluster while using a custom subnet, add it to the subnet in OpenStack.

Additional Networks

You can set additional networks for your machines by defining additionalNetworkIDs parameter in the machine configuration. The parameter is a list of strings with additional network IDs:

additionalNetworkIDs:
- <network1_uuid>
- <network2_uuid>

You can attach this parameter for both controlPlane and compute machines:

Example:

compute:
- name: worker
  platform:
    openstack:
      additionalNetworkIDs:
      - fa806b2f-ac49-4bce-b9db-124bc64209bf
controlPlane:
  name: master
  platform:
    openstack:
      additionalNetworkIDs:
      - fa806b2f-ac49-4bce-b9db-124bc64209bf

Note Allowed address pairs won't be created for the additional networks.

Note The additional networks attached to the Control Plane machine will also be attached to the bootstrap node.

Additional Security Groups

You can set additional security groups for your machines by defining additionalSecurityGroupIDs parameter in the machine configuration. The parameter is a list of strings with additional security group IDs:

additionalSecurityGroupIDs:
- <security_group1_id>
- <security_group2_id>

You can attach this parameter for both controlPlane and compute machines:

Example:

compute:
- name: worker
  platform:
    openstack:
      additionalSecurityGroupIDs:
      - 7ee219f3-d2e9-48a1-96c2-e7429f1b0da7
controlPlane:
  name: master
  platform:
    openstack:
      additionalSecurityGroupIDs:
      - 7ee219f3-d2e9-48a1-96c2-e7429f1b0da7

Note The additional security groups attached to the Control Plane machine will also be attached to the bootstrap node.

Cloud Provider configuration

You may want to modify cloud provider configuration in order to make it work with your OpenStack cloud. This is possible if you'll let the installer generate the manifests before running the installation:

openshift-install --dir <directory> create manifests

Then modify the manifest containing the cloud provider configuration:

vi <directory>/manifests/cloud-provider-config.yaml

As an example in order to tweak support for LoadBalancer Services you can modify options regarding Octavia configuration in the [LoadBalancer] section of config key. In particular:

  • enabled = false disables the Octavia integration.
  • lb-provider = <"amphora" or "ovn"> lets you choose the Octavia provider to use when creating load balancers. Please note that setting "ovn" requires setting lb-method = SOURCE_IP_PORT as this is the only method supported by OVN.
  • floating-network-id = <uuid> is required to be set if your OpenStack cluster has multiple external networks. The network set here will be used by cloud provider to create floating IPs on.

After saving the file you can continue the installation normally:

openshift-install --dir <directory> create cluster

Further customization

For customizing the installation beyond what is possible with openshift-install, refer to the UPI (User Provided Infrastructure) documentation.