1. Documentation
×

Installing a cluster on OpenStack with customizations

Prerequisites

Resource guidelines for installing OKD on OpenStack

To support an OKD installation, your OpenStack quota must meet the following requirements:

Table 1. Recommended resources for a default OKD cluster on OpenStack
Resource Value

Floating IP addresses

3

Ports

15

Routers

1

Subnets

1

RAM

88 GB

vCPUs

22

Volume storage

275 GB

Instances

7

Security groups

3

Security group rules

60

Server groups

2 - plus 1 for each additional availability zone in each machine pool

A cluster might function with fewer than recommended resources, but its performance is not guaranteed.

If OpenStack object storage (Swift) is available and operated by a user account with the swiftoperator role, it is used as the default backend for the OKD image registry. In this case, the volume storage requirement is 175 GB. Swift space requirements vary depending on the size of the image registry.
By default, your security group and security group rule quotas might be low. If you encounter problems, run openstack quota set --secgroups 3 --secgroup-rules 60 <project> as an administrator to increase them.

An OKD deployment comprises control plane machines, compute machines, and a bootstrap machine.

Control plane machines

By default, the OKD installation process creates three control plane machines.

Each machine requires:

  • An instance from the OpenStack quota

  • A port from the OpenStack quota

  • A flavor with at least 16 GB memory and 4 vCPUs

  • At least 100 GB storage space from the OpenStack quota

Compute machines

By default, the OKD installation process creates three compute machines.

Each machine requires:

  • An instance from the OpenStack quota

  • A port from the OpenStack quota

  • A flavor with at least 8 GB memory and 2 vCPUs

  • At least 100 GB storage space from the OpenStack quota

Compute machines host the applications that you run on OKD; aim to run as many as you can.

Bootstrap machine

During installation, a bootstrap machine is temporarily provisioned to stand up the control plane. After the production control plane is ready, the bootstrap machine is deprovisioned.

The bootstrap machine requires:

  • An instance from the OpenStack quota

  • A port from the OpenStack quota

  • A flavor with at least 16 GB memory and 4 vCPUs

  • At least 100 GB storage space from the OpenStack quota

Load balancing requirements for user-provisioned infrastructure

Before you install OKD, you can provision your own API and application ingress load balancing infrastructure to use in place of the default, internal load balancing solution. In production scenarios, you can deploy the API and application Ingress load balancers separately so that you can scale the load balancer infrastructure for each in isolation.

If you want to deploy the API and application Ingress load balancers with a Fedora instance, you must purchase the Fedora subscription separately.

The load balancing infrastructure must meet the following requirements:

  1. API load balancer: Provides a common endpoint for users, both human and machine, to interact with and configure the platform. Configure the following conditions:

    • Layer 4 load balancing only. This can be referred to as Raw TCP or SSL Passthrough mode.

    • A stateless load balancing algorithm. The options vary based on the load balancer implementation.

    Do not configure session persistence for an API load balancer. Configuring session persistence for a Kubernetes API server might cause performance issues from excess application traffic for your OKD cluster and the Kubernetes API that runs inside the cluster.

    Configure the following ports on both the front and back of the load balancers:

    Table 2. API load balancer
    Port Back-end machines (pool members) Internal External Description

    6443

    Bootstrap and control plane. You remove the bootstrap machine from the load balancer after the bootstrap machine initializes the cluster control plane. You must configure the /readyz endpoint for the API server health check probe.

    X

    X

    Kubernetes API server

    22623

    Bootstrap and control plane. You remove the bootstrap machine from the load balancer after the bootstrap machine initializes the cluster control plane.

    X

    Machine config server

    The load balancer must be configured to take a maximum of 30 seconds from the time the API server turns off the /readyz endpoint to the removal of the API server instance from the pool. Within the time frame after /readyz returns an error or becomes healthy, the endpoint must have been removed or added. Probing every 5 or 10 seconds, with two successful requests to become healthy and three to become unhealthy, are well-tested values.

  2. Application Ingress load balancer: Provides an ingress point for application traffic flowing in from outside the cluster. A working configuration for the Ingress router is required for an OKD cluster.

    Configure the following conditions:

    • Layer 4 load balancing only. This can be referred to as Raw TCP or SSL Passthrough mode.

    • A connection-based or session-based persistence is recommended, based on the options available and types of applications that will be hosted on the platform.

    If the true IP address of the client can be seen by the application Ingress load balancer, enabling source IP-based session persistence can improve performance for applications that use end-to-end TLS encryption.

    Configure the following ports on both the front and back of the load balancers:

    Table 3. Application Ingress load balancer
    Port Back-end machines (pool members) Internal External Description

    443

    The machines that run the Ingress Controller pods, compute, or worker, by default.

    X

    X

    HTTPS traffic

    80

    The machines that run the Ingress Controller pods, compute, or worker, by default.

    X

    X

    HTTP traffic

    If you are deploying a three-node cluster with zero compute nodes, the Ingress Controller pods run on the control plane nodes. In three-node cluster deployments, you must configure your application Ingress load balancer to route HTTP and HTTPS traffic to the control plane nodes.

Example load balancer configuration for clusters that are deployed with user-managed load balancers

This section provides an example API and application Ingress load balancer configuration that meets the load balancing requirements for clusters that are deployed with user-managed load balancers. The sample is an /etc/haproxy/haproxy.cfg configuration for an HAProxy load balancer. The example is not meant to provide advice for choosing one load balancing solution over another.

In the example, the same load balancer is used for the Kubernetes API and application ingress traffic. In production scenarios, you can deploy the API and application ingress load balancers separately so that you can scale the load balancer infrastructure for each in isolation.

If you are using HAProxy as a load balancer and SELinux is set to enforcing, you must ensure that the HAProxy service can bind to the configured TCP port by running setsebool -P haproxy_connect_any=1.
Sample API and application Ingress load balancer configuration 
global
  log         127.0.0.1 local2
  pidfile     /var/run/haproxy.pid
  maxconn     4000
  daemon
defaults
  mode                    http
  log                     global
  option                  dontlognull
  option http-server-close
  option                  redispatch
  retries                 3
  timeout http-request    10s
  timeout queue           1m
  timeout connect         10s
  timeout client          1m
  timeout server          1m
  timeout http-keep-alive 10s
  timeout check           10s
  maxconn                 3000
listen api-server-6443 (1)
  bind *:6443
  mode tcp
  option  httpchk GET /readyz HTTP/1.0
  option  log-health-checks
  balance roundrobin
  server bootstrap bootstrap.ocp4.example.com:6443 verify none check check-ssl inter 10s fall 2 rise 3 backup (2)
  server master0 master0.ocp4.example.com:6443 weight 1 verify none check check-ssl inter 10s fall 2 rise 3
  server master1 master1.ocp4.example.com:6443 weight 1 verify none check check-ssl inter 10s fall 2 rise 3
  server master2 master2.ocp4.example.com:6443 weight 1 verify none check check-ssl inter 10s fall 2 rise 3
listen machine-config-server-22623 (3)
  bind *:22623
  mode tcp
  server bootstrap bootstrap.ocp4.example.com:22623 check inter 1s backup (2)
  server master0 master0.ocp4.example.com:22623 check inter 1s
  server master1 master1.ocp4.example.com:22623 check inter 1s
  server master2 master2.ocp4.example.com:22623 check inter 1s
listen ingress-router-443 (4)
  bind *:443
  mode tcp
  balance source
  server worker0 worker0.ocp4.example.com:443 check inter 1s
  server worker1 worker1.ocp4.example.com:443 check inter 1s
listen ingress-router-80 (5)
  bind *:80
  mode tcp
  balance source
  server worker0 worker0.ocp4.example.com:80 check inter 1s
  server worker1 worker1.ocp4.example.com:80 check inter 1s
1 Port 6443 handles the Kubernetes API traffic and points to the control plane machines.
2 The bootstrap entries must be in place before the OKD cluster installation and they must be removed after the bootstrap process is complete.
3 Port 22623 handles the machine config server traffic and points to the control plane machines.
4 Port 443 handles the HTTPS traffic and points to the machines that run the Ingress Controller pods. The Ingress Controller pods run on the compute machines by default.
5 Port 80 handles the HTTP traffic and points to the machines that run the Ingress Controller pods. The Ingress Controller pods run on the compute machines by default.

If you are deploying a three-node cluster with zero compute nodes, the Ingress Controller pods run on the control plane nodes. In three-node cluster deployments, you must configure your application Ingress load balancer to route HTTP and HTTPS traffic to the control plane nodes.

If you are using HAProxy as a load balancer, you can check that the haproxy process is listening on ports 6443, 22623, 443, and 80 by running netstat -nltupe on the HAProxy node.

Enabling Swift on OpenStack

Swift is operated by a user account with the swiftoperator role. Add the role to an account before you run the installation program.

If the OpenStack object storage service, commonly known as Swift, is available, OKD uses it as the image registry storage. If it is unavailable, the installation program relies on the OpenStack block storage service, commonly known as Cinder.

If Swift is present and you want to use it, you must enable access to it. If it is not present, or if you do not want to use it, skip this section.

OpenStack 17 sets the rgw_max_attr_size parameter of Ceph RGW to 256 characters. This setting causes issues with uploading container images to the OKD registry. You must set the value of rgw_max_attr_size to at least 1024 characters.

Before installation, check if your OpenStack deployment is affected by this problem. If it is, reconfigure Ceph RGW.
Prerequisites

  • You have a OpenStack administrator account on the target environment.

  • The Swift service is installed.

  • On Ceph RGW, the account in url option is enabled.

Procedure

To enable Swift on OpenStack:

  1. As an administrator in the OpenStack CLI, add the swiftoperator role to the account that will access Swift:

    $ openstack role add --user <user> --project <project> swiftoperator

Your OpenStack deployment can now use Swift for the image registry.

Configuring an image registry with custom storage on clusters that run on OpenStack

After you install a cluster on OpenStack, you can use a Cinder volume that is in a specific availability zone for registry storage.

Procedure

  1. Create a YAML file that specifies the storage class and availability zone to use. For example:

    apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: custom-csi-storageclass
provisioner: cinder.csi.openstack.org
volumeBindingMode: WaitForFirstConsumer
allowVolumeExpansion: true
parameters:
  availability: <availability_zone_name>

    OKD does not verify the existence of the availability zone you choose. Verify the name of the availability zone before you apply the configuration.

  2. From a command line, apply the configuration:

    $ oc apply -f <storage_class_file_name>
    Example output
    storageclass.storage.k8s.io/custom-csi-storageclass created

  3. Create a YAML file that specifies a persistent volume claim (PVC) that uses your storage class and the openshift-image-registry namespace. For example:

    apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: csi-pvc-imageregistry
  namespace: openshift-image-registry (1)
  annotations:
    imageregistry.openshift.io: "true"
spec:
  accessModes:
  - ReadWriteOnce
  volumeMode: Filesystem
  resources:
    requests:
      storage: 100Gi (2)
  storageClassName: <your_custom_storage_class> (3)
    1 Enter the namespace openshift-image-registry. This namespace allows the Cluster Image Registry Operator to consume the PVC.
    2 Optional: Adjust the volume size.
    3 Enter the name of the storage class that you created.

  4. From a command line, apply the configuration:

    $ oc apply -f <pvc_file_name>
    Example output
    persistentvolumeclaim/csi-pvc-imageregistry created

  5. Replace the original persistent volume claim in the image registry configuration with the new claim:

    $ oc patch configs.imageregistry.operator.openshift.io/cluster --type 'json' -p='[{"op": "replace", "path": "/spec/storage/pvc/claim", "value": "csi-pvc-imageregistry"}]'
    Example output
    config.imageregistry.operator.openshift.io/cluster patched

    Over the next several minutes, the configuration is updated.

Verification

To confirm that the registry is using the resources that you defined:

  1. Verify that the PVC claim value is identical to the name that you provided in your PVC definition:

    $ oc get configs.imageregistry.operator.openshift.io/cluster -o yaml
    Example output
    ...
status:
    ...
    managementState: Managed
    pvc:
      claim: csi-pvc-imageregistry
...

  2. Verify that the status of the PVC is Bound:

    $ oc get pvc -n openshift-image-registry csi-pvc-imageregistry
    Example output
    NAME                   STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS             AGE
csi-pvc-imageregistry  Bound    pvc-72a8f9c9-f462-11e8-b6b6-fa163e18b7b5   100Gi      RWO            custom-csi-storageclass  11m

Verifying external network access

The OKD installation process requires external network access. You must provide an external network value to it, or deployment fails. Before you begin the process, verify that a network with the external router type exists in OpenStack.

Prerequisites
Procedure

  1. Using the OpenStack CLI, verify the name and ID of the 'External' network:

    $ openstack network list --long -c ID -c Name -c "Router Type"
    Example output
    +--------------------------------------+----------------+-------------+
| ID                                   | Name           | Router Type |
+--------------------------------------+----------------+-------------+
| 148a8023-62a7-4672-b018-003462f8d7dc | public_network | External    |
+--------------------------------------+----------------+-------------+

A network with an external router type appears in the network list. If at least one does not, see Creating a default floating IP network and Creating a default provider network.

If the external network’s CIDR range overlaps one of the default network ranges, you must change the matching network ranges in the install-config.yaml file before you start the installation process.

The default network ranges are:

Network Range

machineNetwork

10.0.0.0/16

serviceNetwork

172.30.0.0/16

clusterNetwork

10.128.0.0/14
If the installation program finds multiple networks with the same name, it sets one of them at random. To avoid this behavior, create unique names for resources in OpenStack.

If the Neutron trunk service plugin is enabled, a trunk port is created by default. For more information, see Neutron trunk port.

Defining parameters for the installation program

The OKD installation program relies on a file that is called clouds.yaml. The file describes OpenStack configuration parameters, including the project name, log in information, and authorization service URLs.

Procedure

  1. Create the clouds.yaml file:

    • If your OpenStack distribution includes the Horizon web UI, generate a clouds.yaml file in it.

      Remember to add a password to the auth field. You can also keep secrets in a separate file from clouds.yaml.

    • If your OpenStack distribution does not include the Horizon web UI, or you do not want to use Horizon, create the file yourself. For detailed information about clouds.yaml, see Config files in the OpenStack documentation.

      clouds:
  shiftstack:
    auth:
      auth_url: http://10.10.14.42:5000/v3
      project_name: shiftstack
      username: <username>
      password: <password>
      user_domain_name: Default
      project_domain_name: Default
  dev-env:
    region_name: RegionOne
    auth:
      username: <username>
      password: <password>
      project_name: 'devonly'
      auth_url: 'https://10.10.14.22:5001/v2.0'

  2. If your OpenStack installation uses self-signed certificate authority (CA) certificates for endpoint authentication:

    1. Copy the certificate authority file to your machine.

    2. Add the cacerts key to the clouds.yaml file. The value must be an absolute, non-root-accessible path to the CA certificate:

      clouds:
  shiftstack:
    ...
    cacert: "/etc/pki/ca-trust/source/anchors/ca.crt.pem"

      After you run the installer with a custom CA certificate, you can update the certificate by editing the value of the ca-cert.pem key in the cloud-provider-config keymap. On a command line, run:

      $ oc edit configmap -n openshift-config cloud-provider-config

  3. Place the clouds.yaml file in one of the following locations:

    1. The value of the OS_CLIENT_CONFIG_FILE environment variable

    2. The current directory

    3. A Unix-specific user configuration directory, for example ~/.config/openstack/clouds.yaml

    4. A Unix-specific site configuration directory, for example /etc/openstack/clouds.yaml

      The installation program searches for clouds.yaml in that order.

Setting OpenStack Cloud Controller Manager options

Optionally, you can edit the OpenStack Cloud Controller Manager (CCM) configuration for your cluster. This configuration controls how OKD interacts with OpenStack.

For a complete list of configuration parameters, see the "OpenStack Cloud Controller Manager reference guide" page in the "Installing on OpenStack" documentation.

Procedure

  1. If you have not already generated manifest files for your cluster, generate them by running the following command:

    $ openshift-install --dir <destination_directory> create manifests

  2. In a text editor, open the cloud-provider configuration manifest file. For example:

    $ vi openshift/manifests/cloud-provider-config.yaml

  3. Modify the options according to the CCM reference guide.

    Configuring Octavia for load balancing is a common case for clusters that do not use Kuryr. For example:

    #...
[LoadBalancer]
lb-provider = "amphora" (1)
floating-network-id="d3deb660-4190-40a3-91f1-37326fe6ec4a" (2)
create-monitor = True (3)
monitor-delay = 10s (4)
monitor-timeout = 10s (5)
monitor-max-retries = 1 (6)
#...
    1 This property sets the Octavia provider that your load balancer uses. It accepts "ovn" or "amphora" as values. If you choose to use OVN, you must also set lb-method to SOURCE_IP_PORT.
    2 This property is required if you want to use multiple external networks with your cluster. The cloud provider creates floating IP addresses on the network that is specified here.
    3 This property controls whether the cloud provider creates health monitors for Octavia load balancers. Set the value to True to create health monitors. As of OpenStack 16.2, this feature is only available for the Amphora provider.
    4 This property sets the frequency with which endpoints are monitored. The value must be in the time.ParseDuration() format. This property is required if the value of the create-monitor property is True.
    5 This property sets the time that monitoring requests are open before timing out. The value must be in the time.ParseDuration() format. This property is required if the value of the create-monitor property is True.
    6 This property defines how many successful monitoring requests are required before a load balancer is marked as online. The value must be an integer. This property is required if the value of the create-monitor property is True.

    Prior to saving your changes, verify that the file is structured correctly. Clusters might fail if properties are not placed in the appropriate section.

    You must set the value of the create-monitor property to True if you use services that have the value of the .spec.externalTrafficPolicy property set to Local. The OVN Octavia provider in OpenStack 16.2 does not support health monitors. Therefore, services that have ETP parameter values set to Local might not respond when the lb-provider value is set to "ovn".

    For installations that use Kuryr, Kuryr handles relevant services. There is no need to configure Octavia load balancing in the cloud provider.

  4. Save the changes to the file and proceed with installation.

    You can update your cloud provider configuration after you run the installer. On a command line, run:

    $ oc edit configmap -n openshift-config cloud-provider-config

    After you save your changes, your cluster will take some time to reconfigure itself. The process is complete if none of your nodes have a SchedulingDisabled status.

Obtaining the installation program

Before you install OKD, download the installation file on the host you are using for installation.

Prerequisites

  • You have a computer that runs Linux or macOS, with at least 1.2 GB of local disk space.

Procedure

  1. Download the installation program from https://github.com/openshift/okd/releases.

    • The installation program creates several files on the computer that you use to install your cluster. You must keep the installation program and the files that the installation program creates after you finish installing the cluster. Both of the files are required to delete the cluster.

    • Deleting the files created by the installation program does not remove your cluster, even if the cluster failed during installation. To remove your cluster, complete the OKD uninstallation procedures for your specific cloud provider.

  2. Extract the installation program. For example, on a computer that uses a Linux operating system, run the following command:

    $ tar -xvf openshift-install-linux.tar.gz

  3. Download your installation pull secret from Red Hat OpenShift Cluster Manager. This pull secret allows you to authenticate with the services that are provided by the included authorities, including Quay.io, which serves the container images for OKD components.

    Using a pull secret from Red Hat OpenShift Cluster Manager is not required. You can use a pull secret for another private registry. Or, if you do not need the cluster to pull images from a private registry, you can use {"auths":{"fake":{"auth":"aWQ6cGFzcwo="}}} as the pull secret when prompted during the installation.

    If you do not use the pull secret from Red Hat OpenShift Cluster Manager:

    • Red Hat Operators are not available.

    • The Telemetry and Insights operators do not send data to Red Hat.

    • Content from the Red Hat Ecosystem Catalog Container images registry, such as image streams and Operators, are not available.

Creating the installation configuration file

You can customize the OKD cluster you install on OpenStack.

Prerequisites

  • You have the OKD installation program and the pull secret for your cluster.

Procedure

  1. Create the install-config.yaml file.

    1. Change to the directory that contains the installation program and run the following command:

      $ ./openshift-install create install-config --dir <installation_directory> (1)
      1 For <installation_directory>, specify the directory name to store the files that the installation program creates.

      When specifying the directory:

      • Verify that the directory has the execute permission. This permission is required to run Terraform binaries under the installation directory.

      • Use an empty directory. Some installation assets, such as bootstrap X.509 certificates, have short expiration intervals, therefore you must not reuse an installation directory. If you want to reuse individual files from another cluster installation, you can copy them into your directory. However, the file names for the installation assets might change between releases. Use caution when copying installation files from an earlier OKD version.

        Always delete the ~/.powervs directory to avoid reusing a stale configuration. Run the following command:

        $ rm -rf ~/.powervs

    2. At the prompts, provide the configuration details for your cloud:

      1. Optional: Select an SSH key to use to access your cluster machines.

        For production OKD clusters on which you want to perform installation debugging or disaster recovery, specify an SSH key that your ssh-agent process uses.

      2. Select openstack as the platform to target.

      3. Specify the OpenStack external network name to use for installing the cluster.

      4. Specify the floating IP address to use for external access to the OpenShift API.

      5. Specify a OpenStack flavor with at least 16 GB RAM to use for control plane nodes and 8 GB RAM for compute nodes.

      6. Select the base domain to deploy the cluster to. All DNS records will be sub-domains of this base and will also include the cluster name.

      7. Enter a name for your cluster. The name must be 14 or fewer characters long.

  2. Modify the install-config.yaml file. You can find more information about the available parameters in the "Installation configuration parameters" section.

  3. Back up the install-config.yaml file so that you can use it to install multiple clusters.

    The install-config.yaml file is consumed during the installation process. If you want to reuse the file, you must back it up now.
Additional resources

Configuring the cluster-wide proxy during installation

Production environments can deny direct access to the internet and instead have an HTTP or HTTPS proxy available. You can configure a new OKD cluster to use a proxy by configuring the proxy settings in the install-config.yaml file.

Prerequisites

  • You have an existing install-config.yaml file.

  • You reviewed the sites that your cluster requires access to and determined whether any of them need to bypass the proxy. By default, all cluster egress traffic is proxied, including calls to hosting cloud provider APIs. You added sites to the Proxy object’s spec.noProxy field to bypass the proxy if necessary.

    The Proxy object status.noProxy field is populated with the values of the networking.machineNetwork[].cidr, networking.clusterNetwork[].cidr, and networking.serviceNetwork[] fields from your installation configuration.

    For installations on Amazon Web Services (AWS), Google Cloud Platform (GCP), Microsoft Azure, and OpenStack, the Proxy object status.noProxy field is also populated with the instance metadata endpoint (169.254.169.254).
Procedure

  1. Edit your install-config.yaml file and add the proxy settings. For example:

    apiVersion: v1
baseDomain: my.domain.com
proxy:
  httpProxy: http://<username>:<pswd>@<ip>:<port> (1)
  httpsProxy: https://<username>:<pswd>@<ip>:<port> (2)
  noProxy: example.com (3)
additionalTrustBundle: | (4)
    -----BEGIN CERTIFICATE-----
    <MY_TRUSTED_CA_CERT>
    -----END CERTIFICATE-----
additionalTrustBundlePolicy: <policy_to_add_additionalTrustBundle> (5)
    1 A proxy URL to use for creating HTTP connections outside the cluster. The URL scheme must be http.
    2 A proxy URL to use for creating HTTPS connections outside the cluster.
    3 A comma-separated list of destination domain names, IP addresses, or other network CIDRs to exclude from proxying. Preface a domain with . to match subdomains only. For example, .y.com matches x.y.com, but not y.com. Use * to bypass the proxy for all destinations.
    4 If provided, the installation program generates a config map that is named user-ca-bundle in the openshift-config namespace that contains one or more additional CA certificates that are required for proxying HTTPS connections. The Cluster Network Operator then creates a trusted-ca-bundle config map that merges these contents with the Fedora CoreOS (FCOS) trust bundle, and this config map is referenced in the trustedCA field of the Proxy object. The additionalTrustBundle field is required unless the proxy’s identity certificate is signed by an authority from the FCOS trust bundle.
    5 Optional: The policy to determine the configuration of the Proxy object to reference the user-ca-bundle config map in the trustedCA field. The allowed values are Proxyonly and Always. Use Proxyonly to reference the user-ca-bundle config map only when http/https proxy is configured. Use Always to always reference the user-ca-bundle config map. The default value is Proxyonly.

    The installation program does not support the proxy readinessEndpoints field.

    If the installer times out, restart and then complete the deployment by using the wait-for command of the installer. For example:

    $ ./openshift-install wait-for install-complete --log-level debug

  2. Save the file and reference it when installing OKD.

The installation program creates a cluster-wide proxy that is named cluster that uses the proxy settings in the provided install-config.yaml file. If no proxy settings are provided, a cluster Proxy object is still created, but it will have a nil spec.

Only the Proxy object named cluster is supported, and no additional proxies can be created.

Custom subnets in OpenStack deployments

Optionally, you can deploy a cluster on a OpenStack subnet of your choice. The subnet’s GUID is passed as the value of platform.openstack.machinesSubnet in the install-config.yaml file.

This subnet is used as the cluster’s primary subnet. By default, nodes and ports are created on it. You can create nodes and ports on a different OpenStack subnet by setting the value of the platform.openstack.machinesSubnet property to the subnet’s UUID.

Before you run the OKD installer with a custom subnet, verify that your configuration meets the following requirements:

  • The subnet that is used by platform.openstack.machinesSubnet has DHCP enabled.

  • The CIDR of platform.openstack.machinesSubnet matches the CIDR of networking.machineNetwork.

  • The installation program user has permission to create ports on this network, including ports with fixed IP addresses.

Clusters that use custom subnets have the following limitations:

  • If you plan to install a cluster that uses floating IP addresses, the platform.openstack.machinesSubnet subnet must be attached to a router that is connected to the externalNetwork network.

  • If the platform.openstack.machinesSubnet value is set in the install-config.yaml file, the installation program does not create a private network or subnet for your OpenStack machines.

  • You cannot use the platform.openstack.externalDNS property at the same time as a custom subnet. To add DNS to a cluster that uses a custom subnet, configure DNS on the OpenStack network.

By default, the API VIP takes x.x.x.5 and the Ingress VIP takes x.x.x.7 from your network’s CIDR block. To override these default values, set values for platform.openstack.apiVIPs and platform.openstack.ingressVIPs that are outside of the DHCP allocation pool.

The CIDR ranges for networks are not adjustable after cluster installation. Red Hat does not provide direct guidance on determining the range during cluster installation because it requires careful consideration of the number of created pods per namespace.

Deploying a cluster with bare metal machines

If you want your cluster to use bare metal machines, modify the install-config.yaml file. Your cluster can have both control plane and compute machines running on bare metal, or just compute machines.

Bare-metal compute machines are not supported on clusters that use Kuryr.

Be sure that your install-config.yaml file reflects whether the OpenStack network that you use for bare metal workers supports floating IP addresses or not.
Prerequisites

  • The OpenStack Bare Metal service (Ironic) is enabled and accessible via the OpenStack Compute API.

  • Bare metal is available as a OpenStack flavor.

  • If your cluster runs on an OpenStack version that is more than 16.1.6 and less than 16.2.4, bare metal workers do not function due to a known issue that causes the metadata service to be unavailable for services on OKD nodes.

  • The OpenStack network supports both VM and bare metal server attachment.

  • If you want to deploy the machines on a pre-existing network, a OpenStack subnet is provisioned.

  • If you want to deploy the machines on an installer-provisioned network, the OpenStack Bare Metal service (Ironic) is able to listen for and interact with Preboot eXecution Environment (PXE) boot machines that run on tenant networks.

  • You created an install-config.yaml file as part of the OKD installation process.

Procedure

  1. In the install-config.yaml file, edit the flavors for machines:

    1. If you want to use bare-metal control plane machines, change the value of controlPlane.platform.openstack.type to a bare metal flavor.

    2. Change the value of compute.platform.openstack.type to a bare metal flavor.

    3. If you want to deploy your machines on a pre-existing network, change the value of platform.openstack.machinesSubnet to the OpenStack subnet UUID of the network. Control plane and compute machines must use the same subnet.

      An example bare metal install-config.yaml file
      controlPlane:
    platform:
      openstack:
        type: <bare_metal_control_plane_flavor> (1)
...

compute:
  - architecture: amd64
    hyperthreading: Enabled
    name: worker
    platform:
      openstack:
        type: <bare_metal_compute_flavor> (2)
    replicas: 3
...

platform:
    openstack:
      machinesSubnet: <subnet_UUID> (3)
...
      1 If you want to have bare-metal control plane machines, change this value to a bare metal flavor.
      2 Change this value to a bare metal flavor to use for compute machines.
      3 If you want to use a pre-existing network, change this value to the UUID of the OpenStack subnet.

Use the updated install-config.yaml file to complete the installation process. The compute machines that are created during deployment use the flavor that you added to the file.

The installer may time out while waiting for bare metal machines to boot.

If the installer times out, restart and then complete the deployment by using the wait-for command of the installer. For example:

$ ./openshift-install wait-for install-complete --log-level debug

Cluster deployment on OpenStack provider networks

You can deploy your OKD clusters on OpenStack with a primary network interface on a provider network. Provider networks are commonly used to give projects direct access to a public network that can be used to reach the internet. You can also share provider networks among projects as part of the network creation process.

OpenStack provider networks map directly to an existing physical network in the data center. A OpenStack administrator must create them.

In the following example, OKD workloads are connected to a data center by using a provider network:

A diagram that depicts four OpenShift workloads on OpenStack. Each workload is connected by its NIC to an external data center by using a provider network.

OKD clusters that are installed on provider networks do not require tenant networks or floating IP addresses. The installer does not create these resources during installation.

Example provider network types include flat (untagged) and VLAN (802.1Q tagged).

A cluster can support as many provider network connections as the network type allows. For example, VLAN networks typically support up to 4096 connections.

You can learn more about provider and tenant networks in the OpenStack documentation.

OpenStack provider network requirements for cluster installation

Before you install an OKD cluster, your OpenStack deployment and provider network must meet a number of conditions:

  • The OpenStack networking service (Neutron) is enabled and accessible through the OpenStack networking API.

  • The OpenStack networking service has the port security and allowed address pairs extensions enabled.

  • The provider network can be shared with other tenants.

    Use the openstack network create command with the --share flag to create a network that can be shared.

  • The OpenStack project that you use to install the cluster must own the provider network, as well as an appropriate subnet.

    To create a network for a project that is named "openshift," enter the following command
    $ openstack network create --project openshift
    To create a subnet for a project that is named "openshift," enter the following command
    $ openstack subnet create --project openshift

    To learn more about creating networks on OpenStack, read the provider networks documentation.

    If the cluster is owned by the admin user, you must run the installer as that user to create ports on the network.

    Provider networks must be owned by the OpenStack project that is used to create the cluster. If they are not, the OpenStack Compute service (Nova) cannot request a port from that network.

  • Verify that the provider network can reach the OpenStack metadata service IP address, which is 169.254.169.254 by default.

    Depending on your OpenStack SDN and networking service configuration, you might need to provide the route when you create the subnet. For example:

    $ openstack subnet create --dhcp --host-route destination=169.254.169.254/32,gateway=192.0.2.2 ...

  • Optional: To secure the network, create role-based access control (RBAC) rules that limit network access to a single project.

Deploying a cluster that has a primary interface on a provider network

You can deploy an OKD cluster that has its primary network interface on an OpenStack provider network.

Prerequisites

  • Your OpenStack deployment is configured as described by "OpenStack provider network requirements for cluster installation".

Procedure

  1. In a text editor, open the install-config.yaml file.

  2. Set the value of the platform.openstack.apiVIPs property to the IP address for the API VIP.

  3. Set the value of the platform.openstack.ingressVIPs property to the IP address for the Ingress VIP.

  4. Set the value of the platform.openstack.machinesSubnet property to the UUID of the provider network subnet.

  5. Set the value of the networking.machineNetwork.cidr property to the CIDR block of the provider network subnet.

The platform.openstack.apiVIPs and platform.openstack.ingressVIPs properties must both be unassigned IP addresses from the networking.machineNetwork.cidr block.
Section of an installation configuration file for a cluster that relies on a OpenStack provider network
        ...
        platform:
          openstack:
            apiVIPs: (1)
              - 192.0.2.13
            ingressVIPs: (1)
              - 192.0.2.23
            machinesSubnet: fa806b2f-ac49-4bce-b9db-124bc64209bf
            # ...
        networking:
          machineNetwork:
          - cidr: 192.0.2.0/24
1 In OKD 4.12 and later, the apiVIP and ingressVIP configuration settings are deprecated. Instead, use a list format to enter values in the apiVIPs and ingressVIPs configuration settings.

You cannot set the platform.openstack.externalNetwork or platform.openstack.externalDNS parameters while using a provider network for the primary network interface.

When you deploy the cluster, the installer uses the install-config.yaml file to deploy the cluster on the provider network.

You can add additional networks, including provider networks, to the platform.openstack.additionalNetworkIDs list.

After you deploy your cluster, you can attach pods to additional networks. For more information, see Understanding multiple networks.

Sample customized install-config.yaml file for OpenStack

This sample install-config.yaml demonstrates all of the possible OpenStack customization options.

This sample file is provided for reference only. You must obtain your install-config.yaml file by using the installation program. 
apiVersion: v1
baseDomain: example.com
controlPlane:
  name: master
  platform: {}
  replicas: 3
compute:
- name: worker
  platform:
    openstack:
      type: ml.large
  replicas: 3
metadata:
  name: example
networking:
  clusterNetwork:
  - cidr: 10.128.0.0/14
    hostPrefix: 23
  machineNetwork:
  - cidr: 10.0.0.0/16
  serviceNetwork:
  - 172.30.0.0/16
  networkType: OVNKubernetes
platform:
  openstack:
    cloud: mycloud
    externalNetwork: external
    computeFlavor: m1.xlarge
    apiFloatingIP: 128.0.0.1
pullSecret: '{"auths": ...}'
sshKey: ssh-ed25519 AAAA...

Optional: Configuring a cluster with dual-stack networking

Dual-stack configuration for OpenStack is a Technology Preview feature only. Technology Preview features are not supported with Red Hat production service level agreements (SLAs) and might not be functionally complete. Red Hat does not recommend using them in production. These features provide early access to upcoming product features, enabling customers to test functionality and provide feedback during the development process.

For more information about the support scope of Red Hat Technology Preview features, see Technology Preview Features Support Scope.

You can create a dual-stack cluster on OpenStack. However, the dual-stack configuration is enabled only if you are using an OpenStack network with IPv4 and IPv6 subnets.

OpenStack does not support the following configurations:

  • Conversion of an IPv4 single-stack cluster to a dual-stack cluster network.

  • IPv6 as the primary address family for dual-stack cluster network.

Deploying the dual-stack cluster

For dual-stack networking in OKD clusters, you can configure IPv4 and IPv6 address endpoints for cluster nodes.

Prerequisites

  • You enabled Dynamic Host Configuration Protocol (DHCP) on the subnets.

Procedure

  1. Create a network with IPv4 and IPv6 subnets. The available address modes for ipv6-ra-mode and ipv6-address-mode fields are: stateful, stateless and slaac.

    The dual-stack network MTU must accommodate both the minimum MTU for IPv6, which is 1280, and the OVN-Kubernetes encapsulation overhead, which is 100.

  2. Create the API and Ingress VIPs ports.

  3. Add the IPv6 subnet to the router to enable router advertisements. If you are using a provider network, you can enable router advertisements by adding the network as an external gateway, which also enables external connectivity.

  4. For an IPv4/IPv6 dual-stack cluster where you set IPv4 as the primary endpoint for your cluster nodes, edit the install-config.yaml file like the following example:

    apiVersion: v1
baseDomain: mydomain.test
featureSet: TechPreviewNoUpgrade (1)
compute:
- name: worker
  platform:
    openstack:
      type: m1.xlarge
  replicas: 3
controlPlane:
  name: master
  platform:
    openstack:
      type: m1.xlarge
  replicas: 3
metadata:
  name: mycluster
networking:
  machineNetwork: (2)
  - cidr: "192.168.25.0/24"
  - cidr: "fd2e:6f44:5dd8:c956::/64"
  clusterNetwork: (2)
  - cidr: 10.128.0.0/14
    hostPrefix: 23
  - cidr: fd01::/48
    hostPrefix: 64
  serviceNetwork: (2)
  - 172.30.0.0/16
  - fd02::/112
platform:
  openstack:
    ingressVIPs: ['192.168.25.79', 'fd2e:6f44:5dd8:c956:f816:3eff:fef1:1bad'] (3)
    apiVIPs: ['192.168.25.199', 'fd2e:6f44:5dd8:c956:f816:3eff:fe78:cf36'] (4)
    controlPlanePort: (5)
      fixedIPs: (6)
      - subnet: (7)
          name: subnet-v4
          id: subnet-v4-id
      - subnet: (7)
          name: subnet-v6
          id: subnet-v6-id
      network: (7)
        name: dualstack
        id: network-id
    1 Dual-stack clusters are supported only with the TechPreviewNoUpgrade value.
    2 You must specify an IP address range in the cidr field for both IPv4 and IPv6 address families.
    3 Specify the virtual IP (VIP) address endpoints for the Ingress VIP services to provide an interface to the cluster.
    4 Specify the virtual IP (VIP) address endpoints for the API VIP services to provide an interface to the cluster.
    5 Specify the dual-stack network details that are used by all the nodes across the cluster.
    6 The Classless Inter-Domain Routing (CIDR) of any subnet specified in this field must match the CIDRs listed on networks.machineNetwork.
    7 You can specify a value for name, id, or both.

Installation configuration for a cluster on OpenStack with a user-managed load balancer

The following example install-config.yaml file demonstrates how to configure a cluster that uses an external, user-managed load balancer rather than the default internal load balancer.

apiVersion: v1
baseDomain: mydomain.test
compute:
- name: worker
  platform:
    openstack:
      type: m1.xlarge
  replicas: 3
controlPlane:
  name: master
  platform:
    openstack:
      type: m1.xlarge
  replicas: 3
metadata:
  name: mycluster
networking:
  clusterNetwork:
  - cidr: 10.128.0.0/14
    hostPrefix: 23
  machineNetwork:
  - cidr: 192.168.10.0/24
platform:
  openstack:
    cloud: mycloud
    machinesSubnet: 8586bf1a-cc3c-4d40-bdf6-c243decc603a (1)
    apiVIPs:
    - 192.168.10.5
    ingressVIPs:
    - 192.168.10.7
    loadBalancer:
      type: UserManaged (2)
1 Regardless of which load balancer you use, the load balancer is deployed to this subnet.
2 The UserManaged value indicates that you are using an user-managed load balancer.

Generating a key pair for cluster node SSH access

During an OKD installation, you can provide an SSH public key to the installation program. The key is passed to the Fedora CoreOS (FCOS) nodes through their Ignition config files and is used to authenticate SSH access to the nodes. The key is added to the ~/.ssh/authorized_keys list for the core user on each node, which enables password-less authentication.

After the key is passed to the nodes, you can use the key pair to SSH in to the FCOS nodes as the user core. To access the nodes through SSH, the private key identity must be managed by SSH for your local user.

If you want to SSH in to your cluster nodes to perform installation debugging or disaster recovery, you must provide the SSH public key during the installation process. The ./openshift-install gather command also requires the SSH public key to be in place on the cluster nodes.

Do not skip this procedure in production environments, where disaster recovery and debugging is required.

On clusters running Fedora CoreOS (FCOS), the SSH keys specified in the Ignition config files are written to the /home/core/.ssh/authorized_keys.d/core file. However, the Machine Config Operator manages SSH keys in the /home/core/.ssh/authorized_keys file and configures sshd to ignore the /home/core/.ssh/authorized_keys.d/core file. As a result, newly provisioned OKD nodes are not accessible using SSH until the Machine Config Operator reconciles the machine configs with the authorized_keys file. After you can access the nodes using SSH, you can delete the /home/core/.ssh/authorized_keys.d/core file.
Procedure

  1. If you do not have an existing SSH key pair on your local machine to use for authentication onto your cluster nodes, create one. For example, on a computer that uses a Linux operating system, run the following command:

    $ ssh-keygen -t ed25519 -N '' -f <path>/<file_name> (1)
    1 Specify the path and file name, such as ~/.ssh/id_ed25519, of the new SSH key. If you have an existing key pair, ensure your public key is in the your ~/.ssh directory.

    If you plan to install an OKD cluster that uses the Fedora cryptographic libraries that have been submitted to NIST for FIPS 140-2/140-3 Validation on only the x86_64, ppc64le, and s390x architectures, do not create a key that uses the ed25519 algorithm. Instead, create a key that uses the rsa or ecdsa algorithm.

  2. View the public SSH key:

    $ cat <path>/<file_name>.pub

    For example, run the following to view the ~/.ssh/id_ed25519.pub public key:

    $ cat ~/.ssh/id_ed25519.pub

  3. Add the SSH private key identity to the SSH agent for your local user, if it has not already been added. SSH agent management of the key is required for password-less SSH authentication onto your cluster nodes, or if you want to use the ./openshift-install gather command.

    On some distributions, default SSH private key identities such as ~/.ssh/id_rsa and ~/.ssh/id_dsa are managed automatically.

    1. If the ssh-agent process is not already running for your local user, start it as a background task:

      $ eval "$(ssh-agent -s)"
      Example output
      Agent pid 31874

      If your cluster is in FIPS mode, only use FIPS-compliant algorithms to generate the SSH key. The key must be either RSA or ECDSA.

  4. Add your SSH private key to the ssh-agent:

    $ ssh-add <path>/<file_name> (1)
    1 Specify the path and file name for your SSH private key, such as ~/.ssh/id_ed25519
    Example output
    Identity added: /home/<you>/<path>/<file_name> (<computer_name>)
Next steps

  • When you install OKD, provide the SSH public key to the installation program.

Enabling access to the environment

At deployment, all OKD machines are created in a OpenStack-tenant network. Therefore, they are not accessible directly in most OpenStack deployments.

You can configure OKD API and application access by using floating IP addresses (FIPs) during installation. You can also complete an installation without configuring FIPs, but the installer will not configure a way to reach the API or applications externally.

Enabling access with floating IP addresses

Create floating IP (FIP) addresses for external access to the OKD API and cluster applications.

Procedure

  1. Using the OpenStack CLI, create the API FIP:

    $ openstack floating ip create --description "API <cluster_name>.<base_domain>" <external_network>

  2. Using the OpenStack CLI, create the apps, or Ingress, FIP:

    $ openstack floating ip create --description "Ingress <cluster_name>.<base_domain>" <external_network>

  3. Add records that follow these patterns to your DNS server for the API and Ingress FIPs:

    api.<cluster_name>.<base_domain>.  IN  A  <API_FIP>
*.apps.<cluster_name>.<base_domain>. IN  A <apps_FIP>

    If you do not control the DNS server, you can access the cluster by adding the cluster domain names such as the following to your /etc/hosts file:

    • <api_floating_ip> api.<cluster_name>.<base_domain>

    • <application_floating_ip> grafana-openshift-monitoring.apps.<cluster_name>.<base_domain>

    • <application_floating_ip> prometheus-k8s-openshift-monitoring.apps.<cluster_name>.<base_domain>

    • <application_floating_ip> oauth-openshift.apps.<cluster_name>.<base_domain>

    • <application_floating_ip> console-openshift-console.apps.<cluster_name>.<base_domain>

    • application_floating_ip integrated-oauth-server-openshift-authentication.apps.<cluster_name>.<base_domain>

    The cluster domain names in the /etc/hosts file grant access to the web console and the monitoring interface of your cluster locally. You can also use the kubectl or oc. You can access the user applications by using the additional entries pointing to the <application_floating_ip>. This action makes the API and applications accessible to only you, which is not suitable for production deployment, but does allow installation for development and testing.

  4. Add the FIPs to the install-config.yaml file as the values of the following parameters:

    • platform.openstack.ingressFloatingIP

    • platform.openstack.apiFloatingIP

If you use these values, you must also enter an external network as the value of the platform.openstack.externalNetwork parameter in the install-config.yaml file.

You can make OKD resources available outside of the cluster by assigning a floating IP address and updating your firewall configuration.

Completing installation without floating IP addresses

You can install OKD on OpenStack without providing floating IP addresses.

In the install-config.yaml file, do not define the following parameters:

  • platform.openstack.ingressFloatingIP

  • platform.openstack.apiFloatingIP

If you cannot provide an external network, you can also leave platform.openstack.externalNetwork blank. If you do not provide a value for platform.openstack.externalNetwork, a router is not created for you, and, without additional action, the installer will fail to retrieve an image from Glance. You must configure external connectivity on your own.

If you run the installer from a system that cannot reach the cluster API due to a lack of floating IP addresses or name resolution, installation fails. To prevent installation failure in these cases, you can use a proxy network or run the installer from a system that is on the same network as your machines.

You can enable name resolution by creating DNS records for the API and Ingress ports. For example:

api.<cluster_name>.<base_domain>.  IN  A  <api_port_IP>
*.apps.<cluster_name>.<base_domain>. IN  A <ingress_port_IP>

If you do not control the DNS server, you can add the record to your /etc/hosts file. This action makes the API accessible to only you, which is not suitable for production deployment but does allow installation for development and testing.

Deploying the cluster

You can install OKD on a compatible cloud platform.

You can run the create cluster command of the installation program only once, during initial installation.
Prerequisites

  • You have the OKD installation program and the pull secret for your cluster.

  • You have verified that the cloud provider account on your host has the correct permissions to deploy the cluster. An account with incorrect permissions causes the installation process to fail with an error message that displays the missing permissions.

Procedure

  • Change to the directory that contains the installation program and initialize the cluster deployment:

    $ ./openshift-install create cluster --dir <installation_directory> \ (1)
    --log-level=info (2)
    1 For <installation_directory>, specify the location of your customized ./install-config.yaml file.
    2 To view different installation details, specify warn, debug, or error instead of info.
Verification

When the cluster deployment completes successfully:

  • The terminal displays directions for accessing your cluster, including a link to the web console and credentials for the kubeadmin user.

  • Credential information also outputs to <installation_directory>/.openshift_install.log.

Do not delete the installation program or the files that the installation program creates. Both are required to delete the cluster.
Example output
...
INFO Install complete!
INFO To access the cluster as the system:admin user when using 'oc', run 'export KUBECONFIG=/home/myuser/install_dir/auth/kubeconfig'
INFO Access the OpenShift web-console here: https://console-openshift-console.apps.mycluster.example.com
INFO Login to the console with user: "kubeadmin", and password: "password"
INFO Time elapsed: 36m22s

  • The Ignition config files that the installation program generates contain certificates that expire after 24 hours, which are then renewed at that time. If the cluster is shut down before renewing the certificates and the cluster is later restarted after the 24 hours have elapsed, the cluster automatically recovers the expired certificates. The exception is that you must manually approve the pending node-bootstrapper certificate signing requests (CSRs) to recover kubelet certificates. See the documentation for Recovering from expired control plane certificates for more information.

  • It is recommended that you use Ignition config files within 12 hours after they are generated because the 24-hour certificate rotates from 16 to 22 hours after the cluster is installed. By using the Ignition config files within 12 hours, you can avoid installation failure if the certificate update runs during installation.

Verifying cluster status

You can verify your OKD cluster’s status during or after installation.

Procedure

  1. In the cluster environment, export the administrator’s kubeconfig file:

    $ export KUBECONFIG=<installation_directory>/auth/kubeconfig (1)
    1 For <installation_directory>, specify the path to the directory that you stored the installation files in.

    The kubeconfig file contains information about the cluster that is used by the CLI to connect a client to the correct cluster and API server.

  2. View the control plane and compute machines created after a deployment:

    $ oc get nodes

  3. View your cluster’s version:

    $ oc get clusterversion

  4. View your Operators' status:

    $ oc get clusteroperator

  5. View all running pods in the cluster:

    $ oc get pods -A

Logging in to the cluster by using the CLI

You can log in to your cluster as a default system user by exporting the cluster kubeconfig file. The kubeconfig file contains information about the cluster that is used by the CLI to connect a client to the correct cluster and API server. The file is specific to a cluster and is created during OKD installation.

Prerequisites

  • You deployed an OKD cluster.

  • You installed the oc CLI.

Procedure

  1. Export the kubeadmin credentials:

    $ export KUBECONFIG=<installation_directory>/auth/kubeconfig (1)
    1 For <installation_directory>, specify the path to the directory that you stored the installation files in.

  2. Verify you can run oc commands successfully using the exported configuration:

    $ oc whoami
    Example output
    system:admin
Additional resources
Additional resources

Next steps