×

You can use admission plugins to regulate how OKD functions. The default set of admission plugins for OKD ensures proper functioning for your cluster.

About admission plugins

Admission plugins enforce security standards, resource limits, and configuration requirements across your cluster. These plugins intercept and validate authenticated API requests to ensure compliance with cluster policies.

After a request is authenticated and authorized, the admission plugins ensure that any associated policies are followed.

Admission plugins run in sequence as an admission chain. If any admission plugin in the sequence rejects a request, the whole chain is aborted and an error is returned.

OKD has a default set of admission plugins enabled for each resource type. These are required for proper functioning of the cluster. Admission plugins ignore resources that they are not responsible for.

In addition to the defaults, the admission chain can be extended dynamically through webhook admission plugins that call out to custom webhook servers. There are two types of webhook admission plugins: a mutating admission plugin and a validating admission plugin. The mutating admission plugin runs first and can both modify resources and validate requests. The validating admission plugin validates requests and runs after the mutating admission plugin so that modifications triggered by the mutating admission plugin can also be validated.

Calling webhook servers through a mutating admission plugin can produce side effects on resources related to the target object. In such situations, you must take steps to validate that the end result is as expected.

Dynamic admission should be used cautiously because it impacts cluster control plane operations. When calling webhook servers through webhook admission plugins in OKD 4.21, ensure that you have read the documentation fully and tested for side effects of mutations. Include steps to restore resources back to their original state prior to mutation, in the event that a request does not pass through the entire admission chain.

Default admission plugins

Default validating and mutating admission plugins are enabled in OKD. These default plugins contribute to fundamental control plane functionality, such as ingress policy, cluster resource limit override, and quota policy.

Do not run workloads in or share access to default projects. Default projects are reserved for running core cluster components.

The following default projects are considered highly privileged: default, kube-public, kube-system, openshift, openshift-infra, openshift-node, and other system-created projects that have the openshift.io/run-level label set to 0 or 1. Functionality that relies on admission plugins, such as pod security admission, security context constraints, cluster resource quotas, and image reference resolution, does not work in highly privileged projects.

The following list includes the default validating admission plugins:

  • LimitRanger

  • ServiceAccount

  • PodNodeSelector

  • Priority

  • PodTolerationRestriction

  • OwnerReferencesPermissionEnforcement

  • PersistentVolumeClaimResize

  • RuntimeClass

  • CertificateApproval

  • CertificateSigning

  • CertificateSubjectRestriction

  • autoscaling.openshift.io/ManagementCPUsOverride

  • authorization.openshift.io/RestrictSubjectBindings

  • scheduling.openshift.io/OriginPodNodeEnvironment

  • network.openshift.io/ExternalIPRanger

  • network.openshift.io/RestrictedEndpointsAdmission

  • image.openshift.io/ImagePolicy

  • security.openshift.io/SecurityContextConstraint

  • security.openshift.io/SCCExecRestrictions

  • route.openshift.io/IngressAdmission

  • config.openshift.io/ValidateAPIServer

  • config.openshift.io/ValidateAuthentication

  • config.openshift.io/ValidateFeatureGate

  • config.openshift.io/ValidateConsole

  • operator.openshift.io/ValidateDNS

  • config.openshift.io/ValidateImage

  • config.openshift.io/ValidateOAuth

  • config.openshift.io/ValidateProject

  • config.openshift.io/DenyDeleteClusterConfiguration

  • config.openshift.io/ValidateScheduler

  • quota.openshift.io/ValidateClusterResourceQuota

  • security.openshift.io/ValidateSecurityContextConstraints

  • authorization.openshift.io/ValidateRoleBindingRestriction

  • config.openshift.io/ValidateNetwork

  • operator.openshift.io/ValidateKubeControllerManager

  • ValidatingAdmissionWebhook

  • ResourceQuota

  • quota.openshift.io/ClusterResourceQuota

The following list includes the default mutating admission plugins:

  • NamespaceLifecycle

  • LimitRanger

  • ServiceAccount

  • NodeRestriction

  • TaintNodesByCondition

  • PodNodeSelector

  • Priority

  • DefaultTolerationSeconds

  • PodTolerationRestriction

  • DefaultStorageClass

  • StorageObjectInUseProtection

  • RuntimeClass

  • DefaultIngressClass

  • autoscaling.openshift.io/ManagementCPUsOverride

  • scheduling.openshift.io/OriginPodNodeEnvironment

  • image.openshift.io/ImagePolicy

  • security.openshift.io/SecurityContextConstraint

  • security.openshift.io/DefaultSecurityContextConstraints

  • MutatingAdmissionWebhook

Webhook admission plugins

In addition to OKD default admission plugins, you can implement dynamic admission through webhook admission plugins that call webhook servers. This approach extends the functionality of the admission chain.

Webhook servers are called over HTTP at defined endpoints.

There are two types of webhook admission plugins in OKD:

  • During the admission process, the mutating admission plugin can perform tasks, such as injecting affinity labels.

  • At the end of the admission process, the validating admission plugin validates that an object is configured properly, for example ensuring affinity labels are as expected. If the validation passes, OKD schedules the object as configured.

When an API request comes in, mutating or validating admission plugins use the list of external webhooks in the configuration and call them in parallel:

  • If all of the webhooks approve the request, the admission chain continues.

  • If any of the webhooks deny the request, the admission request is denied and the reason for doing so is based on the first denial.

  • If more than one webhook denies the admission request, only the first denial reason is returned to the user.

  • If an error is encountered when calling a webhook, the request is either denied or the webhook is ignored depending on the error policy set. If the error policy is set to Ignore, the request is unconditionally accepted in the event of a failure. If the policy is set to Fail, failed requests are denied. Using Ignore can result in unpredictable behavior for all clients.

Communication between the webhook admission plugin and the webhook server must use TLS. Generate a CA certificate and use the certificate to sign the server certificate that is used by your webhook admission server. The PEM-encoded CA certificate is supplied to the webhook admission plugin using a mechanism, such as service serving certificate secrets.

The following diagram illustrates the sequential admission chain process within which multiple webhook servers are called.

API admission stage
Figure 1. API admission chain with mutating and validating admission plugins

An example webhook admission plugin use case is where all pods must have a common set of labels. In this example, the mutating admission plugin can inject labels and the validating admission plugin can check that labels are as expected. OKD would subsequently schedule pods that include required labels and reject those that do not.

Some common webhook admission plugin use cases include:

  • Namespace reservation.

  • Limiting custom network resources managed by the SR-IOV network device plugin.

  • Defining tolerations that enable taints to qualify which pods should be scheduled on a node.

  • Pod priority class validation.

The maximum default webhook timeout value in OKD is 13 seconds, and it cannot be changed.

Types of webhook admission plugins

As a cluster administrator, you can call webhook servers through the mutating admission plugin or the validating admission plugin in the API server admission chain.

Mutating admission plugin

The mutating admission plugin is invoked during the mutation phase of the admission process, which allows modification of resource content before it is persisted. One example webhook that can be called through the mutating admission plugin is the Pod Node Selector feature, which uses an annotation on a namespace to find a label selector and add it to the pod specification.

Sample mutating admission plugin configuration
apiVersion: admissionregistration.k8s.io/v1beta1
kind: MutatingWebhookConfiguration
metadata:
  name: <webhook_name>
webhooks:
- name: <webhook_name>
  clientConfig:
    service:
      namespace: default
      name: kubernetes
      path: <webhook_url>
    caBundle: <ca_signing_certificate>
  rules:
  - operations:
    - <operation>
    apiGroups:
    - ""
    apiVersions:
    - "*"
    resources:
    - <resource>
  failurePolicy: <policy>
  sideEffects: None

where:

kind

Specifies a mutating admission plugin configuration.

metadata.name

Specifies the name for the MutatingWebhookConfiguration object. Replace <webhook_name> with the appropriate value.

webhooks.name

Specifies the name of the webhook to call. Replace <webhook_name> with the appropriate value.

webhooks.clientConfig

Specifies information about how to connect to, trust, and send data to the webhook server.

webhooks.clientConfig.service.namespace

Specifies the namespace where the front-end service is created.

webhooks.clientConfig.service.name

Specifies the name of the front-end service.

webhooks.clientConfig.service.path

Specifies the webhook URL used for admission requests. Replace <webhook_url> with the appropriate value.

webhooks.clientConfig.caBundle

Specifies a PEM-encoded CA certificate that signs the server certificate that is used by the webhook server. Replace <ca_signing_certificate> with the appropriate certificate in base64 format.

webhooks.rules

Specifies rules that define when the API server should use this webhook admission plugin.

webhooks.rules.operations

Specifies one or more operations that trigger the API server to call this webhook admission plugin. Possible values are create, update, delete, or connect. Replace <operation> and <resource> with the appropriate values.

webhooks.failurePolicy

Specifies how the policy should proceed if the webhook server is unavailable. Replace <policy> with either Ignore (to unconditionally accept the request in case of a failure) or Fail (to deny the failed request). Using Ignore can result in unpredictable behavior for all clients.

In OKD 4.21, objects created by users or control loops through a mutating admission plugin might return unexpected results, especially if values set in an initial request are overwritten, which is not recommended.

Validating admission plugin

A validating admission plugin is invoked during the validation phase of the admission process. This phase allows the enforcement of invariants on particular API resources to ensure that the resource does not change again. The Pod Node Selector is also an example of a webhook which is called by the validating admission plugin, to ensure that all nodeSelector fields are constrained by the node selector restrictions on the namespace.

Sample validating admission plugin configuration
apiVersion: admissionregistration.k8s.io/v1beta1
kind: ValidatingWebhookConfiguration
metadata:
  name: <webhook_name>
webhooks:
- name: <webhook_name>
  clientConfig:
    service:
      namespace: default
      name: kubernetes
      path: <webhook_url>
    caBundle: <ca_signing_certificate>
  rules:
  - operations:
    - <operation>
    apiGroups:
    - ""
    apiVersions:
    - "*"
    resources:
    - <resource>
  failurePolicy: <policy>
  sideEffects: Unknown

where:

kind

Specifies a validating admission plugin configuration.

metadata.name

Specifies the name for the ValidatingWebhookConfiguration object. Replace <webhook_name> with the appropriate value.

webhooks.name

Specifies the name of the webhook to call. Replace <webhook_name> with the appropriate value.

webhooks.clientConfig

Specifies information about how to connect to, trust, and send data to the webhook server.

webhooks.clientConfig.service.namespace

Specifies the namespace where the front-end service is created.

webhooks.clientConfig.service.name

Specifies the name of the front-end service.

webhooks.clientConfig.service.path

Specifies the webhook URL used for admission requests. Replace <webhook_url> with the appropriate value.

webhooks.caBundle

Specifies a PEM-encoded CA certificate that signs the server certificate that is used by the webhook server. Replace <ca_signing_certificate> with the appropriate certificate in base64 format.

webhooks.rules

Specifies rules that define when the API server should use this webhook admission plugin.

webhooks.rules.operations

Specifies one or more operations that trigger the API server to call this webhook admission plugin. Possible values are create, update, delete, or connect. Replace <operation> and <resource> with the appropriate values.

webhooks.failurePolicy

Specifies how the policy should proceed if the webhook server is unavailable. Replace <policy> with either Ignore (to unconditionally accept the request in case of a failure) or Fail (to deny the failed request). Using Ignore can result in unpredictable behavior for all clients.

Configuring dynamic admission

You can complete high-level steps to configure dynamic admission. These steps extend the admission chain by configuring a webhook admission plugin to call out to a webhook server.

The webhook server is also configured as an aggregated API server. This allows other OKD components to communicate with the webhook that uses internal credentials and facilitates testing that uses the oc command. Additionally, this enables role-based access control (RBAC) into the webhook and prevents token information from other API servers from being disclosed to the webhook.

Prerequisites
  • An OKD account with cluster administrator access.

  • The OKD OpenShift CLI (oc) installed.

  • A published webhook server container image.

Procedure
  1. Build a webhook server container image and make it available to the cluster by using an image registry.

  2. Create a local CA key and certificate and use them to sign the webhook server’s certificate signing request (CSR).

  3. Create a new project for webhook resources:

    $ oc new-project my-webhook-namespace
    • Note that the webhook server might expect a specific name.

  4. Define RBAC rules for the aggregated API service in a file called rbac.yaml:

    apiVersion: v1
    kind: List
    items:
    
    - apiVersion: rbac.authorization.k8s.io/v1
      kind: ClusterRoleBinding
      metadata:
        name: auth-delegator-my-webhook-namespace
      roleRef:
        kind: ClusterRole
        apiGroup: rbac.authorization.k8s.io
        name: system:auth-delegator
      subjects:
      - kind: ServiceAccount
        namespace: my-webhook-namespace
        name: server
    
    - apiVersion: rbac.authorization.k8s.io/v1
      kind: ClusterRole
      metadata:
        annotations:
        name: system:openshift:online:my-webhook-server
      rules:
      - apiGroups:
        - online.openshift.io
        resources:
        - namespacereservations
        verbs:
        - get
        - list
        - watch
    
    - apiVersion: rbac.authorization.k8s.io/v1
      kind: ClusterRole
      metadata:
        name: system:openshift:online:my-webhook-requester
      rules:
      - apiGroups:
        - admission.online.openshift.io
        resources:
        - namespacereservations
        verbs:
        - create
    
    - apiVersion: rbac.authorization.k8s.io/v1
      kind: ClusterRoleBinding
      metadata:
        name: my-webhook-server-my-webhook-namespace
      roleRef:
        kind: ClusterRole
        apiGroup: rbac.authorization.k8s.io
        name: system:openshift:online:my-webhook-server
      subjects:
      - kind: ServiceAccount
        namespace: my-webhook-namespace
        name: server
    
    - apiVersion: rbac.authorization.k8s.io/v1
      kind: RoleBinding
      metadata:
        namespace: kube-system
        name: extension-server-authentication-reader-my-webhook-namespace
      roleRef:
        kind: Role
        apiGroup: rbac.authorization.k8s.io
        name: extension-apiserver-authentication-reader
      subjects:
      - kind: ServiceAccount
        namespace: my-webhook-namespace
        name: server
    
    - apiVersion: rbac.authorization.k8s.io/v1
      kind: ClusterRole
      metadata:
        name: my-cluster-role
      rules:
      - apiGroups:
        - admissionregistration.k8s.io
        resources:
        - validatingwebhookconfigurations
        - mutatingwebhookconfigurations
        verbs:
        - get
        - list
        - watch
      - apiGroups:
        - ""
        resources:
        - namespaces
        verbs:
        - get
        - list
        - watch
    
    - apiVersion: rbac.authorization.k8s.io/v1
      kind: ClusterRoleBinding
      metadata:
        name: my-cluster-role
      roleRef:
        kind: ClusterRole
        apiGroup: rbac.authorization.k8s.io
        name: my-cluster-role
      subjects:
      - kind: ServiceAccount
        namespace: my-webhook-namespace
        name: server

    where:

    apiVersion: rbac.authorization.k8s.io/v1

    For ClusterRoleBinding, specifies authentication and authorization metadata to the webhook server API.

    apiVersion: rbac.authorization.k8s.io/v1

    For ClusterRole, specifies the webhook server that is allowed access cluster resources.

    rules.resources

    For ClusterRole, specifies the location to resources. This example points to the namespacereservations resource.

    apiVersion: rbac.authorization.k8s.io/v1

    For ClusterRole, specifies the enablement of the aggregated API server to create admission reviews.

    apiVersion: rbac.authorization.k8s.io/v1

    For ClusterRoleBinding, specifies the enablement of the webhook server to access cluster resources.

    apiVersion: rbac.authorization.k8s.io/v1

    For RoleBinding, specifies role binding to read the configuration for terminating authentication.

    apiVersion: rbac.authorization.k8s.io/v1

    For ClusterRole, specifies the default cluster role and cluster role bindings for an aggregated API server.

  5. Apply those RBAC rules to the cluster:

    $ oc auth reconcile -f rbac.yaml
  6. Create a YAML file called webhook-daemonset.yaml that is used to deploy a webhook as a daemon set server in a namespace:

    apiVersion: apps/v1
    kind: DaemonSet
    metadata:
      namespace: my-webhook-namespace
      name: server
      labels:
        server: "true"
    spec:
      selector:
        matchLabels:
          server: "true"
      template:
        metadata:
          name: server
          labels:
            server: "true"
        spec:
          serviceAccountName: server
          containers:
          - name: my-webhook-container
            image: <image_registry_username>/<image_path>:<tag>
            imagePullPolicy: IfNotPresent
            command:
            - <container_commands>
            ports:
            - containerPort: 8443
            volumeMounts:
            - mountPath: /var/serving-cert
              name: serving-cert
            readinessProbe:
              httpGet:
                path: /healthz
                port: 8443
                scheme: HTTPS
          volumes:
          - name: serving-cert
            secret:
              defaultMode: 420
              secretName: server-serving-cert

    where:

    spec.template.spec.name

    Specifies the container name. Note that the webhook server might expect a specific container name.

    spec.template.spec.image

    Specifies the path to a webhook server container image. Replace <image_registry_username>/<image_path>:<tag> with the appropriate value.

    spec.template.spec.command

    Specifies webhook container run commands. Replace <container_commands> with the appropriate value.

    spec.template.spec.ports.containerPort

    Specifies the target port within pods. This example uses port 8443.

    spec.template.spec.readinessProbe.port

    Specifies the port used by the readiness probe. This example uses port 8443.

  7. Deploy the daemon set:

    $ oc apply -f webhook-daemonset.yaml
  8. Define a secret for the service serving certificate signer, within a YAML file called webhook-secret.yaml:

    apiVersion: v1
    kind: Secret
    metadata:
      namespace: my-webhook-namespace
      name: server-serving-cert
    type: kubernetes.io/tls
    data:
      tls.crt: <server_certificate>
      tls.key: <server_key>

    where:

    data.tls.crt

    References the signed webhook server certificate. Replace <server_certificate> with the appropriate certificate in base64 format.

    data.tls.key

    References the signed webhook server key. Replace <server_key> with the appropriate key in base64 format.

  9. Create the secret:

    $ oc apply -f webhook-secret.yaml
  10. Define a service account and service, within a YAML file called webhook-service.yaml:

    apiVersion: v1
    kind: List
    items:
    
    - apiVersion: v1
      kind: ServiceAccount
      metadata:
        namespace: my-webhook-namespace
        name: server
    
    - apiVersion: v1
      kind: Service
      metadata:
        namespace: my-webhook-namespace
        name: server
        annotations:
          service.beta.openshift.io/serving-cert-secret-name: server-serving-cert
      spec:
        selector:
          server: "true"
        ports:
        - port: 443
          targetPort: 8443

    where:

    spec.ports.port

    Specifies the port that the service listens on. This example uses port 443.

    spec.ports.targetPort

    Specifies the target port within pods that the service forwards connections to. This example uses port 8443.

  11. Expose the webhook server within the cluster:

    $ oc apply -f webhook-service.yaml
  12. Define a custom resource definition for the webhook server, in a file called webhook-crd.yaml:

    apiVersion: apiextensions.k8s.io/v1beta1
    kind: CustomResourceDefinition
    metadata:
      name: namespacereservations.online.openshift.io
    spec:
      group: online.openshift.io
      version: v1alpha1
      scope: Cluster
      names:
        plural: namespacereservations
        singular: namespacereservation
        kind: NamespaceReservation

    where:

    metadata.name

    Reflects CustomResourceDefinition spec values and is in the format <plural>.<group>. This example uses the namespacereservations resource.

    spec.group

    Specifies the REST API group name.

    spec.version

    Specifies the REST API version name.

    spec.scope

    Specifies accepted values are Namespaced or Cluster.

    spec.names.plural

    Specifies the plural name to be included in URL.

    spec.names.singular

    Specifies the alias seen in oc output.

    spec.names.kind

    Specifies the reference for resource manifests.

  13. Apply the custom resource definition:

    $ oc apply -f webhook-crd.yaml
  14. Configure the webhook server also as an aggregated API server, within a file called webhook-api-service.yaml:

    apiVersion: apiregistration.k8s.io/v1beta1
    kind: APIService
    metadata:
      name: v1beta1.admission.online.openshift.io
    spec:
      caBundle: <ca_signing_certificate>
      group: admission.online.openshift.io
      groupPriorityMinimum: 1000
      versionPriority: 15
      service:
        name: server
        namespace: my-webhook-namespace
      version: v1beta1
    • The spec.caBundle field specifies a PEM-encoded CA certificate that signs the server certificate. This certificate is used by the webhook server. Replace <ca_signing_certificate> with the appropriate certificate in base64 format.

  15. Deploy the aggregated API service:

    $ oc apply -f webhook-api-service.yaml
  16. Define the webhook admission plugin configuration within a file called webhook-config.yaml. This example uses the validating admission plugin:

    apiVersion: admissionregistration.k8s.io/v1beta1
    kind: ValidatingWebhookConfiguration
    metadata:
      name: namespacereservations.admission.online.openshift.io
    webhooks:
    - name: namespacereservations.admission.online.openshift.io
      clientConfig:
        service:
          namespace: default
          name: kubernetes
          path: /apis/admission.online.openshift.io/v1beta1/namespacereservations
        caBundle: <ca_signing_certificate>
      rules:
      - operations:
        - CREATE
        apiGroups:
        - project.openshift.io
        apiVersions:
        - "*"
        resources:
        - projectrequests
      - operations:
        - CREATE
        apiGroups:
        - ""
        apiVersions:
        - "*"
        resources:
        - namespaces
      failurePolicy: Fail

    where:

    metadata.name

    Specifies the name for the ValidatingWebhookConfiguration object. This example uses the namespacereservations resource.

    webhooks.name

    Specifies the name of the webhook to call. This example uses the namespacereservations resource.

    name.clientConfig.service

    Enables access to the webhook server through the aggregated API.

    name.clientConfig.service.path

    Specifies the webhook URL used for admission requests. This example uses the namespacereservation resource.

    name.clientConfig.caBundle

    Specifies a PEM-encoded CA certificate that signs the server certificate that is used by the webhook server. Replace <ca_signing_certificate> with the appropriate certificate in base64 format.

  17. Deploy the webhook:

    $ oc apply -f webhook-config.yaml
  18. Verify that the webhook is functioning as expected. For example, if you have configured dynamic admission to reserve specific namespaces, confirm that requests to create those namespaces are rejected and that requests to create non-reserved namespaces succeed.