Overview

The authentication layer identifies the user associated with requests to the OKD API. The authorization layer then uses information about the requesting user to determine if the request should be allowed.

As an administrator, you can configure authentication using a master configuration file.

Users and Groups

A user in OKD is an entity that can make requests to the OKD API. Typically, this represents the account of a developer or administrator that is interacting with OKD.

A user can be assigned to one or more groups, each of which represent a certain set of users. Groups are useful when managing authorization policies to grant permissions to multiple users at once, for example allowing access to objects within a project, versus granting them to users individually.

In addition to explicitly defined groups, there are also system groups, or virtual groups, that are automatically provisioned by OpenShift. These can be seen when viewing cluster bindings.

In the default set of virtual groups, note the following in particular:

Virtual Group Description

system:authenticated

Automatically associated with all authenticated users.

system:authenticated:oauth

Automatically associated with all users authenticated with an OAuth access token.

system:unauthenticated

Automatically associated with all unauthenticated users.

API Authentication

Requests to the OKD API are authenticated using the following methods:

OAuth Access Tokens
  • Obtained from the OKD OAuth server using the <master>/oauth/authorize and <master>/oauth/token endpoints.

  • Sent as an Authorization: Bearer…​ header

  • Sent as an access_token=…​ query parameter for websocket requests prior to OKD server version 3.6.

  • Sent as a websocket subprotocol header in the form base64url.bearer.authorization.k8s.io.<base64url-encoded-token> for websocket requests in OKD server version 3.6 and later.

X.509 Client Certificates
  • Requires a HTTPS connection to the API server.

  • Verified by the API server against a trusted certificate authority bundle.

  • The API server creates and distributes certificates to controllers to authenticate themselves.

Any request with an invalid access token or an invalid certificate is rejected by the authentication layer with a 401 error.

If no access token or certificate is presented, the authentication layer assigns the system:anonymous virtual user and the system:unauthenticated virtual group to the request. This allows the authorization layer to determine which requests, if any, an anonymous user is allowed to make.

Impersonation

A request to the OKD API can include an Impersonate-User header, which indicates that the requester wants to have the request handled as though it came from the specified user. You impersonate a user by adding the --as=<user> flag to requests.

Before User A can impersonate User B, User A is authenticated. Then, an authorization check occurs to ensure that User A is allowed to impersonate the user named User B. If User A is requesting to impersonate a service account, system:serviceaccount:namespace:name, OKD confirms that User A can impersonate the serviceaccount named name in namespace. If the check fails, the request fails with a 403 (Forbidden) error code.

By default, project administrators and editors can impersonate service accounts in their namespace. The sudoers role allows a user to impersonate system:admin, which in turn has cluster administrator permissions. The ability to impersonate system:admin grants some protection against typos, but not security, for someone administering the cluster. For example, running oc delete nodes --all fails, but running oc delete nodes --all --as=system:admin succeeds. You can grant a user that permission by running this command:

$ oc create clusterrolebinding <any_valid_name> --clusterrole=sudoer --user=<username>

If you need to create a project request on behalf of a user, include the --as=<user> --as-group=<group1> --as-group=<group2> flags in your command. Because system:authenticated:oauth is the only bootstrap group that can create project requests, you must impersonate that group, as shown in the following example:

$ oc new-project <project> --as=<user> \
--as-group=system:authenticated --as-group=system:authenticated:oauth

OAuth

The OKD master includes a built-in OAuth server. Users obtain OAuth access tokens to authenticate themselves to the API.

When a person requests a new OAuth token, the OAuth server uses the configured identity provider to determine the identity of the person making the request.

It then determines what user that identity maps to, creates an access token for that user, and returns the token for use.

OAuth Clients

Every request for an OAuth token must specify the OAuth client that will receive and use the token. The following OAuth clients are automatically created when starting the OKD API:

OAuth Client Usage

openshift-web-console

Requests tokens for the web console.

openshift-browser-client

Requests tokens at <master>/oauth/token/request with a user-agent that can handle interactive logins.

openshift-challenging-client

Requests tokens with a user-agent that can handle WWW-Authenticate challenges.

To register additional clients:

$ oc create -f <(echo '
kind: OAuthClient
apiVersion: oauth.openshift.io/v1
metadata:
 name: demo (1)
secret: "..." (2)
redirectURIs:
 - "http://www.example.com/" (3)
grantMethod: prompt (4)
')
1 The name of the OAuth client is used as the client_id parameter when making requests to <master>/oauth/authorize and <master>/oauth/token.
2 The secret is used as the client_secret parameter when making requests to <master>/oauth/token.
3 The redirect_uri parameter specified in requests to <master>/oauth/authorize and <master>/oauth/token must be equal to (or prefixed by) one of the URIs in redirectURIs.
4 The grantMethod is used to determine what action to take when this client requests tokens and has not yet been granted access by the user. Uses the same values seen in Grant Options.

Service Accounts as OAuth Clients

A service account can be used as a constrained form of OAuth client. Service accounts can only request a subset of scopes that allow access to some basic user information and role-based power inside of the service account’s own namespace:

  • user:info

  • user:check-access

  • role:<any_role>:<serviceaccount_namespace>

  • role:<any_role>:<serviceaccount_namespace>:!

When using a service account as an OAuth client:

  • client_id is system:serviceaccount:<serviceaccount_namespace>:<serviceaccount_name>.

  • client_secret can be any of the API tokens for that service account. For example:

    $ oc sa get-token <serviceaccount_name>
  • To get WWW-Authenticate challenges, set an serviceaccounts.openshift.io/oauth-want-challenges annotation on the service account to true.

  • redirect_uri must match an annotation on the service account. Redirect URIs for Service Accounts as OAuth Clients provides more information.

Redirect URIs for Service Accounts as OAuth Clients

Annotation keys must have the prefix serviceaccounts.openshift.io/oauth-redirecturi. or serviceaccounts.openshift.io/oauth-redirectreference. such as:

serviceaccounts.openshift.io/oauth-redirecturi.<name>

In its simplest form, the annotation can be used to directly specify valid redirect URIs. For example:

"serviceaccounts.openshift.io/oauth-redirecturi.first":  "https://example.com"
"serviceaccounts.openshift.io/oauth-redirecturi.second": "https://other.com"

The first and second postfixes in the above example are used to separate the two valid redirect URIs.

In more complex configurations, static redirect URIs may not be enough. For example, perhaps you want all ingresses for a route to be considered valid. This is where dynamic redirect URIs via the serviceaccounts.openshift.io/oauth-redirectreference. prefix come into play.

For example:

"serviceaccounts.openshift.io/oauth-redirectreference.first": "{\"kind\":\"OAuthRedirectReference\",\"apiVersion\":\"v1\",\"reference\":{\"kind\":\"Route\",\"name\":\"jenkins\"}}"

Since the value for this annotation contains serialized JSON data, it is easier to see in an expanded format:

{
  "kind": "OAuthRedirectReference",
  "apiVersion": "v1",
  "reference": {
    "kind": "Route",
    "name": "jenkins"
  }
}

Now you can see that an OAuthRedirectReference allows us to reference the route named jenkins. Thus, all ingresses for that route will now be considered valid. The full specification for an OAuthRedirectReference is:

{
  "kind": "OAuthRedirectReference",
  "apiVersion": "v1",
  "reference": {
    "kind": ..., (1)
    "name": ..., (2)
    "group": ... (3)
  }
}
1 kind refers to the type of the object being referenced. Currently, only route is supported.
2 name refers to the name of the object. The object must be in the same namespace as the service account.
3 group refers to the group of the object. Leave this blank, as the group for a route is the empty string.

Both annotation prefixes can be combined to override the data provided by the reference object. For example:

"serviceaccounts.openshift.io/oauth-redirecturi.first":  "custompath"
"serviceaccounts.openshift.io/oauth-redirectreference.first": "{\"kind\":\"OAuthRedirectReference\",\"apiVersion\":\"v1\",\"reference\":{\"kind\":\"Route\",\"name\":\"jenkins\"}}"

The first postfix is used to tie the annotations together. Assuming that the jenkins route had an ingress of https://example.com, now https://example.com/custompath is considered valid, but https://example.com is not. The format for partially supplying override data is as follows:

Type Syntax

Scheme

"https://"

Hostname

"//website.com"

Port

"//:8000"

Path

"examplepath"

Specifying a host name override will replace the host name data from the referenced object, which is not likely to be desired behavior.

Any combination of the above syntax can be combined using the following format:

<scheme:>//<hostname><:port>/<path>

The same object can be referenced more than once for more flexibility:

"serviceaccounts.openshift.io/oauth-redirecturi.first":  "custompath"
"serviceaccounts.openshift.io/oauth-redirectreference.first": "{\"kind\":\"OAuthRedirectReference\",\"apiVersion\":\"v1\",\"reference\":{\"kind\":\"Route\",\"name\":\"jenkins\"}}"
"serviceaccounts.openshift.io/oauth-redirecturi.second":  "//:8000"
"serviceaccounts.openshift.io/oauth-redirectreference.second": "{\"kind\":\"OAuthRedirectReference\",\"apiVersion\":\"v1\",\"reference\":{\"kind\":\"Route\",\"name\":\"jenkins\"}}"

Assuming that the route named jenkins has an ingress of https://example.com, then both https://example.com:8000 and https://example.com/custompath are considered valid.

Static and dynamic annotations can be used at the same time to achieve the desired behavior:

"serviceaccounts.openshift.io/oauth-redirectreference.first": "{\"kind\":\"OAuthRedirectReference\",\"apiVersion\":\"v1\",\"reference\":{\"kind\":\"Route\",\"name\":\"jenkins\"}}"
"serviceaccounts.openshift.io/oauth-redirecturi.second": "https://other.com"

API Events for OAuth

In some cases the API server returns an unexpected condition error message that is difficult to debug without direct access to the API master log. The underlying reason for the error is purposely obscured in order to avoid providing an unauthenticated user with information about the server’s state.

A subset of these errors is related to service account OAuth configuration issues. These issues are captured in events that can be viewed by non-administrator users. When encountering an unexpected condition server error during OAuth, run oc get events to view these events under ServiceAccount.

The following example warns of a service account that is missing a proper OAuth redirect URI:

$ oc get events | grep ServiceAccount
1m         1m          1         proxy                    ServiceAccount                                  Warning   NoSAOAuthRedirectURIs   service-account-oauth-client-getter   system:serviceaccount:myproject:proxy has no redirectURIs; set serviceaccounts.openshift.io/oauth-redirecturi.<some-value>=<redirect> or create a dynamic URI using serviceaccounts.openshift.io/oauth-redirectreference.<some-value>=<reference>

Running oc describe sa/<service-account-name> reports any OAuth events associated with the given service account name.

$ oc describe sa/proxy | grep -A5 Events
Events:
  FirstSeen     LastSeen        Count   From                                    SubObjectPath   Type            Reason                  Message
  ---------     --------        -----   ----                                    -------------   --------        ------                  -------
  3m            3m              1       service-account-oauth-client-getter                     Warning         NoSAOAuthRedirectURIs   system:serviceaccount:myproject:proxy has no redirectURIs; set serviceaccounts.openshift.io/oauth-redirecturi.<some-value>=<redirect> or create a dynamic URI using serviceaccounts.openshift.io/oauth-redirectreference.<some-value>=<reference>

The following is a list of the possible event errors:

No redirect URI annotations or an invalid URI is specified

Reason                  Message
NoSAOAuthRedirectURIs   system:serviceaccount:myproject:proxy has no redirectURIs; set serviceaccounts.openshift.io/oauth-redirecturi.<some-value>=<redirect> or create a dynamic URI using serviceaccounts.openshift.io/oauth-redirectreference.<some-value>=<reference>

Invalid route specified

Reason                  Message
NoSAOAuthRedirectURIs   [routes.route.openshift.io "<name>" not found, system:serviceaccount:myproject:proxy has no redirectURIs; set serviceaccounts.openshift.io/oauth-redirecturi.<some-value>=<redirect> or create a dynamic URI using serviceaccounts.openshift.io/oauth-redirectreference.<some-value>=<reference>]

Invalid reference type specified

Reason                  Message
NoSAOAuthRedirectURIs   [no kind "<name>" is registered for version "v1", system:serviceaccount:myproject:proxy has no redirectURIs; set serviceaccounts.openshift.io/oauth-redirecturi.<some-value>=<redirect> or create a dynamic URI using serviceaccounts.openshift.io/oauth-redirectreference.<some-value>=<reference>]

Missing SA tokens

Reason                  Message
NoSAOAuthTokens         system:serviceaccount:myproject:proxy has no tokens
Sample API Event Caused by a Possible Misconfiguration

The following steps represent one way a user could get into a broken state and how to debug or fix the issue:

  1. Create a project utilizing a service account as an OAuth client.

    1. Create YAML for a proxy service account object and ensure it uses the route proxy:

      vi serviceaccount.yaml

      Add the following sample code:

      apiVersion: v1
      kind: ServiceAccount
      metadata:
        name: proxy
        annotations:
          serviceaccounts.openshift.io/oauth-redirectreference.primary: '{"kind":"OAuthRedirectReference","apiVersion":"v1","reference":{"kind":"Route","name":"proxy"}}'
    2. Create YAML for a route object to create a secure connection to the proxy:

      vi route.yaml

      Add the following sample code:

      apiVersion: route.openshift.io/v1
      kind: Route
      metadata:
        name: proxy
      spec:
        to:
          name: proxy
        tls:
          termination: Reencrypt
      apiVersion: v1
      kind: Service
      metadata:
        name: proxy
        annotations:
          service.alpha.openshift.io/serving-cert-secret-name: proxy-tls
      spec:
        ports:
        - name: proxy
          port: 443
          targetPort: 8443
        selector:
          app: proxy
    3. Create a YAML for a deployment configuration to launch a proxy as a sidecar:

      vi proxysidecar.yaml

      Add the following sample code:

      apiVersion: extensions/v1beta1
      kind: Deployment
      metadata:
        name: proxy
      spec:
        replicas: 1
        selector:
          matchLabels:
            app: proxy
        template:
          metadata:
            labels:
              app: proxy
          spec:
            serviceAccountName: proxy
            containers:
            - name: oauth-proxy
              image: openshift/oauth-proxy:v1.0.0
              imagePullPolicy: IfNotPresent
              ports:
              - containerPort: 8443
                name: public
              args:
              - --https-address=:8443
              - --provider=openshift
              - --openshift-service-account=proxy
              - --upstream=http://localhost:8080
              - --tls-cert=/etc/tls/private/tls.crt
              - --tls-key=/etc/tls/private/tls.key
              - --cookie-secret=SECRET
              volumeMounts:
              - mountPath: /etc/tls/private
                name: proxy-tls
      
            - name: app
              image: openshift/hello-openshift:latest
            volumes:
            - name: proxy-tls
              secret:
                secretName: proxy-tls
    4. Create the objects

      oc create -f serviceaccount.yaml
      oc create -f route.yaml
      oc create -f proxysidecar.yaml
  2. Run oc edit sa/proxy to edit the service account and change the serviceaccounts.openshift.io/oauth-redirectreference annotation to point to a Route that does not exist.

    apiVersion: v1
    imagePullSecrets:
    - name: proxy-dockercfg-08d5n
    kind: ServiceAccount
    metadata:
      annotations:
        serviceaccounts.openshift.io/oauth-redirectreference.primary: '{"kind":"OAuthRedirectReference","apiVersion":"v1","reference":{"kind":"Route","name":"notexist"}}'
    ...
  3. Review the OAuth log for the service to locate the server error:

    The authorization server encountered an unexpected condition that prevented it from fulfilling the request.
  4. Run oc get events to view the ServiceAccount event:

    oc get events | grep ServiceAccount
    
    23m        23m         1         proxy                    ServiceAccount                                  Warning   NoSAOAuthRedirectURIs   service-account-oauth-client-getter   [routes.route.openshift.io "notexist" not found, system:serviceaccount:myproject:proxy has no redirectURIs; set serviceaccounts.openshift.io/oauth-redirecturi.<some-value>=<redirect> or create a dynamic URI using serviceaccounts.openshift.io/oauth-redirectreference.<some-value>=<reference>]

Integrations

All requests for OAuth tokens involve a request to <master>/oauth/authorize. Most authentication integrations place an authenticating proxy in front of this endpoint, or configure OKD to validate credentials against a backing identity provider. Requests to <master>/oauth/authorize can come from user-agents that cannot display interactive login pages, such as the CLI. Therefore, OKD supports authenticating using a WWW-Authenticate challenge in addition to interactive login flows.

If an authenticating proxy is placed in front of the <master>/oauth/authorize endpoint, it should send unauthenticated, non-browser user-agents WWW-Authenticate challenges, rather than displaying an interactive login page or redirecting to an interactive login flow.

To prevent cross-site request forgery (CSRF) attacks against browser clients, Basic authentication challenges should only be sent if a X-CSRF-Token header is present on the request. Clients that expect to receive Basic WWW-Authenticate challenges should set this header to a non-empty value.

If the authenticating proxy cannot support WWW-Authenticate challenges, or if OKD is configured to use an identity provider that does not support WWW-Authenticate challenges, users can visit <master>/oauth/token/request using a browser to obtain an access token manually.

OAuth Server Metadata

Applications running in OKD may need to discover information about the built-in OAuth server. For example, they may need to discover what the address of the <master> server is without manual configuration. To aid in this, OKD implements the IETF OAuth 2.0 Authorization Server Metadata draft specification.

Thus, any application running inside the cluster can issue a GET request to https://openshift.default.svc/.well-known/oauth-authorization-server to fetch the following information:

{
  "issuer": "https://<master>", (1)
  "authorization_endpoint": "https://<master>/oauth/authorize", (2)
  "token_endpoint": "https://<master>/oauth/token", (3)
  "scopes_supported": [ (4)
    "user:full",
    "user:info",
    "user:check-access",
    "user:list-scoped-projects",
    "user:list-projects"
  ],
  "response_types_supported": [ (5)
    "code",
    "token"
  ],
  "grant_types_supported": [ (6)
    "authorization_code",
    "implicit"
  ],
  "code_challenge_methods_supported": [ (7)
    "plain",
    "S256"
  ]
}
1 The authorization server’s issuer identifier, which is a URL that uses the https scheme and has no query or fragment components. This is the location where .well-known RFC 5785 resources containing information about the authorization server are published.
2 URL of the authorization server’s authorization endpoint. See RFC 6749.
3 URL of the authorization server’s token endpoint. See RFC 6749.
4 JSON array containing a list of the OAuth 2.0 RFC 6749 scope values that this authorization server supports. Note that not all supported scope values are advertised.
5 JSON array containing a list of the OAuth 2.0 response_type values that this authorization server supports. The array values used are the same as those used with the response_types parameter defined by "OAuth 2.0 Dynamic Client Registration Protocol" in RFC 7591.
6 JSON array containing a list of the OAuth 2.0 grant type values that this authorization server supports. The array values used are the same as those used with the grant_types parameter defined by OAuth 2.0 Dynamic Client Registration Protocol in RFC 7591.
7 JSON array containing a list of PKCE RFC 7636 code challenge methods supported by this authorization server. Code challenge method values are used in the code_challenge_method parameter defined in Section 4.3 of RFC 7636. The valid code challenge method values are those registered in the IANA PKCE Code Challenge Methods registry. See IANA OAuth Parameters.

Obtaining OAuth Tokens

The OAuth server supports standard authorization code grant and the implicit grant OAuth authorization flows.

When requesting an OAuth token using the implicit grant flow (response_type=token) with a client_id configured to request WWW-Authenticate challenges (like openshift-challenging-client), these are the possible server responses from /oauth/authorize, and how they should be handled:

Status Content Client response

302

Location header containing an access_token parameter in the URL fragment (RFC 4.2.2)

Use the access_token value as the OAuth token

302

Location header containing an error query parameter (RFC 4.1.2.1)

Fail, optionally surfacing the error (and optional error_description) query values to the user

302

Other Location header

Follow the redirect, and process the result using these rules

401

WWW-Authenticate header present

Respond to challenge if type is recognized (e.g. Basic, Negotiate, etc), resubmit request, and process the result using these rules

401

WWW-Authenticate header missing

No challenge authentication is possible. Fail and show response body (which might contain links or details on alternate methods to obtain an OAuth token)

Other

Other

Fail, optionally surfacing response body to the user

Authentication Metrics for Prometheus

OKD captures the following Prometheus system metrics during authentication attempts:

  • openshift_auth_basic_password_count counts the number of oc login user name and password attempts.

  • openshift_auth_basic_password_count_result counts the number of oc login user name and password attempts by result (success or error).

  • openshift_auth_form_password_count counts the number of web console login attempts.

  • openshift_auth_form_password_count_result counts the number of web console login attempts by result (success or error).

  • openshift_auth_password_total counts the total number of oc login and web console login attempts.