IBM Cloud Docs
Multicloud apps with Istio

Multicloud apps with Istio

By using the App Identity and Access Adapter, you can centralize all your identity management in a single place.

The App Identity and Access Adapter is not currently supported.

Because enterprises use clouds from multiple providers or a combination of on and off-premise solutions, heterogeneous deployment models can help you to preserve existing infrastructure and avoid vendor lock-in. The Adapter can be configured to work with any OIDC-compliant identity provider, such as App ID. The service enables the Adapter to control authentication and authorization policies in all environments, including front end and backend applications. And, it does it all without any change to your code or the need to redeploy your application.

What can the App Identity and Access Adapter do for you? Check out the following video to learn more.

Multicloud architecture

A multicloud computing environment combines multiple cloud and / or private computing environments into a single network architecture. By distributing workloads across multiple environments, you might find improved resiliency, flexibility, and greater cost-efficiency. To achieve the benefits, it's common to use a container-based application with an orchestration layer, such as Kubernetes.

App Identity and Access Adapter architecture diagram
Figure 1. Multicloud deployment - achieved with the App Identity and Access Adapter

Understanding Istio and the Adapter

Istio is an open source service mesh that layers transparently onto existing distributed applications that can integrate with Kubernetes. To reduce the complexity of deployments, Istio provides behavioral insights and operational control over the service mesh as a whole. When App ID is combined with Istio, it becomes a scalable, integrated identity solution for multicloud architectures that does not require any custom application code changes. For more information, check out What is Istio.

Istio uses an Envoy proxy sidecar to mediate all inbound and outbound traffic for all services in the service mesh. By using the proxy, Istio extracts information about traffic, also known as telemetry, that is sent to the Istio component called Mixer to enforce policy decisions. The App Identity and Access Adapter extends the Mixer functionality by analyzing the telemetry (attributes) against custom policies to control identity and access management into and across the service mesh. The access management policies are linked to particular Kubernetes services and can be finely tuned to specific service endpoints. For more information about policies and telemetry, see the Istio documentation.

Due to an Istio limitation, the App Identity and Access Adapter currently stores user session information internally and does not persist the information across replicas or over failover configurations. When you use the Adapter, limit your workloads to a single replica until the limitation is addressed.

Protecting front-end apps

If you're using a browser-based application, you can use the Open ID Connect (OIDC) / OAuth 2.0 authorization_grant flow to authenticate your users. When an unauthenticated user is detected, they are automatically redirected to the authentication page. When the authentication completes, the browser is redirected to an implicit /oidc/callback endpoint where the Adapter intercepts the request. At this point, the Adapter obtains tokens from the identity provider and then redirects the user back to their originally requested URL.

To view the user session information, including the session tokens, you can look in the Authorization header.

Authorization: Bearer <accessToken> <IDToken>

You can also log out authenticated users. When an authenticated user accesses any protected endpoint with oidc/logout appended as shown in the following example, they are logged out.

https://myhost/path/oidc/logout

If needed, a refresh token can be used to automatically acquire new access and identity tokens without your user's needing to reauthenticate. If the configured identity provider returns a refresh token, it's persisted in the session and used to retrieve new tokens when the identity token expires.

Protecting backend apps

The Adapter can be used in collaboration with the OAuth 2.0 JWT Bearer flow to protect service APIs by validating JWT Bearer tokens. The Bearer authorization flow expects a request to contain an Authorization header with a valid access token and an optional identity token. The expected header structure is Authorization=Bearer {access_token} [{id_token}]. Unauthenticated clients are returned an HTTP 401 response status with a list of the scopes that are needed to obtain authorization. If the tokens are invalid or expired, the API strategy returns an HTTP 401 response with an optional error component that says Www-Authenticate=Bearer scope="{scope}" error="{error}".

For more information about tokens and how they're used, see understanding tokens.

Before you begin

Before you get started, be sure that you installed the following prerequisites.

Installing the Adapter

To install the chart, initialize Helm in your cluster, define the options that you want to use, and then run the installation command.

  1. If you're working with IBM Cloud Kubernetes Service, be sure to log in and set the context for your cluster.

  2. Verify that you enabled Istio policy enforcement. If not, turn it on.

  3. Add the repository.

    helm repo add appidentityandaccessAdapter https://raw.githubusercontent.com/ibm-cloud-security/app-identity-and-access-Adapter/master/helm/appidentityandaccessAdapter
    
  4. Install the chart.

    helm install --name appidentityandaccessAdapter appidentityandaccessAdapter/appidentityandaccessAdapter
    

    You can specify an image tag during installation by setting the image.tag flag. For example, --set image.tag=0.5.0. You can also install the chart locally. To do so, clone the repo by running git clone git@github.com:ibm-cloud-security/app-identity-and-access-Adapter.git before you run the installation command.

Applying an authorization and authentication policy

An authentication or authorization policy is a set of conditions that must be met before a request can access a resource access. By defining an identity provider's service configuration and a policy that outlines when a particular flow must be used, you can control access to any resource in your service mesh. To see example CRDs, check out the samples directory.

To create a policy:

  1. Define a configuration.
  2. Register the endpoint.

Defining a configuration

Depending on whether you're protecting front end or backend applications, create a policy configuration with one of the following options.

  • For front-end applications: Browser-based applications that require user authentication can be configured to use the OIDC / OAuth 2.0 authentication flow. To define an OidcConfig CRD containing the client used to facilitate the authentication flow with the Identity provider, use the following example as a guide.

    apiVersion: "security.cloud.ibm.com/v1"
    kind: OidcConfig
    metadata:
       name:      oidc-provider-config
       namespace: sample-namespace
    spec:
       discoveryUrl: https://us-south.appid.cloud.ibm.com/oauth/v4/<tenantID>/.well-known/openid-configuration
       clientId:     <clientID>
       clientSecret: <randomlyGeneratedClientSecret>
       clientSecretRef:
             name: <nameOfKubeSecret>
             key: <keyInKubeSecret>
    
    Table 1. YAML configuration file components explained
    Field Type Required Description
    discoveryUrl string Yes A well-known endpoint that provides a JSON document of OIDC/OAuth 2.0 configuration information.
    clientId string Yes An identifier for the client that is used for authentication.
    clientSecret string *No A plain text secret that is used to authenticate the client. If not provided, a clientSecretRef must exist.
    clientSecretRef object No A reference secret that is used to authenticate the client. The reference can be used in place of the clientSecret.
    clientSecretRef.name string Yes The name of the Kubernetes Secret that contains the clientSecret.
    clientSecretRef.key string Yes The field within the Kubernetes Secret that holds the clientSecret.
  • For backend applications: The OAuth 2.0 Bearer token spec defines a pattern for protecting APIs by using JSON Web Tokens (JWTs). By using the following configuration as an example, define a JwtConfig CRD that contains the public key resource, which is used to validate token signatures.

    apiVersion: "security.cloud.ibm.com/v1"
    kind: JwtConfig
    metadata:
       name:      jwt-config
       namespace: sample-app
    spec:
       jwksUrl: https://us-south.appid.cloud.ibm.com/oauth/v4/<tenantID>/publickeys
    

Registering application endpoints

Register application endpoints within a Policy CRD to validate incoming requests and enforce authentication rules. Each Policy applies exclusively to the Kubernetes namespace in which the object lives and can specify the services, paths, and methods that you want to protect.

apiVersion: "security.cloud.ibm.com/v1"
kind: Policy
metadata:
  name:      samplepolicy
  namespace: sample-app
spec:
  targets:
    -
      serviceName: <svcSampleApp>
      paths:
        - exact: /web/home
          method: ALL
          policies:
            - policyType: oidc
              config: <oidcProviderConfig>
              rules:
                - claim: scope
                  match: ALL
                  source: access_token
                  values:
                    - appid_default
                    - openid
                - claim: amr
                  match: ANY
                  source: id_token
                  values:
                    - cloud_directory
                    - google

        - exact: /web/user
          method: GET
          policies:
            - policyType: oidc
              config: <oidcProviderConfig>
              redirectUri: https://github.com/ibm-cloud-security/app-identity-and-access-Adapter
        - prefix: /
          method: ALL
          policies:
            -
              policyType: jwt
              config: <jwtConfig>
Table 2. Understanding the service object components
Service Object Type Required Description
serviceName string Yes The name of Kubernetes service in the Policy namespace that you want to protect.
paths array[Path Object] Yes A list of path objects that define the endpoints that you want to protect. If not specified, all paths are protected.
Table 2. Understanding the path object components
Path Object Type Required Description
exact or prefix string Yes The path that you want to apply the policies on. Options include exact and prefix. exact matches the provided endpoints exactly with the last / trimmed. prefix matches the endpoints that begin with the route prefix that you provide.
method enum No The HTTP method protected. Valid options ALL, GET, PUT, POST, DELETE, PATCH - Defaults to ALL:
policies array[Policy] No The OIDC/JWT policies that you want to apply.
Table 2. Understanding the policy object components
Policy Object Type Required Description
policyType enum Yes The type of OIDC policy. Options include: jwt or oidc.
config string Yes The name of the provider config that you want to use.
redirectUri string No The URL that you want the user to be redirected after successful authentication, default: the original request URL.
rules array[Rule] No The set of rules that you want to use for token validation.
Table 2. Understanding the policy object components
Rule Object Type Required Description
claim string Yes The claim that you want to validate.
match enum No The criteria required for claim validation. Options include: ALL, ANY, or NOT. The default is set to ALL.
source enum No The token where you want to apply the rule. Options include: access_token or id_token. The default is set to access_token.
values array[string] Yes The required set of values for validation.

Deleting the Adapter

To remove the Adapter and all associated CRDs, you must delete the Helm chart and the associated signing and encryption keys.

helm delete --purge appidentityandaccessAdapter
kubectl delete secret appidentityandaccessAdapter-keys -n istio-system

Configuring logging

By default, logs are styled as JSON and provided at an info visibility level to provide for ease of integration with external logging systems. To update the logging configuration, you can use the Helm chart. Supported logging levels include range [-1, 7] as shown in Zap core. For more information about the levels, see the Zap core documentation.

When you're manually viewing JSON logs, you might want to tail the logs and "pretty print" them by using jq.

Adapter

To see the Adapter logs, you can use kubectl or access the pod from the appidentityandaccessAdapter pod from the Kubernetes console.

alias Adapter_logs="kubectl -n istio-system logs -f $(kubectl -n istio-system get pods -lapp=appidentityandaccessAdapter -o jsonpath='{.items[0].metadata.name}')"
Adapter_logs | jq

Mixer

If the Adapter does not appear to receive requests, check the Mixer logs to ensure that it connected successfully to the Adapter.

alias mixer_logs="kubectl -n istio-system logs -f $(kubectl -n istio-system get pods -lapp=telemetry -o jsonpath='{.items[0].metadata.name}') -c mixer"
mixer_logs | jq