Logging into your cluster using Pinniped

Prerequisites

This how-to guide assumes that you have already configured the following Pinniped server-side components within your Kubernetes cluster(s):

1. If you would like to use the Pinniped Supervisor for federated authentication across multiple Kubernetes clusters then you have already:
   1. Installed the Pinniped Supervisor with working ingress.
   2. Configured a FederationDomain to issue tokens for your downstream clusters.
   3. Configured an OIDCIdentityProvider, LDAPIdentityProvider, or ActiveDirectoryIdentityProvider for the Supervisor as the source of your user’s identities. Various examples of configuring these resources can be found in these guides.
2. In each cluster for which you would like to use Pinniped for authentication, you have installed the Concierge.
3. In each cluster’s Concierge, you have configured an authenticator. For example, if you are using the Pinniped Supervisor, then you have configured each Concierge to use the Supervisor for authentication.

You should have also already installed the pinniped command-line client, which is used to generate Pinniped-compatible kubeconfig files, and is also a kubectl plugin to enable the Pinniped-based login flow.

Overview

1. A cluster admin uses Pinniped to generate a kubeconfig for each cluster, and shares the kubeconfig for each cluster with all users of that cluster.
2. A cluster user uses kubectl with the generated kubeconfig given to them by the cluster admin. kubectl interactively prompts the user to log in using their own unique identity.

Key advantages of using the Pinniped Supervisor

Although you can choose to use Pinniped without using the Pinniped Supervisor, there are several key advantages of choosing to use the Pinniped Supervisor to manage identity across fleets of Kubernetes clusters.

1. The Supervisor makes it easy to bring your own OIDC, LDAP, or Active Directory identity provider to act as the source of user identities. It also allows you to configure how identities and group memberships in the identity provider map to identities and group memberships in the Kubernetes clusters.

2. A generated kubeconfig for a cluster will be specific for that cluster, however it will not contain any specific user identity or credentials. This kubeconfig file can be safely shared with all cluster users. When the user runs kubectl commands using this kubeconfig, they will be interactively prompted to log in using their own unique identity from the identity provider configured in the Supervisor.

3. The Supervisor will provide a federated identity across all clusters that use the same FederationDomain. The user will be prompted by kubectl to interactively authenticate once per day, and then will be able to use all clusters from the same FederationDomain for the rest of the day without being asked to authenticate again. This federated identity is secure because behind the scenes the Supervisor is issuing very short-lived credentials that are uniquely scoped to each cluster.

Generate a Pinniped-compatible kubeconfig file

You will need to generate a Pinniped-compatible kubeconfig file for each cluster in which you have installed the Concierge. This requires admin-level access to each cluster, so this would typically be performed by the same user who installed the Concierge.

For each cluster, use pinniped get kubeconfig to generate the new kubeconfig file for that cluster.

It is typically sufficient to run this command with no arguments, aside from pointing the command at your admin kubeconfig. The command uses the same rules as kubectl to find your admin kubeconfig:

“By default, kubectl looks for a file named config in the $HOME/.kube directory. You can specify other kubeconfig files by setting the KUBECONFIG environment variable or by setting the --kubeconfig flag.”

For example, if your admin kubeconfig file were at the path $HOME/admin-kubeconfig.yaml, then you could use:

pinniped get kubeconfig \
  --kubeconfig "$HOME/admin-kubeconfig.yaml" > pinniped-kubeconfig.yaml

The new Pinniped-compatible kubeconfig YAML will be output as stdout, and can be redirected to a file.

Various default behaviors of pinniped get kubeconfig can be overridden using its command-line options. One flag of note is --upstream-identity-provider-flow browser_authcode to choose end-user kubectl login via a web browser (the default for OIDCIdentityProviders), and --upstream-identity-provider-flow cli_password to choose end-user kubectl login via CLI username/password prompts (the default for LDAPIdentityProviders and ActiveDirectoryIdentityProviders).

If the cluster is using a Pinniped Supervisor’s FederationDomain to provide authentication services, and if that FederationDomain allows multiple identity providers, then you will need to specify which identity provider you would like to use in the resulting kubeconfig with the --upstream-identity-provider-name and/or --upstream-identity-provider-type flags. You may call pinniped get kubeconfig multiple times to generate multiple kubeconfigs for the cluster.

By default, the resulting kubeconfig will contain the absolute path to the Pinniped CLI binary that was used to run pinniped get kubeconfig. However, this absolute path may not work on the local machines for your end users to whom you distribute the kubeconfig, since they may have the Pinniped CLI installed elsewhere. You can optionally set the absolute path or executable name in the resulting kubeconfig by using the --pinniped-cli-path argument. For example, when using pinniped get kubeconfig --pinniped-cli-path=pinniped then the resulting kubeconfig will include pinniped as the command to execute the Pinniped CLI, and during user login it will find the CLI by searching the user’s PATH for a binary named pinniped. This also works on Windows, where using pinniped as the command can find a binary named pinniped.exe in the user’s PATH. Alternatively, you could use pinniped get kubeconfig --pinniped-cli-path=/usr/local/bin/pinniped if you have reason to believe that your end users' machines will always have the Pinniped CLI installed in /usr/local/bin.

Use the generated kubeconfig with kubectl to access the cluster

A cluster user will typically be given a Pinniped-compatible kubeconfig by their cluster admin. They can use this kubeconfig with kubectl just like any other kubeconfig, as long as they have also installed the pinniped CLI tool at the same absolute path where it is referenced inside the kubeconfig’s YAML. The pinniped CLI will act as a kubectl plugin to manage the user’s authentication to the cluster.

For example, if the kubeconfig were saved at $HOME/pinniped-kubeconfig.yaml:

kubectl get namespaces \
  --kubeconfig "$HOME/pinniped-kubeconfig.yaml"

This command, when configured to use the Pinniped-compatible kubeconfig, will invoke the pinniped CLI behind the scenes as an ExecCredential plugin to authenticate the user to the cluster.

If the Pinniped Supervisor is used for authentication to that cluster, then the user’s authentication experience will depend on which type of identity provider was configured.

• For an OIDC identity provider, there are two supported client flows:

  1. When using the default browser-based flow, kubectl will open the user’s web browser and direct it to the login page of their OIDC Provider. This login flow is controlled by the provider, so it may include two-factor authentication or other features provided by the OIDC Provider. If the user’s browser is not available, then kubectl will instead print a URL which can be visited in a browser (potentially on a different computer) to complete the authentication.

  2. When using the optional CLI-based flow, kubectl will interactively prompt the user for their username and password at the CLI. Alternatively, the user can set the environment variables PINNIPED_USERNAME and PINNIPED_PASSWORD for the kubectl process to avoid the interactive prompts. Note that the optional CLI-based flow must be enabled by the administrator in the OIDCIdentityProvider configuration before use (see allowPasswordGrant in the API docs for more details).

• For LDAP and Active Directory identity providers, there are also two supported client flows:

  1. When using the default CLI-based flow, kubectl will interactively prompt the user for their username and password at the CLI. Alternatively, the user can set the environment variables PINNIPED_USERNAME and PINNIPED_PASSWORD for the kubectl process to avoid the interactive prompts.

  2. When using the optional browser-based flow, kubectl will open the user’s web browser and direct it to a login page hosted by the Pinniped Supervisor. When the user enters their username and password, the Supervisor will authenticate the user using the LDAP or Active Directory provider. If the user’s browser is not available, then kubectl will instead print a URL which can be visited in a browser (potentially on a different computer) to complete the authentication. Unlike the optional flow for OIDC providers described above, this optional flow does not need to be configured in the LDAPIdentityProvider or ActiveDirectoryIdentityProvider resource, so it is always available for end-users.

The flow selected by the --upstream-identity-provider-flow CLI flag may be overridden by using the PINNIPED_UPSTREAM_IDENTITY_PROVIDER_FLOW environment variable for the CLI at runtime. This environment variable may be set to the same values as the CLI flag (browser_authcode or cli_password). This allows a user to switch flows based on their needs without editing their kubeconfig file.

Once the user completes authentication, the kubectl command will automatically continue and complete the user’s requested command. For the example above, kubectl would list the cluster’s namespaces.

Authorization

Pinniped provides authentication (usernames and group memberships) but not authorization. Kubernetes authorization is often provided by the Kubernetes RBAC system on each cluster.

In the example above, if the user gets an access denied error, then they may need authorization to list namespaces. For example, an admin could grant the user “edit” access to all cluster resources via the user’s username:

kubectl create clusterrolebinding my-user-can-edit \
  --clusterrole edit \
  --user my-username@example.com

Alternatively, an admin could create role bindings based on the group membership of the users in the upstream identity provider, for example:

kubectl create clusterrolebinding my-auditors \
  --clusterrole view \
  --group auditors

Session and credential caching by the CLI

Temporary session credentials such as ID, access, and refresh tokens are stored in:

• $HOME/.config/pinniped/sessions.yaml (macOS/Linux)
• %USERPROFILE%/.config/pinniped/sessions.yaml (Windows).

Temporary cluster credentials such mTLS client certificates are stored in:

• $HOME/.config/pinniped/credentials.yaml (macOS/Linux)
• %USERPROFILE%/.config/pinniped/credentials.yaml (Windows).

Deleting the contents of these directories is equivalent to performing a client-side logout.