ocp-sso-token 1.2.1

Obtain an OCP OAuth token for an SSO IdP with Kerberos support

Non-interactive CLI login for OpenShift clusters via OpenID-connected IdP with Kerberos support

This package can be found

• on Copr as @cki/ocp-sso-token
• on PyPI as ocp-sso-token

The Copr packages are provided for all active Fedora releases, CentOS Stream/EPEL Next 9 and RHEL/EPEL 9.

The code is tested on the supported Python versions 3.8, 3.9, 3.10, 3.11 and 3.12.

Quickstart

# from Copr...
sudo dnf copr enable @cki/ocp-sso-token
sudo dnf install ocp-sso-token
# ...or from PyPI
pip install ocp-sso-token

# get a Kerberos token
kinit [e-mail]
# create a kubectl context with a temporary token via SSO
ocp-sso-token https://API.CLUSTER:6443 --context CONTEXT --namespace NAMESPACE
# and try it out!
oc --context CONTEXT get pod

Problem: several manual steps to log into an OpenShift cluster via OIDC without ROPC

To log into an OpenShift cluster on the command line, oc login supports user/password authentication for various identity providers like LDAP or OIDC with ROPC grant flow.

If no provider with password support is configured, the user is referred to the OAuth login page to obtain a temporary token interactively. After selecting the right provider, the user is forwarded to authenticate with the SSO provider, and redirected back to the cluster afterwards. Another click reveals the temporary token that can now be used for the CLI tools.

For an OpenID provider that supports Kerberos tickets, the authentication with the SSO provider happens transparently. For such setups, logging into a cluster via the CLI roughly requires the following steps:

• run oc login and click on the link, or visit a bookmark for the cluster login page
• click on the button for the OpenID provider
• watch the webpages forwarding to each other
• click on the link to reveal the temporary token
• use the shown temporary token/oc login command to log into the cluster

These steps must be performed daily and per cluster.

Approach: automate all the steps above

The Python script in this repository automates all the steps to obtain the temporary token so that the following is possible:

kinit $user@$domain

# either save the token directly in the specified context in ~/.kube/config...
ocp-sso-token $server --context $context --namespace $namespace
oc --context $context get pod

# ...or use the token with oc login
oc login --server $server --token $(ocp-sso-token $server)
oc get pod

Installing the script

# from Copr
sudo dnf copr enable @cki/ocp-sso-token
sudo dnf install ocp-sso-token

# from PyPI
pip install ocp-sso-token

# from source
pip install --user git+https://gitlab.com/cki-project/ocp-sso-token

Using the script to log into an OpenShift cluster via OIDC

usage: ocp-sso-token [-h] [--identity-providers IDENTITY_PROVIDERS]
                     [--context CONTEXT] [--namespace NAMESPACE]
                     api_url

Obtain an OCP OAuth token for a Kerberos ticket

positional arguments:
  api_url               Cluster API URL like https://api.cluster:6443

optional arguments:
  -h, --help            show this help message and exit
  --identity-providers IDENTITY_PROVIDERS
                        Identity provider names (default: SSO,OpenID)
  --context CONTEXT     Instead of printing the token, store it in the given context (default:
                        None)
  --namespace NAMESPACE
                        Namespace to use for --context (default: None)

If your identity provider name is not included in the defaults shown above, add it via --identity-providers. The first matching identity provider will be used.

With --context, the token is directly added to the Kubernetes configuration file (~/.kube/config or KUBECONFIG). Otherwise, the token is printed to the console.

Running a smoke test

kinit [e-mail]
server=https://api.cluster:6443; oc --server $server --token $(ocp-sso-token $server) get project

Logging into clusters with custom context names

With --context and --namespace, the obtained tokens are directly configured in the specified contexts:

$ kinit [e-mail]
$ ocp-sso-token https://api.cluster1:6443 --context cluster1 --namespace project1
$ ocp-sso-token https://api.cluster2:6443 --context cluster2 --namespace project2

$ oc config get-contexts
CURRENT   NAME       CLUSTER             AUTHINFO            NAMESPACE
          cluster1   api-cluster1:6443   api-cluster1:6443   project1
          cluster2   api-cluster2:6443   api-cluster2:6443   project2

$ oc --context cluster1 get pod
$ oc --context cluster2 get pod

Logging into clusters via oc login

Without --context, the obtained tokens are printed to the console and can be used with oc login. Logging into clusters this way creates context names automatically. This is most useful with a single cluster, but works with multiple clusters as well:

$ kinit [e-mail]
$ server=https://api.cluster1:6443; oc login --server $server --token $(ocp-sso-token $server)
$ server=https://api.cluster2:6443; oc login --server $server --token $(ocp-sso-token $server)

$ oc config get-contexts
CURRENT   NAME                                        CLUSTER             AUTHINFO                            NAMESPACE
          default/api-cluster1:6443/[e-mail]   api-cluster1:6443   [e-mail]/api-cluster1:6443   default
*         default/api-cluster2:6443/[e-mail]   api-cluster2:6443   [e-mail]/api-cluster2:6443   default

$ oc --namespace project2 get pod
$ oc --context default/api-cluster1:6443/[e-mail] --namespace project1 get pod
$ oc --context default/api-cluster2:6443/[e-mail] --namespace project2 get pod

ssl.SSLCertVerificationError

If you get an error similar to the following, most likely the intranet certificates are not used by the requests module as the certifi module ships its own certificate store:

ssl.SSLCertVerificationError: [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:992)

On Fedora/Debian-based distributions, this can be solved by using the certifi module from the distribution package repositories, e.g. via

pip uninstall certifi              # if certifi was already installed via pip
dnf install python3-certifi        # Fedora-based distributions
apt install python3-certifi        # Debian-based distributions

It is also possible to force the use of a specific certificate bundle via something like

export REQUESTS_CA_BUNDLE=/etc/pki/tls/certs/ca-bundle.crt

Creating a development setup and running the tests

Installing development dependencies:

pip install -e .[dev]

Running linting/tests:

tox

Creating a release

1. Create a release MR with an update of the version number in ocp_sso_token/__init__.py, e.g to '3.1.4'

2. Create an annotated tag with the same version prefixed with v and enter the release notes as the tag message, e.g.

   git tag v3.1.4 -a
   

   For the release notes, list the important changes and include the merge requests that introduced them, e.g.

   - Run tests in parallel on multiple Python versions (!8)
   

3. Push the tag to GitLab, e.g.

   git push origin v3.1.4
   

4. Wait for the tag pipeline to finish

5. Check the resulting GitLab and PyPI releases