Launching kubectl for cluster within Rancher UI fails in a cluster after following the CIS Benchmark Hardening Guide for Kubernetes

This document (000020198) is provided subject to the disclaimer at the end of this document.

Situation

Issue

Attempting to launch kubectl in the Rancher v2.x UI, for a cluster upon which the Rancher CIS Hardening Guide has been applied, results in a Closed Code: 1006 message. Further, using the browser developer tools to inspect requests when opening this page reveals the API request to initiate the connection (https:///v3/clusters/?shell=true) receiving a HTTP 403 response.

Pre-requisites

An RKE CLI or Rancher v2.x launched Kubernetes cluster, with the Rancher v2.1.x, v2.2.x or v2.3.x CIS Hardening Guide applied.

Root cause

This behaviour is caused by CIS Control 1.1.12, which specifies that the DenyEscalatingExec Admission Controller should be enabled on the Kubernetes API Server.

The terminal for the Rancher UI is provided by exec'ing into a cattle-node-agent Pod, whilst Pods within this DaemonSet run in Privileged mode. As a result the exec to open the terminal session is denied by the DenyEscalatingExec Admission Controller.

Workaround

You can workaround the issue by removing DenyEscalatingExec from the list of enable-admission-plugins in extra_args for the kube-api service.

Resolution

This issue is tracked in the Rancher GitHub issue #19439.

Disclaimer

This Support Knowledgebase provides a valuable tool for SUSE customers and parties interested in our products and solutions to acquire information, ideas and learn from one another. Materials are provided for informational, personal or non-commercial use within your organization and are presented "AS IS" WITHOUT WARRANTY OF ANY KIND.