Air-gapped Kubernetes Cluster
You set up an air gapped kubernetes cluster by installing nodes on an isolated network. You must host all container images and control plane binaries on local servers. Many teams build these private networks to protect sensitive data from outside threats. Once inside the private network, admins often grant broad cluster roles to speed up work. Broad permissions let one bad command break live apps or expose private data. How an air gapped kubernetes cluster works An air gapped kubernetes clust
Your coding agents run with your access.
Claude Code reaches prod with a developer's credentials, and approval prompts get clicked through. hoop.dev checks every query, masks sensitive data, and holds risky changes for a human.
You set up an air gapped kubernetes cluster by installing nodes on an isolated network. You must host all container images and control plane binaries on local servers. Many teams build these private networks to protect sensitive data from outside threats. Once inside the private network, admins often grant broad cluster roles to speed up work. Broad permissions let one bad command break live apps or expose private data.
How an air gapped kubernetes cluster works
An air gapped kubernetes cluster runs with zero outbound paths to the public internet. Engineers must load operating system packages, container runtimes, and cluster binaries onto local storage first. A private image registry inside the closed network holds every container image for the nodes. When a pod starts, the node pulls the image from this local registry. Network firewalls block any traffic that tries to leave or enter the local zone.
Inside the isolated network, the API server still controls who can run commands. Kubernetes uses Role-Based Access Control to check every user and service account request. A Role sets permissions inside one namespace, while a ClusterRole covers the whole cluster. A RoleBinding grants those rules in one namespace, and a ClusterRoleBinding grants them everywhere. All permissions are additive, which means the system has no deny rules at all.
Admins also use subresources in these rules to control pod logs and shell commands. When you write a rule, you name the subresource with a slash, like pods/log. You can also set rules for pods/exec to control who opens a container shell. This native setup gives you a strong foundation for local user permissions.
What native network isolation and roles leave open
Network isolation stops outside attackers from reaching your nodes over the public web. Native roles and bindings also let you decide which users can reach each namespace. Together, these controls handle network boundaries and basic login checks very well. They make sure only approved engineers on the local network can talk to the API.
Kubernetes RBAC: architecture patterns and best practices
Continue reading? Get the full guide.
Free. No spam. Unsubscribe anytime.
Still, network walls and static roles only check who logs in to the cluster. They do not inspect what individual commands or requests do once a session starts. If a role allows pods/exec, the user can run any binary inside that container. Because permissions are additive, you cannot add a rule to block one dangerous command. Native RBAC also cannot require a peer review before a single shell command runs.
Data exposure is another major gap inside an isolated cluster. When an engineer reads pod logs or queries a pod, raw data returns in full. Native roles do not rewrite sensitive fields like emails, phone numbers, or card numbers. If a user has read rules, the cluster sends back every secret value in plain text. Teams need a way to check commands in flight and mask sensitive output fields.
Add request guardrails and response masking beside Kubernetes
Cluster roles and service accounts govern who reaches the Kubernetes API. Still, a broad role or agent can delete workloads or print sensitive values from ConfigMaps and logs. In practice, hoop.dev runs beside the API server as a sidecar for your Kubernetes API and checks every request and response alongside RBAC.
With hoop.dev Kubernetes controls, guardrails can allow listing Secret names while refusing reads of Secret data before execution. Data masking redacts matched keys in returned JSON responses and log streams in memory. Furthermore, hoop.dev holds SOC 2 Type II and is MIT licensed on GitHub.
Frequently asked questions
How do nodes pull images in an air gapped kubernetes cluster?
Nodes pull images from a private registry hosted inside the isolated network. Engineers must copy all required container images into this local registry before deploying pods.
Can native Kubernetes RBAC block specific shell commands inside a pod?
No, native RBAC only checks if a user has permission for the pods/exec subresource. Once the shell opens, RBAC cannot read or block the commands the user types.
How do Role and ClusterRole rules differ in an isolated cluster?
A Role applies permission rules to resources inside one single namespace. A ClusterRole applies rules across all namespaces or to cluster-wide resources like nodes.
Related guides
In this post