flux2/rfcs/0004-multi-tenancy

RFC-0004 Flux Multi-Tenancy

Status: provisional

Creation date: 2021-11-15

Last update: 2021-12-17

Summary

This RFC explains the mechanisms available in Flux for implementing multi-tenancy, defines two models for multi-tenancy, and gives reference implementations for those models.

Motivation

To this point, the Flux project has provided examples of multi-tenancy, but not explained exactly how they relate to Flux's authorisation model. This RFC explains two multi-tenancy implementations, their security properties, and how they are implemented within the authorisation model as defined in RFC-0001.

Goals

• Explain the mechanisms available in Flux for supporting multi-tenancy
• Define two models for multi-tenancy, "soft multi-tenancy" and "hard multi-tenancy".
• Explain when each model is appropriate.
• Describe a reference implementation of each model with Flux.

Non-Goals

• Give an exhaustive account of multi-tenancy implementations in general.
• Provide an end-to-end workflow of how to set up multi-tenancy with Flux.

Introduction

Flux allows different organizations and/or teams to share the same Kubernetes control plane; this is referred to as "multi-tenancy". To make this safe, Flux supports segmentation and isolation of resources by using namespaces and role-based access control ("RBAC"), and integrating with Kubernetes Cluster API.

The following subsections explain the existing mechanisms used for safe multi-tenancy.

Flux's authorisation model

Flux defers to Kubernetes' native RBAC to specify which operations are authorised when processing the custom resources in the Flux API. By default, this means operations are constrained by the service account under which the controllers run, which (again, by default) has the cluster-admin role bound to it. This is convenient for a deployment in which all users are trusted.

In a multi-tenant deployment, each tenant needs to be restricted in the operations that can be done on their behalf. Since tenants control Flux via its API objects, this becomes a matter of attaching RBAC rules to Flux API objects. There are two mechanisms that do this, "impersonation" and "remote apply".

Impersonation

The Kustomize controller and Helm controller both apply arbitrary sets of Kubernetes configuration to a cluster. These controllers are subject to authorisation on two counts:

• when accessing Kubernetes resources that are needed for a particular "apply" operation -- for example, a secret referenced in the field .spec.valuesFrom in a HelmRelease;
• when creating, watching, updating and deleting Kubernetes resources in the process of applying a piece of configuration.

To give users control over this authorisation, these two controllers will impersonate (assume the identity of) a service account mentioned in the apply specification (e.g., the field .spec.serviceAccountName in a Kustomization object or in a HelmRelease object) for both accessing resources and applying configuration. This lets a user constrain the operations mentioned above with RBAC.

As stated in RFC-0003, the platform admins can configure Flux to enforce service account impersonation by setting a default service account name when .spec.serviceAccountName is not specified.

Remote apply

The Kustomize controller and Helm controller are able to apply a set of configuration to a cluster other than the cluster in which they run. If the specification refers to a secret containing a "kubeconfig" file, the controller will construct a client using that kubeconfig, then the client used to apply the specified set of configuration. The effect of this is that the configuration will be applied as the user given in the kubeconfig; often this is a user with the cluster-admin role bound to it, but not necessarily so.

Assumptions made by the multi-tenancy models

User Roles

The tenancy models assume two types of user: platform admins and tenants. Besides installing Flux, all the other operations (deploy applications, configure ingress, policies, etc) do not require users to have direct access to the Kubernetes API. Flux acts as a proxy between users and the Kubernetes API, using Git as source of truth for the cluster desired state. Changes to the clusters and workloads configuration can be made in a collaborative manner, where the various teams responsible for the delivery process propose, review and approve changes via pull request workflows.

Platform Admins

The platform admins have unrestricted access to Kubernetes API. They are responsible for installing Flux and granting Flux access to the sources (Git, Helm, OCI repositories) that make up the cluster(s) control plane desired state. The repository(s) owned by the platform admins are reconciled on the cluster(s) by Flux, under the cluster-admin Kubernetes cluster role.

Example of operations performed by platform admins:

• Bootstrap Flux onto cluster(s).
• Extend the Kubernetes API with custom resource definitions and validation webhooks.
• Configure various controllers for ingress, storage, logging, monitoring, progressive delivery, etc.
• Set up namespaces for tenants and define their level of access with Kubernetes RBAC.
• Onboard tenants by registering their Git repositories with Flux.

Tenants

The tenants have restricted access to the cluster(s) according to the Kubernetes RBAC configured by the platform admins. The repositories owned by tenants are reconciled on the cluster(s) by Flux, under the Kubernetes account(s) assigned by platform admins.

Example of operations performed by tenants:

• Register their sources with Flux (GitRepositories, HelmRepositories and Buckets).
• Deploy workload(s) into their namespace(s) using Flux custom resources (Kustomizations and HelmReleases).
• Automate application updates using Flux custom resources (ImageRepositories, ImagePolicies and ImageUpdateAutomations).
• Configure the release pipeline(s) using Flagger custom resources (Canaries and MetricsTemplates).
• Setup webhooks and alerting for their release pipeline(s) using Flux custom resources (Receivers and Alerts).

Tenancy Models

The Kubernetes tenancy models supported by Flux are: soft multi-tenancy and hard multi-tenancy.

For an overview of the Kubernetes multi-tenant architecture please consult the following documentation:

• Three Tenancy Models For Kubernetes
• GKE multi-tenancy overview
• EKS multi-tenancy best practices

Soft Multi-Tenancy

With soft multi-tenancy, the platform admins use Kubernetes constructs such as namespaces, accounts, roles and role bindings to create a logical separation between tenants.

When Flux deploys workloads from a repository belonging to a tenant, it uses the Kubernetes account assigned to that tenant to perform the git-to-cluster reconciliation. By leveraging Kubernetes RBAC, Flux can ensure that the operations performed by tenants are restricted to their namespaces.

Note that with this model, tenants share cluster-wide resources such as ClusterRoles, CustomResourceDefinitions, IngressClasses, StorageClasses, and they cannot create or alter these resources. If a tenant adds a cluster-scoped resource definition to their repository, Flux will fail the git-to-cluster reconciliation due to Kubernetes RBAC restrictions.

To restrict the reconciliation of tenant's sources, a Kubernetes service account name can be specified in Flux Kustomizations and HelmReleases under .spec.serviceAccountName. Please consult the Flux
documentation for more details:

• Kustomization API: Role-based access control
• HelmRelease API: Role-based access control
• Flux multi-tenancy example repository

Note that with soft multi-tenancy, true tenant isolation requires security measures beyond Kubernetes RBAC. Please refer to the Kubernetes security considerations documentation for more details on how to harden shared clusters.

Tenants Onboarding

When onboarding tenants, platform admins have the option to assign namespaces, set permissions and register the tenants main repositories onto clusters.

The Flux CLI offers an easy way of generating all the Kubernetes manifests needed to onboard tenants:

• flux create tenant command generates namespaces, service accounts and Kubernetes RBAC with restricted access to the cluster resources, given tenants access only to their namespaces.
• flux create secret git command generates SSH keys used by Flux to clone the tenants repositories.
• flux create source git command generates the configuration that tells Flux which repositories belong to tenants.
• flux create kustomization command generates the configuration that tells Flux how to reconcile the manifests found in the tenants repositories.

Once the tenants main repositories are registered on the cluster(s), the tenants can configure their app delivery in Git using Kubernetes namespace-scoped resources such as Deployments, Services, Flagger Canaries, Flux GitRepositories, Kustomizations, HelmRepositories, HelmReleases, ImageUpdateAutomations, Alerts, Receivers, etc.

Hard Multi-Tenancy

With hard multi-tenancy, the platform admins create dedicated clusters for each tenant.

When the tenants's clusters are created with Kubernetes Cluster API, the Flux instance installed on the management cluster is responsible for reconciling the cluster definitions belonging to tenants.

To enable GitOps for the tenant's clusters, the platform admins can configure the Flux instance running on the management cluster to connect to the tenant's cluster using the kubeconfig generated by the Cluster API provider or by creating kubeconfig secrets for the clusters created by other means than Cluster API.

To configure Flux reconciliation of remote clusters, a Kubernetes secret containing a kubeConfig can be specified in Flux Kustomizations and HelmReleases under .spec.kubeConfig.secretRef. Please consult the Flux API documentation for more details:

• Kustomization API: Remote Clusters
• HelmRelease API: Remote Clusters

Note that with hard multi-tenancy, tenants have full access to cluster-wide resources, so they have the option to manage Flux independently of platform admins, by deploying a Flux instance on each cluster.

Caveats

When using a Kubernetes Cluster API provider, the kubeConfig secret is automatically generated and Flux can make use of it without any manual actions. For clusters created by other means than Cluster API, the platform team has to create the kubeConfig secrets to allow Flux access to the remote clusters.

As of Flux v0.24 (Nov 2021), we don't provide any guidance for cluster admins on how to generate the kubeConfig secrets.

Implementation History

• Soft multi-tenancy based on service account impersonation was first released in flux2 v0.0.1.
• Generating namespaces and RBAC for defining tenants with flux create tenant was first released in flux2 v0.1.0.
• Hard multi-tenancy based on remote cluster reconciliation was first released in flux2 v0.2.0.
• Soft multi-tenancy end-to-end workflow example was first published on 27 Nov 2020 at fluxcd/flux2-multi-tenancy.
• Soft multi-tenancy CVE-2021-41254 "Privilege escalation to cluster admin on multi-tenant environments" was fixed in flux2 v0.15.0.