Skip to content

Latest commit

 

History

History
162 lines (106 loc) · 6.06 KB

README.md

File metadata and controls

162 lines (106 loc) · 6.06 KB

SPIRE Controller Manager

Build Status Pre-Production Phase

A Kubernetes Controller manager which facilitates the registration of workloads and establishment of federation relationships.

How it Works

Custom Resources

ClusterSPIFFEID

The ClusterSPIFFEID resource is a cluster scoped CRD that describes the shape of the identity that is applied to workloads, as well as selectors that describe which workloads the identity applies to.

ClusterFederatedTrustDomain

The ClusterFederatedTrustDomain resource is a cluster scoped CRD that describes a federation relationship for the cluster.

ClusterStaticEntry

The ClusterStaticEntry resource is a cluster scoped CRD that describes a static SPIRE registration entry. It is typically used for registering workloads that do not run in the Kubernetes cluster but otherwise need to be part of the trust domain (e.g. downstream nested SPIRE servers).

Reconciliation

Workload Registration

To facilitate workload registration, the SPIRE Controller manager registers controllers against the following resources:

When changes are detected on these resources, a workload reconciliation process is triggered. This process determines which SPIRE entries should exist based on the existing Pods and ClusterSPIFFEID resources which apply to those pods, as well as static entries declared via ClusterStaticEntry resources. The reconciliation process creates, updates, and deletes entries on SPIRE server as appropriate to match the declared state.

Federation

To facilitate federation, the SPIRE Controller manager registers controllers against the following resources:

When changes are detected on these resources, a federation relationship reconciliation process is triggered. This process determines which SPIRE federation relationships should exist based on the existing ClusterFederatedTrustDomain resources. It creates, updates, and deletes federation relationships as appropriate to match the declared state.

Deployment

The SPIRE Controller Manager is designed to be deployed in the same pod as the SPIRE Server. It communicates with the SPIRE Server API using a private Unix Domain Socket within a shared volume. It requires configuration for the environment where it is being deployed.

The demo includes sample configuration for deploying the SPIRE Controller Manager, SPIRE, and the SPIFFE CSI driver, including requisite RBAC and Webhook configuration.

Upgrading

The SPIRE Controller Manager must have the correct set of Custom Resources and the manager-role that corresponds to the version to be installed.

Before upgrading, please install custom resources from config/crd and verify that manager-role is up-to-date.

Compatibility

The SPIRE APIs used by the SPIRE Controller Manager are generally stable and supported since at least SPIRE v1.0. However, the API has gained support for additional entry fields beyond what was supported in SPIRE v1.0. Notably, these include the jwt_svid_ttl, hint and the store_svid fields. The ClusterStaticEntry CRD allows these fields to be set, however, a SPIRE server that does not support these fields will not retain them. This means if these fields are set on a ClusterStaticEntry with an older version of SPIRE, the SPIRE Controller Manager will continously try to reconcile SPIRE server. In order to use these fields, you must be on a version of SPIRE Server which supports them.

At the moment, SPIRE Controller Manager will silently try and reconcile these fields over and over. Future updates may cause the SPIRE Controller Manager to fail when an unsupporting SPIRE Server is encounted while these fields are set.

The hint field is supported as of SPIRE 1.6.3.

The jwt_svid_ttl field is supported as of SPIRE 1.5.0.

The store_svid field is supported as of SPIRE 1.1.0.

Demo

Link

Troubleshooting

Workloads

Workload Not Registered

ClusterSPIFFEID Not Defined

Define a ClusterSPIFFEID that applies to the workload pod.

Workload Pod Excluded by ClusterSPIFFEID PodSelector or NamespaceSelector

Adjust the ClusterSPIFFEID selectors.

Failed to Render Templates Against Workload Pod or Node

Check the ClusterSPIFFEID status for entry render failures. Check logs to determine why the rendering failed.

Failed to Register with SPIRE Server

Check logs for API failures talking to SPIRE Server.

Federation

Federation Relationship Missing

ClusterFederatedTrustDomain Not Defined

Define a ClusterFederatedTrustDomain for the target trust domain.

ClusterFederatedTrustDomain TrustDomain Conflict

Ensure each ClusterFederatedTrustDomain resource has a unique trust domain. The controller will only ignore all but the oldest ClusterFederatedTrustDomain resource with a conflicting trust domain.

Workload Not Federated With Trust Domain

Check the ClusterSPIFFEID for the workload. The federatesWith field must include the federated trust domain.

Security

Reporting a Vulnerability

Vulnerabilities can be reported by sending an email to security@spiffe.io. A confirmation email will be sent to acknowledge the report within 72 hours. A second acknowledgement will be sent within 7 days when the vulnerability has been positively or negatively confirmed.