k8s-workload-registrar

README

SPIRE Kubernetes Workload Registrar

The SPIRE Kubernetes Workload Registrar implements a Kubernetes ValidatingAdmissionWebhook that facilitates automatic workload registration within Kubernetes.

Configuration

Command Line Configuration

The registrar has the following command line flags:

Flag	Description	Default
-config	Path on disk to the HCL Configuration file	k8s-workload-registrar.conf

HCL Configuration

The configuration file is a required by the registrar. It contains HCL encoded configurables.

Key	Type	Required?	Description	Default
log_level	string	required	Log level (one of "panic","fatal","error","warn", "warning","info","debug","trace")	"info"
log_path	string	optional	Path on disk to write the log
trust_domain	string	required	Trust domain of the SPIRE server
agent_socket_path	string	optional	Path to the Unix domain socket of the SPIRE agent. Required if server_address is not a unix domain socket address.
server_address	string	required	Address of the spire server. A local socket can be specified using unix:///path/to/socket. This is not the same as the agent socket.
server_socket_path	string	optional	Path to the Unix domain socket of the SPIRE server, equivalent to specifying a server_address with a "unix://..." prefix
cluster	string	required	Logical cluster to register nodes/workloads under. Must match the SPIRE SERVER PSAT node attestor configuration.
pod_label	string	optional	The pod label used for Label Based Workload Registration
pod_annotation	string	optional	The pod annotation used for Annotation Based Workload Registration
mode	string	optional	How to run the registrar, either using a "webhook", "reconcile" or "crd". See Differences for more details.	"webhook"
disabled_namespaces	[]string	optional	Comma seperated list of namespaces to disable auto SVID generation for	"kube-system", "kube-public"

The following configuration directives are specific to "webhook" mode:

Key	Type	Required?	Description	Default
addr	string	required	Address to bind the HTTPS listener to	":8443"
cert_path	string	required	Path on disk to the PEM-encoded server TLS certificate	"cert.pem"
key_path	string	required	Path on disk to the PEM-encoded server TLS key	"key.pem"
cacert_path	string	required	Path on disk to the CA certificate used to verify the client (i.e. API server)	"cacert.pem"
insecure_skip_client_verification	boolean	required	If true, skips client certificate verification (in which case cacert_path is ignored). See Security Considerations for more details.	false

The following configuration directives are specific to "crd" mode:

Key	Type	Required?	Description	Default
add_svc_dns_name	bool	optional	Enable adding service names as SAN DNS names to endpoint pods	true
leader_election	bool	optional	Enable leader election for controller manager. Enabling this will ensure there is only one active controller manager.	false
metrics_bind_addr	string	optional	The address the metric endpoint binds to. The special value of "0" disables metrics.	":8080"
pod_controller	bool	optional	Enable auto generation of SVIDs for new pods that are created	true
webhook_enabled	bool	optional	Enable a validating webhook to ensure CRDs are properly fomatted and there are no duplicates. Only needed if manually creating entries	false
webhook_cert_dir	string	optional	Directory for certificates when enabling validating webhook. The certificate and key must be named tls.crt and tls.key.	"/run/spire/serving-certs"
webhook_port	int	optional	The port to use for the validating webhook.	9443

The following configuration directives are specific to "reconcile" mode:

Key	Type	Required?	Description	Default
leader_election	bool	optional	Enable/disable leader election. Enable if you have multiple registrar replicas running.	false
metrics_addr	string	optional	Address to expose metrics on, use 0 to disable.	0
controller_name	string	optional	Forms part of the spiffe IDs used for parent IDs	"spire-k8s-registrar"
add_pod_dns_names	bool	optional	Enable/disable adding k8s DNS names to pod SVIDs.	false
cluster_dns_zone	string	optional	The DNS zone used for services in the k8s cluster.	"cluster.local"

Example

log_level = "debug"
trust_domain = "domain.test"
server_socket_path = "/run/spire/sockets/registration.sock"
cluster = "production"

Workload Registration

When running in webhook, reconcile, or crd mode with pod_controller=true entries will be automatically created for Pods. There are three workload registration modes. If you use Service Account Based, don't specify either pod_label or pod_annotation. If you use Label Based, specify only pod_label. If you use Annotation Based, specify only pod_annotation.

It may take several seconds for newly created SVIDs to become available to workloads.

Service Account Based Workload Registration

The SPIFFE ID granted to the workload is derived from the 1) service account or 2) a configurable pod label or 3) a configurable pod annotation.

Service account derived workload registration maps the service account into a SPIFFE ID of the form spiffe://<TRUSTDOMAIN>/ns/<NAMESPACE>/sa/<SERVICEACCOUNT>. For example, if a pod came in with the service account blog in the production namespace, the following registration entry would be created:

Entry ID      : 200d8b19-8334-443d-9494-f65d0ad64eb5
SPIFFE ID     : spiffe://example.org/ns/production/sa/blog
Parent ID     : ...
TTL           : default
Selector      : k8s:ns:production
Selector      : k8s:pod-name:example-workload-98b6b79fd-jnv5m

Label Based Workload Registration

Label based workload registration maps a pod label value into a SPIFFE ID of the form spiffe://<TRUSTDOMAIN>/<LABELVALUE>. For example if the registrar was configured with the spire-workload label and a pod came in with spire-workload=example-workload, the following registration entry would be created:

Entry ID      : 200d8b19-8334-443d-9494-f65d0ad64eb5
SPIFFE ID     : spiffe://example.org/example-workload
Parent ID     : ...
TTL           : default
Selector      : k8s:ns:production
Selector      : k8s:pod-name:example-workload-98b6b79fd-jnv5m

Pods that don't contain the pod label are ignored.

Annotation Based Workload Registration

Annotation based workload registration maps a pod annotation value into a SPIFFE ID of the form spiffe://<TRUSTDOMAIN>/<ANNOTATIONVALUE>. By using this mode, it is possible to freely set the SPIFFE ID path. For example if the registrar was configured with the spiffe.io/spiffe-id annotation and a pod came in with spiffe.io/spiffe-id: production/example-workload, the following registration entry would be created:

Entry ID      : 200d8b19-8334-443d-9494-f65d0ad64eb5
SPIFFE ID     : spiffe://example.org/production/example-workload
Parent ID     : ...
TTL           : default
Selector      : k8s:ns:production
Selector      : k8s:pod-name:example-workload-98b6b79fd-jnv5m

Pods that don't contain the pod annotation are ignored.

Deployment

The registrar can either be deployed as standalone deployment, or as a container in the SPIRE server pod. If it is deployed standalone then it will require manual creation of an admin registration entry which will match the registrar deployment.

If it is deployed as a container within the SPIRE server pod then it talks to SPIRE server via a Unix domain socket. It will need access to a shared volume containing the socket file.

Reconcile Mode Configuration

To use reconcile mode you need to create appropriate roles and bind them to the ServiceAccount you intend to run the controller as. An example can be found in mode-reconcile/config/role.yaml, which you would apply with kubectl apply -f mode-reconcile/config/role.yaml

CRD Mode Configuration

The following configuration is required before "crd" mode can be used:

1. The SpiffeId CRD needs to be applied: kubectl apply -f mode-crd/config/spiffeid.spiffe.io_spiffeids.yaml
   • The SpiffeId CRD is namespace scoped
2. The appropriate ClusterRole need to be applied. kubectl apply -f mode-crd/config/crd_role.yaml
   • This creates a new ClusterRole named spiffe-crd-role
3. The new ClusterRole needs a ClusterRoleBinding to the SPIRE Server ServiceAccount. Change the name of the ServiceAccount and then: kubectl apply -f mode-crd/config/crd_role_binding.yaml
   • This creates a new ClusterRoleBinding named spiffe-crd-rolebinding
4. If you would like to manually create SpiffeId custom resources, then a validating webhook is needed to prevent misconfigurations and improve security: kubectl apply -f mode-crd/config/webhook.yaml
   • This creates a new ValidatingWebhookConfiguration and Service, both named k8s-workload-registrar
   • Make sure to add your CA Bundle to the ValidatingWebhookConfiguration where it says <INSERT BASE64 CA BUNDLE HERE>
   • Additionally a Secret that volume mounts the certificate and key to use for the webhook. See webhook_cert_dir configuration option above.

CRD mode Security Considerations

It is imperative to only grant trusted users access to manually create SpiffeId custom resources. Users with access have the ability to issue any SpiffeId to any pod in the namespace.

If allowing users to manually create SpiffeId custom resources it is important to use the Validating Webhook. The Validating Webhook ensures that registration entries created have a namespace selector that matches the namespace the resource was created in. This ensures that the manually created entries can only be consumed by workloads within that namespace.

Webhook Mode Configuration

The registrar will need access to its server keypair and the CA certificate it uses to verify clients.

The following K8S objects are required to set up the validating admission controller:

• Service pointing to the registrar port within the spire-server container
• ValidatingWebhookConfiguration configuring the registrar as a validating admission controller.

Additionally, unless you disable client authentication (insecure_skip_client_verification), you will need:

• Config with a user entry for the registrar service client containing the client certificate/key the API server should use to authenticate with the registrar.
• AdmissionConfiguration describing where the API server can locate the file containing the Config. This file is passed to the API server via the --admission-control-config-file flag.

For convenience, a command line utility is provided to generate authentication material and relevant Kubernetes configuration YAML.

$ go run generate-config.go
.... YAML configuration dump ....

Webhook mode Security Considerations

The registrar authenticates clients by default. This is a very important aspect of the overall security of the registrar since the registrar can be used to provide indirect access to the SPIRE server registration API, albeit scoped. It is NOT recommended to skip client verification (via the insecure_skip_client_verification configurable) unless you fully understand the risks.

Migrating away from the webhook

The k8s ValidatingWebhookConfiguration will need to be removed or pods may fail admission. If you used the default configuration this can be done with:

kubectl validatingwebhookconfiguration delete k8s-workload-registrar-webhook

DNS names

Both "reconcile" and "crd" mode provide the ability to add DNS names to registration entries for pods. They currently have different ideas about what names should be added, with "reconcile" adding every possible name that can be used to access a pod (via a service or directly), and "crd" mode limiting itself to <service>.<namespace>.svc. This functionality defaults off for "reconcile" mode and on for "crd" mode.

Warning: Some software is known to "validate" DNS and IP SANs provided in client certificates by using reverse DNS. There is no guarantee that a client in Kubernetes will be seen to connect from an IP address with valid reverse DNS matching one of the names generated by either of these DNS name implementation, in which case such validation will fail. If you are intending to use X509-SVIDs to authenticate clients to such services you will need to disable adding dns names to entries. This is known to affect etcd.

Differences between modes

The "webhook" mode uses a Validating Admission Webhook to capture pod creation/deletion events at admission time. It was the first of the registrar implementations, but suffers from the following problems:

• Race conditions between add and delete for StatefulSets will regularly lead to StatefulSets without entries;
• Unavailability of the webhook either has to block admission entirely, or you'll end up with pods with no entries;
• Spire server errors have to block admission entirely, or you'll end up with pods with no entries;
• It will not clean up left behind entries for pods deleted while the webhook/spire-server was unavailable;
• Entries are not parented to individual Nodes, all SVIDs are flooded to all agents in a cluster, which severely limits scalability. Use of the "webhook" mode is thus strongly discouraged, but it remains the default for backward compatibility reasons.

The "reconcile" mode and "crd" mode both make use of reconciling controllers instead of webhooks. "reconcile" mode, and "crd" mode with the pod_controller enabled, have similar automated workload creation functionality to webhook, but they do not suffer from the same race conditions, are capable of recovering from (and cleaning up after) failure of the registrar, and both also ensure that automatically created entries for Pods are limited to the appropriate Nodes to prevent SVID flooding. When used in this way, "reconcile" may be slightly faster to create new entries than "crd" mode, and requires less configuration.

"crd" mode additionally provides a namespaced SpiffeID custom resource. These are used internally by the registrar, but may also be manually created to allow creation of arbitrary Spire Entries. If you intend to manage SpiffeID custom resources directly then it is strongly encouraged to run the controller with the "crd" mode's webhook enabled.

SPIFFE ID Custom Resources

A sample SPIFFE ID custom resource for "crd" mode is below:

apiVersion: spiffeid.spiffe.io/v1beta1
kind: SpiffeID
metadata:
  name: my-spiffe-id
  namespace: my-namespace
spec:
  dnsNames:
  - my-dns-name
  selector:
    namespace: my-namespace
    podName: my-pod-name
  spiffeId: spiffe://example.org/my-spiffe-id
  parentId: spiffe://example.org/spire/server

The supported selectors are:

• arbitrary -- Arbitrary selectors
• containerName -- Name of the container
• containerImage -- Container image used
• namespace -- Namespace to match for this SPIFFE ID
• nodeName -- Node name to match for this SPIFFE ID
• podLabel -- Pod label name/value to match for this SPIFFE ID
• podName -- Pod name to match for this SPIFFE ID
• podUID -- Pod UID to match for this SPIFFE ID
• serviceAccount -- ServiceAccount to match for this SPIFFE ID

Note: Specifying DNS Names is optional.

Spire enforces that spiffeId+parentId+selectors are unique. The optional "crd" mode webhook