This guide walks you through getting a SPIRE Server and SPIRE Agent running in a Kubernetes cluster, and configuring a workload container to access SPIRE.
In this introduction to SPIRE on Kubernetes you will learn how to:
- Create the appropriate Kubernetes namespaces and service accounts to deploy SPIRE
- Deploy the SPIRE Server as a Kubernetes statefulset
- Deploy the SPIRE Agent as a Kubernetes deaemonset
- Configure a registration entry for a workload
- Fetch an x509-SVID over the SPIFFE Workload API
- Learn where to find resources for more complex installations
This steps in this guide have been tested on these Kubernetes versions: 1.13.1, 1.12.4, and 1.10.12.
Obtain the Required Files
This deployment this guide walks you through setting up requires a number of .yaml files and you must run all commands in the directory in which those files reside.
To obtain this directory of files clone https://github.com/spiffe/spire-tutorials and obtain the .yaml files from the spire-tutorials/k8s subdirectory.
Configure Kubernetes Namespace for SPIRE Components
Follow these steps to configure the spire namespace in which SPIRE Server and SPIRE Agent are deployed.
Create the namespace:
$ kubectl apply -f spire-namespace.yaml
Run the following command and verify that spire is listed in the output:
$ kubectl get namespaces
Configure SPIRE Server
Create Server Bundle Configmap, Role & ClusterRoleBinding
For the server to function, it is necessary for it to provide agents with certificates that they can use to verify the identity of the server when establishing a connection.
In a deployment such as this, where the agent and server share the same cluster, SPIRE can be configured to automatically generate these certificates on a periodic basis and update a configmap with contents of the certificate. To do that, the server needs the ability to get and patch a configmap object in the
To allow the server to read and write to this configmap, a ClusterRole must be created that confers the appropriate entitlements to Kubernetes RBAC, and that ClusterRoleBinding must be associated with the service account created in the previous step.
Create the server’s service account, configmap and associated role bindings as follows:
$ kubectl apply \ -f server-account.yaml \ -f spire-bundle-configmap.yaml \ -f server-cluster-role.yaml
Create Server Configmap
The server is configured in the Kubernetes configmap specified in server-configmap.yaml, which specifies a number of important directories, notably /run/spire/data and /run/spire/config. These volumes are bound in when the server container is deployed.
Deploy the server configmap and statefulset by applying the following files via kubectl:
$ kubectl apply \ -f server-configmap.yaml \ -f server-statefulset.yaml \ -f server-service.yaml
This creates a statefulset called spire-server in the spire namespace and starts up a spire-server pod, as demonstrated in the output of the following two commands:
$ kubectl get statefulset --namespace spire NAME READY AGE spire-server 1/1 86m $ kubectl get pods --namespace spire NAME READY STATUS RESTARTS AGE spire-server-0 1/1 Running 0 86m $ kubectl get services --namespace spire NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE spire-server NodePort 10.107.205.29 <none> 8081:30337/TCP 88m
Configure and deploy the SPIRE Agent
To allow the agent read access to the kubelet API to perform workload attestation, a Service Account and ClusterRole must be created that confers the appropriate entitlements to Kubernetes RBAC, and that ClusterRoleBinding must be associated with the service account created in the previous step.
$ kubectl apply \ -f agent-account.yaml \ -f agent-cluster-role.yaml
Apply the agent-configmap.yaml configuration file to create the agent configmap and deploy the Agent as a daemonset that runs one instance of each Agent on each Kubernetes worker node.
$ kubectl apply \ -f agent-configmap.yaml \ -f agent-daemonset.yaml
This creates a daemonset called spire-agent in the spire namespace and starts up a spire-agent pod along side spire-server, as demonstrated in the output of the following two commands:
$ kubectl get daemonset --namespace spire NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE spire-agent 1 1 1 1 1 <none> 6m45s $ kubectl get pods --namespace spire NAME READY STATUS RESTARTS AGE spire-agent-88cpl 1/1 Running 0 6m45s spire-server-b95945658-4wbkd 1/1 Running 0 103m
In order to enable SPIRE to perform workload attestation – which allows the agent to identify the workload to attest to its agent – you must register the workload in the server. This tells SPIRE how to identify the workload and which SPIFFE ID to give it.
Create a new registration entry for the node, specifying the SPIFFE ID to allocate to the node:
$ kubectl exec -n spire spire-server-0 -- \ /opt/spire/bin/spire-server entry create \ -spiffeID spiffe://example.org/ns/spire/sa/spire-agent \ -selector k8s_sat:cluster:demo-cluster \ -selector k8s_sat:agent_ns:spire \ -selector k8s_sat:agent_sa:spire-agent \ -node
Create a new registration entry for the workload, specifying the SPIFFE ID to allocate to the workload:
$ kubectl exec -n spire spire-server-0 -- \ /opt/spire/bin/spire-server entry create \ -spiffeID spiffe://example.org/ns/default/sa/default \ -parentID spiffe://example.org/ns/spire/sa/spire-agent \ -selector k8s:ns:default \ -selector k8s:sa:default
Configure a Workload Container to Access SPIRE
In this step, you configure a workload container to access SPIRE. Specifically, you are configuring the workload container to access the Workload API UNIX domain socket.
The client-deployment.yaml file configures a no-op container using the spire-k8s docker image used for the server and agent. Examine the
volumes configuration stanzas to see how the UNIX domain
agent.sock is bound in.
You can test that the agent socket is accessible from an application container by issuing the following sequence of commands:
Apply the deployment file:
$ kubectl apply -f client-deployment.yaml
Obtain the pod hash for the workload container pod:
$ kubectl get pods NAME READY STATUS RESTARTS AGE client-6f9659bd44-m98vv 1/1 Running 0 18s
Obtain a shell connection to the running pod:
$ kubectl exec -it client-6f9659bd44-m98vv /bin/sh
Verify the container can access the socket:
/opt/spire/bin/spire-agent api fetch -socketPath /run/spire/sockets/agent.sock
If the agent is not running, you’ll see an error message such as “no such file or directory” or “connection refused”.
If the agent is running, you’ll see a list of SVIDs.
Tear Down All Components
Delete the workload container:
$ kubectl delete deployment client
Run the following commands to delete all deployments and configurations for the agent, server, and namespace:
$ kubectl delete namespace spire
Considerations When Using Minikube
If you are using minikube to run this tutorial, when starting your cluster you should pass some additional configuration flags.
minikube start \ --extra-config=apiserver.service-account-signing-key-file=/var/lib/minikube/certs/sa.key \ --extra-config=apiserver.service-account-key-file=/var/lib/minikube/certs/sa.pub \ --extra-config=apiserver.service-account-issuer=api \ --extra-config=apiserver.service-account-api-audiences=api,spire-server \ --extra-config=apiserver.authorization-mode=RBAC \ --extra-config=kubelet.authentication-token-webhook=true
Considerations For A Production Environment
When deploying SPIRE in a production environment the following considerations should be made.
In the Create Server Configmap step: set the the cluster name in the
k8s_sat NodeAttestor entry to the name you provide in the agent-configmap.yaml configuration file.
If your Kubernetes cluster supports projected service account tokens, consider using the built-in Projected Service Account Token k8s Node Attestor for authenticating the SPIRE agent to the server. Projected Service Account Tokens are more tightly scoped than regular service account tokens, and thus more secure.
As configured, the SPIRE agent does not verify the identity of the Kubernetes kubelet when requesting metadata for workload attestation. For additional security, you may wish to configure the Kubernetes workload attestor to perform this verification on compatible Kubernetes distributions by setting
false. Read more
- Review the SPIRE Documentation to learn how to configure SPIRE for your environment.