One of the challenges in adopting Kubernetes and microservices is the development and testing workflow. Creating and maintaining a full development environment with many microservices and their dependencies is complex and hard.
Service Preview addresses this challenge by connecting your CI system or local development infrastructure to the Kubernetes cluster, and dynamically routing specific requests to your local environment.
When Service Preview is used, incoming requests get routed by Ambassador to a Traffic Agent, which then routes traffic to the microservice. When a request meets a specific criteria (e.g., it has a specific HTTP header value), the Traffic Agent will route that request to the microservice running locally. The following video shows Service Preview in more detail:
There are three main components to Service Preview:
The Traffic Agent, which controls routing to the microservice. The Traffic Agent is deployed as a sidecar on the same pod as your microservice (behind the scenes, it's a special configuration of the basic Ambassador Edge Stack image).
The Traffic Manager, which manages the different instances of the Traffic Agent, and is deployed in the cluster.
The Edge Control local client, which runs in your local environment (Linux or Mac OS X). The client is the command line interface to the Traffic Manager.
To get started with Service Preview, you'll need to install Traffic Manager, configure a Traffic Agent, and then download and install the
The Traffic Manager is the central point of communication between Traffic Agents in the cluster and Edge Control Daemons on developer workstations.
- Ensure that you have a valid Ambassador Edge Stack license key installed in your cluster.
- Save the manifest below into a file called
- Apply the manifest to your cluster with
kubectl apply -f traffic-manager.yaml.
# This is traffic-manager.yaml---apiVersion: v1kind: Servicemetadata:name: telepresence-proxynamespace: ambassadorspec:type: ClusterIPclusterIP: Noneselector:app: telepresence-proxyports:- name: sshdprotocol: TCPport: 8022- name: apiprotocol: TCPport: 8081---apiVersion: apps/v1kind: Deploymentmetadata:name: telepresence-proxynamespace: ambassadorlabels:app: telepresence-proxyspec:replicas: 1selector:matchLabels:app: telepresence-proxytemplate:metadata:labels:app: telepresence-proxyspec:containers:- name: telepresence-proxyimage: quay.io/datawire/aes:1.4.3command: [ "traffic-manager" ]ports:- name: sshdcontainerPort: 8022env:- name: AMBASSADOR_LICENSE_FILEvalue: /.config/ambassador/license-keyvolumeMounts:- mountPath: /tmp/ambassador-pod-infoname: ambassador-pod-info- mountPath: /.config/ambassadorname: ambassador-edge-stack-secretsreadOnly: truerestartPolicy: AlwaysterminationGracePeriodSeconds: 0volumes:- downwardAPI:items:- fieldRef:fieldPath: metadata.labelspath: labelsname: ambassador-pod-info- name: ambassador-edge-stack-secretssecret:secretName: ambassador-edge-stack
Any microservice running in a cluster with a traffic manager can opt in to intercept functionality by including the Traffic Agent in its pods.
- Since the Traffic Agent is built on Ambassador Edge Stack, it needs the same RBAC permissions that Ambassador does. The easiest way to provide this is to create a
ServiceAccountin your service's namespace, bound to the
---apiVersion: v1kind: ServiceAccountmetadata:name: traffic-agentnamespace: defaultlabels:product: aes---apiVersion: rbac.authorization.k8s.io/v1beta1kind: ClusterRoleBindingmetadata:name: traffic-agentlabels:product: aesroleRef:apiGroup: rbac.authorization.k8s.iokind: ClusterRolename: ambassadorsubjects:- kind: ServiceAccountname: traffic-agentnamespace: default
(If you want to include the Traffic Agent with multiple services, they can all use the same
Copy the above YAML into
traffic-agent-rbac.yaml and, if necessary, edit the two
namespaces appropriately. Apply it:
$ kubectl apply -f traffic-agent-rbac.yamlserviceaccount/traffic-agent createdclusterrolebinding.rbac.authorization.k8s.io/traffic-agent created
- Next, you'll need to modify the YAML for each microservice to include the Traffic Agent. We'll start with a set of manifests for a simple microservice:
# This is hello.yaml---apiVersion: v1kind: Servicemetadata:name: hellolabels:app: hellospec:selector:app: helloports:- protocol: TCPport: 80targetPort: 8000 # Application port---apiVersion: apps/v1kind: Deploymentmetadata:name: hellolabels:app: hellospec:replicas: 1selector:matchLabels:app: hellotemplate:metadata:labels:app: hellospec:containers: # Application container- name: helloimage: datawire/hello-world:latestports:- containerPort: 8000 # Application port
In order to run the sidecar:
- you need to include the Traffic Agent container in the microservice pod;
- you need to switch the microservice's
Servicedefinition to point to the Traffic Agent's listening port (either 8080 or 8443); and
- you need to tell the Traffic Agent how to set up for the microservice, using environment variables.
Here is a modified set of manifests that includes the Traffic Agent (with notes below):
# This is hello-intercept.yaml---apiVersion: v1kind: Servicemetadata:name: hellolabels:app: hellospec:selector:app: helloports:- protocol: TCPport: 80targetPort: 9900 # Traffic Agent listen port (note 1)---apiVersion: apps/v1kind: Deploymentmetadata:name: hellolabels:app: hellospec:replicas: 1selector:matchLabels:app: hellotemplate:metadata:labels:app: hellospec:containers:- name: hello # Application container (note 2)image: datawire/hello-world:latestports:- containerPort: 8000 # Application port- name: traffic-agent # Traffic Agent container (note 3)image: quay.io/datawire/aes:1.4.3 # (note 4)ports:- containerPort: 9900 # Traffic Agent listen portenv:- name: AGENT_SERVICE # Name to use for intercepting (note 5)value: hello- name: AGENT_PORT # Port on which to talk to the microservice (note 6)value: "8000"- name: AGENT_MANAGER_NAMESPACE # Namespace for contacting the Traffic Manager (note 7)value: ambassador- name: AMBASSADOR_NAMESPACE # Namespace in which this microservice is running (note 8)valueFrom:fieldRef:fieldPath: metadata.namespace- name: AGENT_LISTEN_PORT # Port on which to listen for connections (note 9)value: "9900"serviceAccountName: traffic-agent
Key points include:
- Note 1: The
Servicenow points to the Traffic Agent’s listen port (9900) instead of the application’s port (8000).
- Note 2: The microservice's application container is actually unchanged.
- Note 3: The Traffic Agent's container has been added.
- Note 4: The Traffic Agent is included in the AES image.
- Note 5: The
AGENT_SERVICEenvironment variable is mandatory. It sets the name that the Traffic Agent will report to the Traffic Manager for this microservice: you will have to provide this name to intercept this microservice.
- Note 6: The
AGENT_PORTenvironment variable is mandatory. It tells the Traffic Agent the local port on which the microservice is listening.
- Note 7: The
AGENT_MANAGER_NAMESPACEenvironment variable tells the Traffic Agent the namespace in which it will be able to find the Traffic Manager. If not present, it defaults to the
- Note 8: The
AMBASSADOR_NAMESPACEenvironment variable is mandatory. It lets the Traffic Agent tell the Traffic Manager the namespace in which the microservice is running.
- Note 9: The
AGENT_LISTEN_PORTenvironment variable tells the Traffic Agent the port on which to listen for incoming connections. The
Servicemust point to this port (see Note 1). If not present, it defaults to port 9900.
If other microservices in the cluster expect to speak TLS to this microservice, tell the Traffic Agent to terminate TLS:
- Set the
AGENT_TLS_TERM_SECRETenvironment variable to the name of a Kubernetes Secret that contains a TLS certificate
- The Traffic Agent will terminate TLS on its listen port (9900 by default) using the named certificate
- The Traffic Agent will not accept cleartext communication when configured to terminate TLS
If this microservice expects incoming requests to speak TLS, tell the Traffic Agent to originate TLS:
- Set the
AGENT_TLS_ORIG_SECRETenvironment variable to the name of a Kubernetes Secret that contains a TLS certificate
- The Traffic Agent will use that certificate originate HTTPS requests to the application