Funqy Knative Events Binding
Quarkus Funqy Knative Events builds off of the Funqy HTTP extension to allow you to route and process Knative Events within a Funqy function.
The guide walks through quickstart code to show you how you can deploy and invoke on Funqy functions with Knative Events.
Prerequisites
To complete this guide, you need:
-
Roughly 1 hour
-
An IDE
-
JDK 17+ installed with
JAVA_HOME
configured appropriately -
Apache Maven 3.9.9
-
Optionally the Quarkus CLI if you want to use it
-
Read about Funqy Basics. This is a short read!
-
Have gone through the Knative Tutorial, specifically Brokers and Triggers
Setting up Knative
Setting up Knative locally in a Minikube environment is beyond the scope of this guide. It is advised to follow this Knative Tutorial put together by Red Hat. It walks through how to set up Knative on Minikube or OpenShift in a local environment.
Specifically you should run the Brokers and Triggers tutorial as this guide requires that you can invoke on a Broker to trigger the quickstart code. |
Read about Cloud Events
The Cloud Event specification is a good read to give you an even greater understanding of Knative Events.
The Quickstart
Clone the Git repository: git clone https://github.com/quarkusio/quarkus-quickstarts.git
, or download an archive.
The solution is located in the funqy-knative-events-quickstart
directory.
The Quickstart Flow
The quickstart works by manually sending an HTTP request containing a Cloud Event to the Knative Broker using curl
.
The Knative Broker receives the request and triggers the startup of the Funqy container built by the quickstart.
The event triggers the invocation of a chain of Funqy functions. The output of one function triggers the
invocation of another Funqy function.
Funqy and Cloud Events
When living within a Knative Events environment, Funqy functions are triggered by a specific Cloud Event type. You can have multiple Funqy functions within a single application/deployment, but they must be triggered by a specific Cloud Event Type. The exception to this rule is if there is only one Funqy function in the application. In that case, the event is pushed to that function irregardless of the Cloud Event type.
Currently, Funqy can only consume JSON-based data. It supports both Binary and Structured mode of execution, but the data component of the Cloud Event message must be JSON. This JSON must also be marshallable to and from the Java parameters and return types of your functions.
The Code
Let’s start looking at our quickstart code so that you can understand how Knative Events map to Funqy. Open up SimpleFunctionChain.java
The first function we’ll look at is defaultChain
.
import io.quarkus.funqy.Funq;
public class SimpleFunctionChain {
@Funq
public String defaultChain(String input) {
log.info("*** defaultChain ***");
return input + "::" + "defaultChain";
}
As is, a Funqy function has a default Cloud Event mapping. By default, the Cloud Event type must match
the function name for the function to trigger. If the function returns output,
the response is converted into a Cloud Event and returned to the Broker to be routed to other triggers.
The default Cloud Event type for this response is the function name + .output
. The default Cloud Event source is the function name.
So, for the defaultChain
function, the Cloud Event type that triggers the function is defaultChain
. It generates
a response that triggers a new Cloud Event whose type is defaultChain.output
and the event source is defaultChain
.
While the default mapping is simple, it might not always be feasible. You can change this default mapping through configuration. Let’s look at the next function:
import io.quarkus.funqy.Funq;
public class SimpleFunctionChain {
@Funq
public String configChain(String input) {
log.info("*** configChain ***");
return input + "::" + "configChain";
}
The configChain
function has its Cloud Event mapping changed by configuration within application.properties.
quarkus.funqy.knative-events.mapping.configChain.trigger=defaultChain.output
quarkus.funqy.knative-events.mapping.configChain.response-type=annotated
quarkus.funqy.knative-events.mapping.configChain.response-source=configChain
In this case, the configuration maps the incoming Cloud Event type defaultChain.output
to the configChain
function.
The configChain
function maps its response to the annotated
Cloud Event type, and the Cloud Event source configChain
.
-
quarkus.funqy.knative-events.mapping.{function name}.trigger
sets the Cloud Event type that triggers a specific function. It is possible to use the special value*
as a catch-all value. The function will in this case be used for all event types. -
quarkus.funqy.knative-events.mapping.{function name}.response-type
sets the Cloud Event type of the response -
quarkus.funqy.knative-events.mapping.{function name}.resource-source
sets the Cloud Event source of the response
The Funqy Knative Events extension also has annotations to do this Cloud Event mapping to your functions. Take a look at the
annotatedChain
method
import io.quarkus.funqy.Funq;
import io.quarkus.funqy.knative.events.CloudEventMapping;
public class SimpleFunctionChain {
@Funq
@CloudEventMapping(trigger = "annotated", responseSource = "annotated", responseType = "lastChainLink")
public String annotatedChain(String input) {
log.info("*** annotatedChain ***");
return input + "::" + "annotatedChain";
}
If you use the @CloudEventMapping
annotation on your function you can map the Cloud Event type trigger
and the Cloud Event response. In this example the annotatedChain
function will be triggered
by the annotated
Cloud Event type and the response will be mapped to a lastChainLink
type
and annotated
Cloud Event source.
So, if you look at all the functions defined within SimpleFunctionChain
you’ll notice that one function triggers the next.
The last function that is triggered is lastChainLink
.
import io.quarkus.funqy.Context;
import io.quarkus.funqy.Funq;
public class SimpleFunctionChain {
@Funq
public void lastChainLink(String input, @Context CloudEvent event) {
log.info("*** lastChainLink ***");
log.info(input + "::" + "lastChainLink");
}
}
There are two things to notice about this function. One, it has no output. Your functions are not
required to return output. Second, there is an additional event
parameter to the function.
If you want to know additional information about the incoming Cloud Event, you can inject the
io.quarkus.funqy.knative.events.CloudEvent
interface using the Funqy @Context
annotation. The CloudEvent
interface exposes information
about the triggering event.
public interface CloudEvent {
String id();
String specVersion();
String source();
String subject();
OffsetDateTime time();
}
Maven
If you look at the POM, you’ll see that it is a typical Quarkus POM that pulls in one Funqy dependency:
<dependency>
<groupId>io.quarkus</groupId>
<artifactId>quarkus-funqy-knative-events</artifactId>
</dependency>
Dev mode and Testing
Funqy Knative Events support dev mode and unit testing using RestAssured. You can invoke on Funqy Knative Events functions using the same invocation model as Funqy HTTP using normal HTTP requests, or Cloud Event Binary mode, or Structured Mode. All invocation modes are supported at the same time.
So, if you open up the unit test code in FunqyTest.java you’ll see that its simply using RestAssured to make HTTP invocations to test the functions.
Funqy also works with Quarkus Dev mode!
Build the Project
First build the Java artifacts:
quarkus build
./mvnw install
Next, a docker image is required by Knative, so you’ll need to build that next:
docker build -f src/main/docker/Dockerfile.jvm -t yourAccountName/funqy-knative-events-quickstart .
Make sure to replace yourAccountName
with your docker or quay account name when you run docker build
. The
Dockerfile is a standard Quarkus dockerfile. No special Knative magic.
Push your image to docker hub or quay
docker push yourAccountName/funqy-knative-events-quickstart
Again, make sure to replace yourAccountName
with your docker or quay account name when you run docker push
.
Deploy to Kubernetes/OpenShift
The first step is to set up the broker in our namespace. Following is an example command from the Knative cli.
kn broker create default \
--namespace knativetutorial
The broker we have created is called default
, this broker will receive the cloud events.
The broker is also referenced in the function YAML files.
The second step is to define a Kubernetes/OpenShift service to point to the Docker image you created and pushed during build. Take a look at funqy-service.yaml
apiVersion: serving.knative.dev/v1
kind: Service
metadata:
name: funqy-knative-events-quickstart
spec:
template:
metadata:
name: funqy-knative-events-quickstart-v1
annotations:
autoscaling.knative.dev/target: "1"
spec:
containers:
- image: docker.io/yourAccountName/funqy-knative-events-quickstart
This is a standard Kubernetes service definition YAML file.
Make sure you change the image URL to point to the image you built and pushed earlier! |
For our quickstart, one Kubernetes service will contain all functions. There’s no reason you couldn’t break up this quickstart into multiple different projects and deploy a service for each function. For simplicity, and to show that you don’t have to have a deployment per function, the quickstart combines everything into one project, image, and service.
Deploy the service:
kubectl apply -n knativetutorial -f src/main/k8s/funqy-service.yaml
The next step is to deploy Knative Event triggers for each of the event types. As noted in the code section, each Funqy function is mapped to a specific Cloud Event type. You must create Knative Event triggers that map a Cloud Event and route it to a specific Kubernetes service. We have 4 different triggers.
apiVersion: eventing.knative.dev/v1alpha1
kind: Trigger
metadata:
name: defaultchain
spec:
broker: default
filter:
attributes:
type: defaultChain
subscriber:
ref:
apiVersion: serving.knative.dev/v1
kind: Service
name: funqy-knative-events-quickstart
The spec:filter:attributes:type
maps a Cloud Event type to the Kubernetes service defined in spec:subscriber:ref
.
When a Cloud Event is pushed to the Broker, it will trigger the spin up of the service mapped to that event.
There’s a trigger YAML file for each of our 4 Funqy functions. Deploy them all:
kubectl apply -n knativetutorial -f src/main/k8s/defaultChain-trigger.yaml
kubectl apply -n knativetutorial -f src/main/k8s/configChain-trigger.yaml
kubectl apply -n knativetutorial -f src/main/k8s/annotatedChain-trigger.yaml
kubectl apply -n knativetutorial -f src/main/k8s/lastChainLink-trigger.yaml
Run the demo
You’ll need two different terminal windows. One to do a curl request to the Broker, the other to watch the pod log files, so you can see the messages flowing through the Funqy function event chain.
Make sure you have the stern
tool installed. See the Knative Tutorial setup for information on that. Run stern
to look for logs outputted by our Funqy deployment
stern funq user-container
Open a separate terminal. You’ll first need to learn the URL of the broker. Execute this command to find it.
kubectl get broker default -o jsonpath='{.status.address.url}'
This will provide you a URL similar to e.g.: http://broker-ingress.knative-eventing.svc.cluster.local/knativetutorial/default
. Remember this URL.
Next thing we need to do is ssh into our Kubernetes cluster so that we can send a curl request to our broker. The following command will create a simple OS pod so we can curl into our functions.
kubectl -n knativetutorial apply -f src/main/k8s/curler.yaml
You might need to wait a couple of seconds until the curler pod comes up. Run the following to get bash access to the curler pod:
kubectl -n knativetutorial exec -it curler -- /bin/bash
You will now be in a shell within the Kubernetes cluster. Within the shell, execute this curl command , the broker address is an example and might differ based on your project or broker name.
curl -v "http://default-broker.knativetutorial.svc.cluster.local" \
-X POST \
-H "Ce-Id: 1234" \
-H "Ce-Specversion: 1.0" \
-H "Ce-Type: defaultChain" \
-H "Ce-Source: curl" \
-H "Content-Type: application/json" \
-d '"Start"'
This posts a Knative Event to the broker, which will trigger the defaultChain
function. As discussed earlier, the output
of defaultChain
triggers an event that is posted to configChain
which triggers an event posted to annotatedChain
then
finally to the lastChainLink
function. You can see this flow in your stern
window. Something like this should
be outputted.
funqy-knative-events-quickstart-v1-deployment-59bb88bcf4-9jwdx user-container 2020-05-12 13:44:02,256 INFO [org.acm.fun.SimpleFunctionChain] (executor-thread-1) *** defaultChain ***
funqy-knative-events-quickstart-v1-deployment-59bb88bcf4-9jwdx user-container 2020-05-12 13:44:02,365 INFO [org.acm.fun.SimpleFunctionChain] (executor-thread-2) *** configChain ***
funqy-knative-events-quickstart-v1-deployment-59bb88bcf4-9jwdx user-container 2020-05-12 13:44:02,394 INFO [org.acm.fun.SimpleFunctionChain] (executor-thread-1) *** annotatedChain ***
funqy-knative-events-quickstart-v1-deployment-59bb88bcf4-9jwdx user-container 2020-05-12 13:44:02,466 INFO [org.acm.fun.SimpleFunctionChain] (executor-thread-2) *** lastChainLink ***
funqy-knative-events-quickstart-v1-deployment-59bb88bcf4-9jwdx user-container 2020-05-12 13:44:02,467 INFO [org.acm.fun.SimpleFunctionChain] (executor-thread-2) Start::defaultChain::configChain::annotatedChain::lastChainLink