- Introduction
- Configuration
- Resiliency
- Service Discovery and Load Balancing
- Testing
- Integration with Microcks
- Benchmarking with Hyperfoil
- Running the Application
- Running Locally via Docker Compose
- Deploying to Kubernetes
This is the Fight REST API microservice. It is a reactive HTTP microservice exposing an API for performing fights between Heroes and Villains, in a location obtained from the Location Service. Once a winner has been determined, the Narration service creates a narration of the fight.
Each fight and it's corresponding narration is then persisted into a MongoDB database and can be retrieved via the REST API. This service is implemented using RESTEasy Reactive with reactive endpoints and Quarkus MongoDB Reactive with Panache's active record pattern.
Fight messages are also published on an Apache Kafka topic called fights
. The event-statistics service listens for these events. Messages are stored in Apache Avro format and the fight schema is automatically registered in the Apicurio Schema Registry. This all uses built-in extensions from Quarkus.
The following table lists the available REST endpoints. The OpenAPI document for the REST endpoints is also available.
Path | HTTP method | Response Status | Response Object | Description |
---|---|---|---|---|
/api/fights |
GET |
200 |
List<Fight> |
All Fights. Empty array ([] ) if none. |
/api/fights |
POST |
200 |
Fight |
Performs a fight. |
/api/fights |
POST |
400 |
Invalid FightRequest passed in request body (or no request body found) |
|
/api/fights/randomfighters |
GET |
200 |
Fighters |
Finds random fighters |
/api/fights/randomlocation |
GET |
200 |
FightLocation |
Finds a random fight location |
/api/fights/{id} |
GET |
200 |
Fight |
Fight with id == {id} |
/api/fights/{id} |
GET |
404 |
No Fight with id == {id} found |
|
/api/fights/narrate |
POST |
200 |
String |
Performs a narration of the given Fight |
/api/fights/narrate |
POST |
400 |
String |
Invalid Fight passed in request body (or no request body found) |
/api/fights/narrate/image |
POST |
200 |
FightImage |
Generate an image and caption using DALL-E for a narration |
/api/fights/narrate/image |
POST |
400 |
Invalid narration passed in | |
/api/fights/hello/heroes |
GET |
200 |
String |
Invokes the "hello" endpoint of the Heroes microservice |
/api/fights/hello/villains |
GET |
200 |
String |
Invokes the "hello" endpoint of the Villains microservice |
/api/fights/hello/narration |
GET |
200 |
String |
Invokes the "hello" endpoint of the Narration microservice |
/api/fights/hello/locations |
GET |
200 |
String |
Invokes the "hello" endpoint of the Location microservice |
The FightConfig
stores all the application-specific configuration that can be overridden at runtime.
The FightService
class uses timeouts from SmallRye Fault Tolerance to protect against calls to the downstream Hero, Villain, Narration, and Location services. Tests for these conditions can be found in FightServiceTests
.
The FightService
class uses fallbacks from SmallRye Fault Tolerance to protect against calls to the downstream Hero, Villain, Narration, and Location services. Tests for these conditions can be found in FightServiceTests
.
Retry logic to the downstream Hero, Villain, Narration, and Location services is implemented in the clients for each service.
The HeroRestClient
is implemented using the reactive rest client. All of its configuration can be found in application.properties
under the quarkus.rest-client.hero-client
key. This client is not exposed outside of the io.quarkus.sample.superheroes.fight.client
package.
Instead, the HeroClient
class wraps the HeroRestClient
and adds some resiliency to it:
- The downstream Hero service returns a
404
if no randomHero
is found.HeroClient
handles this case and simulates the service returning nothing. - In the event the downstream Hero service returns an error,
HeroClient
adds 3 retries with a 200ms delay between each retry.
The VillainClient
is implemented using the JAX-RS client API with the RESTEasy Reactive client. All of its configuration can be found in application.properties
under the fight.villain.client-base-url
key.
- The downstream Villain service returns a
404
if no randomVillain
is found.VillainClient
handles this case and simulates the service returning nothing. - In the event the downstream Villain service returns an error,
VillainClient
adds 3 retries with a 200ms delay between each retry.
The NarrationClient
is implemented using the reactive rest client. All of its configuration can be found in application.properties
under the quarkus.rest-client.narration-client
key.
The LocationClient
is implemented using the Quarkus gRPC client. All of its configuration can be found in application.properties
under the quarkus.grpc.clients.locations
key.
The fight service implements service discovery and client-side load balancing when making downstream calls to the rest-heroes
, rest-villains
, and rest-narration
services. The service discovery is implemented in Quarkus using SmallRye Stork.
Stork integrates directly with the Quarkus REST Client Reactive. This means that there is no additional code needed in order to take advantage of Stork's service discovery and client-side load balancing.
Important
You could disable Stork completely for the HeroRestClient
by setting quarkus.rest-client.hero-client.url
to any non-Stork URL (i.e. something that doesn't start with stork://
). Similarly, you could disable Stork completely for the VillainClient
by setting fight.villain.client-base-url
to any non-Stork URL or for the NarrationClient
by setting quarkus.rest-client.narration-client.url
to any non-Stork URL.
In local development mode, as well as when running via Docker Compose, SmallRye Stork is configured using static list discovery. In this mode, the downstream URLs are statically defined in an address list. In application.properties
, see the quarkus.stork.hero-service.service-discovery.address-list
, quarkus.stork.villain-service.service-discovery.address-list
, and quarkus.stork.narration-service.service-discovery.address-list
properties.
When running in Kubernetes, Stork is configured to use the Kubernetes Service Discovery. In this mode, Stork will read the Kubernetes Service
s for the rest-heroes
, rest-villains
, and rest-narration
services to obtain the instance information. Additionally, the instance information has been configured to refresh every minute. See the rest-fights-config
ConfigMap in the Kubernetes deployment descriptors. Look for the quarkus.stork.*
properties within the various ConfigMap
s.
All of the other Stork service discovery mechanisms (Consul and Eureka) can be used simply by updating the configuration appropriately according to the Stork documentation.
In all cases, the default load balancing algorithm used is round robin. All of the other load balancing algorithms (random, least requests, least response time, and power of two choices) are available on the application's classpath, so feel free to play around with them by updating the configuration appropriately according to the Stork documentation.
This application has a full suite of tests, including an integration test suite.
- The test suite uses Wiremock for mocking http calls (see
HeroesVillainsNarrationWiremockServerResource
) to the downstream Hero, Villain, and Narration services. - The test suite configures the application to use the in-memory connector from SmallRye Reactive Messaging (see the
%test.mp.messaging.outgoing.fights
configuration inapplication.properties
) for verifying interactions with Kafka. - The integration test suite uses Quarkus Dev Services (see
KafkaConsumerResource
) to interact with a Kafka instance so messages placed onto the Kafka broker by the application can be verified.
Pact is a code-first tool for testing HTTP and message integrations using contract tests
. Contract tests assert that inter-application messages conform to a shared understanding that is documented in a contract. Without contract testing, the only way to ensure that applications will work correctly together is by using expensive and brittle integration tests.
Eric Deandrea and Holly Cummins recently spoke about contract testing with Pact and used the Quarkus Superheroes for their demos. Watch the replay and view the slides if you'd like to learn more about contract testing.
The rest-fights
application is both a Pact Consumer and a Pact Provider. As a Consumer, it should be responsible for defining the contracts between itself and its providers (rest-heroes
, rest-villains
, rest-narration
, & grpc-locations
). As a Provider, is should run provider verification tests against contracts produced by consumers.
As this README states, contracts generally should be hosted in a Pact Broker and then automatically discovered in the provider verification tests.
One of the main goals of the Superheroes application is to be super simple and just "work" by anyone who may clone this repo. That being said, we can't make any assumptions about where a Pact broker may be or any of the credentials required to access it.
The FightServiceConsumerContractTests.java
test class generates the rest-fights-rest-heroes.json
, rest-fights-rest-villains.json
, rest-fights-rest-narration.json
, and rest-fights-grpc-locations.json
contracts while also providing mock instances of the rest-heroes
, rest-villains
, rest-narration
, and grpc-locations
providers.
Note
The grpc-locations
service uses gRPC/protobuf and not REST, therefore the consumer contract tests between rest-fights
and grpc-locations
use the Pact protobuf plugin. There is no installation necessary. When the tests execute the plugin will be automatically installed.
The contracts are committed into the provider's version control simply for easy of use and reproducibility.
Additionally, the Pact contract is committed into this application's source tree inside the src/test/resources/pacts
directory.
The consumer contract tests and provider verification tests ARE executed during this project's CI/CD processes. They run against any pull requests and any commits back to the main
branch.
The Pact tests use the Quarkus Pact extension. This extension is recommended to give the best user experience and ensure compatibility.
Microcks is an open source tool for API mocking and testing. It allows developers to turn an API contract or Postman Collection into live mocks.
This can be especially useful while developing applications that have downstream dependencies, like the rest-fights
application does. The rest-fights
application depends on rest-heroes
, rest-villains
, rest-narration
, and grpc-locations
. Due to the resiliency built into rest-fights
, the rest-fights
application can function without the downstream dependencies. However, it would be nice to be able to live code in rest-fights
and have mock data get served to rest-fights
.
This problem is what the Microcks Quarkus extension solves. This extension has been integrated into rest-fights
so that when live coding in dev mode, mock responses from rest-heroes
, rest-villains
, rest-narration
, and grpc-locations
will automatically get served.
Furthermore, the Microcks user interface is accessible from the Quarkus Dev UI:
If you wish to disable this functionality, simply add -Dquarkus.profile=no-microcks
when you run Quarkus dev mode (via Maven, Gradle, or the Quarkus CLI). In this case, the Microcks dev service will be disabled and the rest-fights
application will attempt to make live calls to the downstream services.
There are some Hyperfoil benchmarks in this directory. See the README for more details.
First you need to start up all of the downstream services (Heroes Service, Villains Service, and Location Service - the Narration Service and Event Statistics Service are optional).
The application runs on port 8082
(defined by quarkus.http.port
in application.properties
).
From the quarkus-super-heroes/rest-fights
directory, simply run ./mvnw quarkus:dev
to run Quarkus Dev Mode, or running quarkus dev
using the Quarkus CLI. The application will be exposed at http://localhost:8082
and the Quarkus Dev UI will be exposed at http://localhost:8082/q/dev
. Quarkus Dev Services will ensure the MongoDB instance, an Apache Kafka instance, and an Apicurio Schema Registry are all started and configured.
Note
Running the application outside of Quarkus Dev Mode requires standing up a MongoDB instance, an Apache Kafka instance, and an Apicurio Schema Registry and binding them to the app.
Furthermore, since this service also communicates with additional downstream services (rest-heroes, rest-villains, & rest-narration), those would need to be stood up as well, although this service does have fallbacks in case those other services aren't available.
By default, the application is configured with the following:
Description | Environment Variable | Java Property | Value |
---|---|---|---|
Database Host | QUARKUS_MONGODB_HOSTS |
quarkus.mongodb.hosts |
localhost:27021 |
Database username | QUARKUS_MONGODB_CREDENTIALS_USERNAME |
quarkus.mongodb.credentials.username |
superfight |
Database password | QUARKUS_MONGODB_CREDENTIALS_PASSWORD |
quarkus.mongodb.credentials.password |
superfight |
Kafka Bootstrap servers | KAFKA_BOOTSTRAP_SERVERS |
kafka.bootstrap.servers |
PLAINTEXT://localhost:9092 |
Apicurio Schema Registry | MP_MESSAGING_CONNECTOR_SMALLRYE_KAFKA_APICURIO_REGISTRY_URL |
mp.messaging.connector.smallrye-kafka.apicurio.registry.url |
http://localhost:8086/apis/registry/v2 |
Heroes Service URL | QUARKUS_REST_CLIENT_HERO_CLIENT_URL |
quarkus.rest-client.hero-client.url |
stork://hero-service |
Villains Service URL | FIGHT_VILLAIN_CLIENT_BASE_URL |
fight.villain.client-base-url |
stork://villain-service |
Narration Service URL | QUARKUS_REST_CLIENT_NARRATION_CLIENT_URL |
quarkus.rest-client.narration-client.url |
stork://narration-service |
Location Service Host | QUARKUS_GRPC_CLIENTS_LOCATIONS_HOST |
quarkus.grpc.clients.locations.host |
localhost |
Location Service Port | QUARKUS_GRPC_CLIENTS_LOCATIONS_PORT |
quarkus.grpc.clients.locations.port |
localhost |
Pre-built images for this application can be found at quay.io/quarkus-super-heroes/rest-fights
.
Pick one of the versions of the application from the table below and execute the appropriate docker compose command from the quarkus-super-heroes/rest-fights
directory.
Note
You may see errors as the applications start up. This may happen if an application completes startup before one if its required services (i.e. database, kafka, etc). This is fine. Once everything completes startup things will work fine.
Description | Image Tag | Docker Compose Run Command |
---|---|---|
JVM Java 21 | java21-latest |
docker compose -f deploy/docker-compose/java21.yml up --remove-orphans |
Native | native-latest |
docker compose -f deploy/docker-compose/native.yml up --remove-orphans |
The above Docker Compose files are meant for standing up this application and the required database, Kafka broker, and Apicurio Schema Registry only. If you want to stand up this application and its downstream services (rest-villains, rest-heroes, rest-narration, & grpc-locations), pick one of the versions from the table below and execute the appropriate docker compose command from the quarkus-super-heroes/rest-fights
directory.
Note
You may see errors as the applications start up. This may happen if an application completes startup before one if its required services (i.e. database, kafka, etc). This is fine. Once everything completes startup things will work fine.
Description | Image Tag | Docker Compose Run Command |
---|---|---|
JVM Java 21 | java21-latest |
docker compose -f deploy/docker-compose/java21-all-downstream.yml up --remove-orphans |
Native | native-latest |
docker compose -f deploy/docker-compose/native-all-downstream.yml up --remove-orphans |
If you want to develop the Fights service (i.e. via Quarkus Dev Mode) but want to stand up just it's downstream services (rest-villains, rest-heroes, rest-narration, & grpc-locations), pick one of the versions from the table below and execute the appropriate docker compose command from the quarkus-super-heroes
directory.
Note
You may see errors as the applications start up. This may happen if an application completes startup before one if its required services (i.e. database, kafka, etc). This is fine. Once everything completes startup things will work fine.
Description | Image Tag | Docker Compose Run Command |
---|---|---|
JVM Java 21 | java21-latest |
docker compose -f rest-heroes/deploy/docker-compose/java21.yml -f rest-villains/deploy/docker-compose/java21.yml -f rest-narration/deploy/docker-compose/java21.yml -f grpc-locations/deploy/docker-compose/java21.yml up --remove-orphans |
Native | native-latest |
docker compose -f rest-heroes/deploy/docker-compose/native.yml -f rest-villains/deploy/docker-compose/native.yml -f rest-narration/deploy/docker-compose/native.yml -f grpc-locations/deploy/docker-compose/java21.yml up --remove-orphans |
If you want to stand up the entire system, follow these instructions.
Once started the application will be exposed at http://localhost:8082
. The Apicurio Schema Registry will be exposed at http://localhost:8086
.
The application can be deployed to Kubernetes using pre-built images or by deploying directly via the Quarkus Kubernetes Extension. Each of these is discussed below.
Pre-built images for this application can be found at quay.io/quarkus-super-heroes/rest-fights
.
Deployment descriptors for these images are provided in the deploy/k8s
directory. There are versions for OpenShift, Minikube, Kubernetes, and Knative.
Note
The Knative variant can be used on any Knative installation that runs on top of Kubernetes or OpenShift. For OpenShift, you need OpenShift Serverless installed from the OpenShift operator catalog. Using Knative has the benefit that services are scaled down to zero replicas when they are not used.
Pick one of the versions of the application from the table below and deploy the appropriate descriptor from the deploy/k8s
directory.
Description | Image Tag | OpenShift Descriptor | Minikube Descriptor | Kubernetes Descriptor | Knative Descriptor |
---|---|---|---|---|---|
JVM Java 21 | java21-latest |
java21-openshift.yml |
java21-minikube.yml |
java21-kubernetes.yml |
java21-knative.yml |
Native | native-latest |
native-openshift.yml |
native-minikube.yml |
native-kubernetes.yml |
native-knative.yml |
The application is exposed outside of the cluster on port 80
.
These are only the descriptors for this application and the required database, Kafka broker, and Apicurio Schema Registry only. If you want to deploy this application and its downstream services (rest-villains, rest-heroes, rest-narration, & grpc-locations), pick one of the versions of the application from the table below and deploy the appropriate descriptor from the rest-fights/deploy/k8s
directory.
Description | Image Tag | OpenShift Descriptor | Minikube Descriptor | Kubernetes Descriptor | Knative Descriptor |
---|---|---|---|---|---|
JVM Java 21 | java21-latest |
java21-openshift-all-downstream.yml |
java21-minikube-all-downstream.yml |
java21-kubernetes-all-downstream.yml |
java21-knative-all-downstream.yml |
Native | native-latest |
native-openshift-all-downstream.yml |
native-minikube-all-downstream.yml |
native-kubernetes-all-downstream.yml |
native-knative-all-downstream.yml |
Each application is exposed outside of the cluster on port 80
.
If you want to deploy the entire system, follow these instructions.
Following the deployment section of the Quarkus Kubernetes Extension Guide (or the deployment section of the Quarkus OpenShift Extension Guide if deploying to OpenShift), you can run one of the following commands to deploy the application and any of its dependencies (see Kubernetes (and variants) resource generation of the automation strategy document) to your preferred Kubernetes distribution.
Note
For non-OpenShift or minikube Kubernetes variants, you will most likely need to push the image to a container registry by adding the -Dquarkus.container-image.push=true
flag, as well as setting the quarkus.container-image.registry
, quarkus.container-image.group
, and/or the quarkus.container-image.name
properties to different values.
Target Platform | Java Version | Command |
---|---|---|
Kubernetes | 21 | ./mvnw clean package -Dquarkus.profile=kubernetes -Dquarkus.kubernetes.deploy=true -DskipTests |
OpenShift | 21 | ./mvnw clean package -Dquarkus.profile=openshift -Dquarkus.container-image.registry=image-registry.openshift-image-registry.svc:5000 -Dquarkus.container-image.group=$(oc project -q) -Dquarkus.kubernetes.deploy=true -DskipTests |
Minikube | 21 | ./mvnw clean package -Dquarkus.profile=minikube -Dquarkus.kubernetes.deploy=true -DskipTests |
Knative | 21 | ./mvnw clean package -Dquarkus.profile=knative -Dquarkus.kubernetes.deploy=true -DskipTests |
Knative (on OpenShift) | 21 | ./mvnw clean package -Dquarkus.profile=knative-openshift -Dquarkus.container-image.registry=image-registry.openshift-image-registry.svc:5000 -Dquarkus.container-image.group=$(oc project -q) -Dquarkus.kubernetes.deploy=true -DskipTests |
You may need to adjust other configuration options as well (see Quarkus Kubernetes Extension configuration options and Quarkus OpenShift Extension configuration options).
Tip
The do_build
function in the generate-k8s-resources.sh
script uses these extensions to generate the manifests in the deploy/k8s
directory.