Example of a microservice that works on Spring Boot.
- Java 8
To make setting up microservices in micro-time we needed to extract the common building blocks to separate libraries. In this template we are using 4finance custom library, i.e. micro-infra-spring (available in jcenter).
There is a version with web application connected with backend side. You can find it on the branch of this repo: boot-microservice-gui.
You can remove all the caching related libraries, annotations and com.ofg.twitter packages. Next adjust all the properties and then you have an empty project - you're now ready to go.
You can run
./gradlew setup --no-daemon
and follow the instructions that will help you set up your new microservice in no time!
./gradlew bootRun<Profile>
Where <Profiles> is a comma separated list of profiles, eg. bootRunProd, bootRunProd,smokeTests. If no profile provided then dev' and 'springCloud are set as default. There's also possible to pass any property using -D switch.
Run Application class with the following VM args:
-DAPP_ENV="prod" -DCONFIG_FOLDER="properties" -Dencrypt.key="secretEncryptKey" -Dspring.profiles.active=dev
To run it just
- create a jar -
./gradlew clean build - go to
build/libsto find the fat jar - execute
java -jarwith the same arguments as running from IntelliJ IDEA
java -jar boot-microservice.jar runnerArgs
Running:
./gradlew docker
will create /build/docker/Dockerfile.
You can build and run it using docker-compose.yml file in project root:
docker-compose up -d
Run this command to see an output:
docker-compose logs
You can use Dockerfile to create Docker image:
docker build -t boot-microservice build/docker
Self-sufficient Docker image with our sample microservice can be started as follows:
docker run \
-e spring.profiles.active=dev \
-e APP_ENV=prod \
-e ENCRYPT_KEY=secretEncryptKey \
-p 8080:8095 \
boot-microservice
Test with curl localhost:8080/ping. Notice that we run it in dev profile (in-memory embedded ZooKeeper and stubs) and we re-map 8095 port to 8080 on host machine.
If you want to run microservice with real service discovery via ZooKeeper and Graphite, first prepare Docker images for that:
docker run --name zookeeper jplock/zookeeper
docker run --name graphite -p 8081:80 kamon/grafana_graphite
After the first time these containers can be executed with docker start -a zookeeper and docker start -a graphite shorthands. Now you can run arbitrary number of microservices and they will all register themselves in ZooKeeper/Graphite instances:
docker run \
-p 8080:8080 \
-e spring.profiles.active=prod \
--link zookeeper:zk \
--link graphite:gr \
boot-microservice \
--service.resolver.url=zk:2181 \
--graphite.host=gr
We link microservice container with ZooKeeper container aliasing it to zk. This way microservice sees ZooKeeper container under zk hostname and we can simply point service.resolver.url=zk:2181 (zk:2181 is a valid network address from boot-microservice container's perspective). Same applies to Graphite.
Your microservice (assuming it exposes 8080 port) will be visible outside under 8080 as well. Moreover you can browse to localhost:8081 and browse Grafana.
Below you can find description of the most crucial parts of the application's production code.
contains Spring Boot autoconfiguration and contains main method
Spring boot auto-configure application and enable support for 4finance's "Micro Infra Spring". That enables module like ServiceDiscovery, Swagger, CorrelationId filters, custom RestTemplate, custom exception handling, health check controllers etc. Check out Micro Infra Spring's Boot Starter module's Spring factories set up
If you want only certain modules of the system just check out 4finance's Micro Infra Spring's readme.
Below you can find description of the most crucial parts of the application's test code.
extends com.ofg.infrastructure.base.IntegrationSpec Spock Specification class that initializes Spring web-context.
extends com.ofg.infrastructure.base.MvcIntegrationSpec Spock Specification class that initializes Spring web-context and provides some additional Mvc related fields.
extends com.ofg.infrastructure.base.MvcWiremockIntegrationSpec Spock Specification class that extends the MvcIntegrationSpec spec. Additionally it provides WireMock related fields and methods.
you may wonder - how on earth does the collaborator collerator respond to with 200 when you post him at /1 ?! It's all about Consumer Driven Contracts and our implementation called stub-runner-spring.
What happens under the hood is that the stubs are downloaded from 4finance Bintray account. A jar of the stub-runner-examples is downloaded and unpacked to a temporary folder and all the tests are ran against it. The stub is in fact here twitter-places-collerator stub!
The paths to the repo, the module name and artifactId are here - twitter-places-analyzer.yaml.
Twitter places analyzer, searches through tweets for places. Then analyzers send those to Collerators.
Hit PUT at:
/api/{pairId}
with list of tweets:
[
{
"created_at": "Sat Jul 26 09:15:10 +0000 2014",
"id": 492961315070439424,
"id_str": "492961315070439424",
"geo": null,
"coordinates":
{
"coordinates":
[
-75.14310264,
40.05701649
],
"type":"Point"
},
},
{
"created_at": "Sat Jul 26 09:15:10 +0000 2014",
"id": 492961315070439424,
"id_str": "492961315070439424",
"geo": null,
"coordinates":
{
"coordinates":
[
-75.14310264,
40.05701649
],
"type":"Point"
},
}
]And it will hit collectors at /{pairId} with tweets taken from twitter
[
{
"pair_id" : 1,
"tweet_id" : "492967299297845248",
"place" :
{
"name":"Washington",
"country_code": "US"
},
"probability" : "2",
"origin" : "twitter_place_section"
},
{
"pair_id" : 2,
"tweet_id" : "123187246819263981"
},
]
This is an extension of the non-GUI boot-microservice. To read about the backend side, please refer to boot-microservice.
The webapp is generated by yeoman using the angular generator.
It uses grunt that will automatically 'compile' your whole application, allowing cool dev mode with live reloads.
Inside you will find bower for javascript dependency management and node with npm that grunt uses.
Before first use, build your whole application with gradle build. It will download auto-magically all npms and bower dependencies.
Then run you application (for example from Idea, just run main in com.ofg.twitter.Application specifying
the correct -Dspring.profiles.active).
Now your application (backend) works. But you still need js+html. And since this is 2015, you don't just write html anymore, you have to use a shitload of libs :)
Install npm if you don't have it already. For example on Debian-based linux run:
sudo apt-get install npm
Then install grunt.
sudo npm install -g grunt-cli
And now make a symbolic link, because nodejs from Debian repos has a wrong name
ln -s /usr/bin/nodejs /usr/bin/node
Next, go to src/main/web and type npm install. This will download all needed libraries defined in package.json file and install bower components at the end. After this you can finally grunt serve. This will run a local webserver on port 9000, your application will
automatically open in the browser and from now on on every change in you webapp the browser will automatically refresh
(no need to hit cmd-R all the time!).
Easy, right? Writing HTML in 2015 is simple... nooooot! :D
When you build your application there is a special directory src/main/web/dist created, where grunt puts your minified, compacted
and production-ready application. Then the whole folder will be bundled in your jar's static folder which will make
it available when you run the jar.
Once you expose some REST services on your backend, you will probably want to call them from Angular.
To make it possible in the Development mode, you will have to expose them via proxy (your dev page is available
on port 9000 while the app is on 8080, remember?). Look into src/main/web/Gruntfile.js and inside connect you will have
proxies section like this
connect: {
proxies: [
{context: '/info', host: 'localhost', port: 8080},
{context: '/city', host: 'localhost', port: 8080},
{context: '/api', host: 'localhost', port: 8080}
],Just add whatever you wish. If you don't like exposing every service explicitly, you can expose them all in spring under some common path like /rest, and then you have to specify only the /rest in the proxy.
Type gradle cleanGUIDeps