How To Microservice

Perhaps a misnomer, this is a personal reference for a small Java REST app that more or less adheres to the principles of the 12 Factor App.

Run

./gradlew run

The application is packaged to be deployed in one of two ways:

As a systemd service

# build rpm
./gradlew rpm
# install in CentOS 7 vagrant machine for testing
vagrant up

As a docker container

# build docker image
./gradlew jibDockerBuilder
# run docker container
docker run --rm gilday/how-to-microservice

Jersey for REST

The app embraces Jersey and the JAX-RS API instead of the Servlet API. Avoiding the servlet API makes the app more portable. Specifically, it allows the Jersey Test Framework to use the JdkHttpServerFactory which initializes quicker than a full servlet container.

Jetty for HTTP

The app espouses "container-less" deployments i.e. the build artifact is not a WAR that must be deployed to a servlet container. Rather, the app embeds the Jetty web server as a dependency. The main class App configures and starts the Jetty web server.

Jackson for JSON

Configures Jersey to use Jackson and a customized ObjectMapper for JSON serialization.

Logback for Logging

Application logging is done via the SLF4J logging API with a Logback implementation. Logback is configured to send all messages to STDOUT which makes development easy to debug and allows production deployments to "treat logs as event streams" as presribed by the 12 Factor App.

HK2 for IOC

HK2 is a light-weight IOC framework (JSR-330) which is used internally and exposed by Jersey. While it is possible to use other IOC frameworks with Jersey, HK2 does a fine job and doesn't require additional dependencies.

Typesafe Config for Configuration

Typesafe Config is a comprehensive, light-weight configuration library for JVM apps. It has good support for overriding default configuration parameters with external config files or environment variables. Its flexibility makes it appropriate for different types of deployments such as external configuration files in /etc/how-to-microservice or environment variables from a PaaS such as Heroku.

Application plugin for easy deployments

The standard Gradle Application plugin generates distribution tar and zip files which contain the application, its dependencies, and scripts for starting the application. The easiest way to use the Application plugin to run the application is ./gradlew run. Use ./gradlew distZip and ./gradlew distTar to build redistributable packages.

nebula for RPM deployments

Netflix's Nebula gradle plugins include a plugin for building RPM distributions. Gradle is configured to build an RPM which installs the app as a systemd daemon on Enterprise Linux 7. The RPM does the following

• includes Java 11 as a dependency
• adds an unprivileged user to own the app's files and process
• installs a systemd unit file to register the app as a systemd daemon
• installs an external configuration file to /etc/how-to-microservice/settings.conf
• installs a script for setting envrionment variable JVM_OPTS for passing arguments such as javax.net.ssl.keystore

The project uses a Vagrant file to test the RPM build. After building an RPM with gradle clean build rpm, one can use vagrant up to create a new CentOS 7 VirtualBox machine which installs the newly built RPM and runs the daemon. VirtualBox forwards traffic from port 8001 to the machine's port 8000 for testing the app e.g. curl http://localhost:8001/greeting.

jib for container deployments

Project is configured to use jib to build a container image pushed to Docker Hub at gilday/how-to-microservice. Deployments to Docker Hub requires authentication. To build a test the image with a local docker daemon, use the jibDockerBuild task.