Deploy A Backend App As An Android Engineer

With the Skills You Already Have

Original photo by SpaceX on Unsplash

Update 10/24/18: Deploy A Backend App As An Android Engineer — Iterate Faster With Ktor

There’s a lot that can be done as a mobile developer working with existing APIs to spin up a proof of concept. Although when logic needs to run constantly in the background to check data for a condition a backend app is a better fit as you don’t want to turn your mobile device into a mobile server. If you ask a backend engineer working at a startup to a medium sized company company in San Francisco you may receive the suggestion to learn Javascript (specifically Node.js), with the promise of it being quick to write once familiar with Javascript and easy to deploy.

After asking a handful of Node.js developers I had the same conversation with my friend Fawn, who has been doing Node for a few years. I was expecting the same answer as she reaffirmed the above estimating it would take me ~2–3 weeks to ramp up with Javascript (Js). Then, she suggested an alternative:

It‘s useful to learn Js, but you already know how to do this in Java and can build it today.

Me:

Fawn was right. Most Android engineers have the skills to build basic backend apps given the best practices and libraries they already have. For instance, Retrofit can be used to make network requests with RxJava to handle the stream of incoming and outgoing data and do advanced manipulations while keeping the code bloat free with Kotlin. Using this approach I’ve started deploying backend instances to collect and manipulate data from multiple APIs. I’m working with Google Cloud Platform’s (GCP) AppEngine with Firebase integration to save data that can be accessed easily from a mobile/web client. Firebase can also integrate directly with Google’s BigQuery for further analytics (From Firestore to BigQuery with Firebase Functions).

Google Cloud Platform is easy to use once setup and insanely powerful. That said, knowing what you need to know is a research project in itself. After you have a rough outline of the steps required for your project then you can begin filtering through documentation and guides for the ones that match the end goal. The hope is that the steps and resources below save you a significant portion of time and allow you to jump right to the execution phase.

Hosting a Kotlin server app appears to have more setup than what I’ve seen advertised for Node.js with “one-click” deploys as seen on Digital Ocean, but once the process is setup with IntelliJ it takes under a minute to start the deploy. If you get into more complex computation there is also the inherit benefit of multi-threading with Kotlin/Java.

Overview

Figuring out the required steps required a scavenger hunt. I sifted through overlapping Google and IntelliJ documentation and guides regarding AppEngine Flexible vs Standard (still unsure of the important difference, I just know Flexible integrates well with both IntelliJ and GCP), Apache and Maven integration, Java website hosting, etc. None of the notes simply outlined how a Kotlin/Java needs to be formatted in order to be hosted onto a server.

Once the app logic is finished build a .JAR file which is what will run on the server.
Once the .JAR file is built you can test whether it runs successfully locally in IntelliJ.
Deploy a .JAR file to the server with deployment configurations which will run on the server until paused or stopped.

Now into the details…

The good news is after an app is setup for the first time you can skip to Step 4: Managing The Server which takes less than 1 minute to setup a redeploy.

Setup Local Development Environment — Step 1 of 4

GCP AppEngine Setup

1. Follow the steps for Install the latest Cloud Tools…

a) Unzip directory and store in Library directory > call one level outside of cloud directory.

./google-cloud-sdk/install.sh

b) Restart Terminal for gcloud commands to work.

2. Install App Engine and Login.

sudo gcloud components install app-engine-java

...

gcloud auth application-default login

IntelliJ Configuration

Follow the steps under Configuring your development environment.
Use an existing gradle or maven project in IntelliJ or create a new IntelliJ project (Make sure to download the IntelliJ Google Cloud Tools plugin and restart).

Create GCP Project And Integrate With IntelliJ — Step 2 of 4

GCP Setup

Create New Project under GCP console with Quickstart steps and ensure Billing is enabled for the project.

Integrate With IntelliJ

Setup normal Java build configuration under Edit Configurations with projectModuleName_main.
Either enable current open project with GCP, start a new one, or use an existing project from a GitHub repo.

For current open project: choose Tools > Google Cloud Tools > Add AppEngine Support > Google AppEngine Flexible> select to create .yaml when given the choice and make sure it’s under src/main/appengine/app.yaml > when adding framework support choose projectModuleName (not projectName_main or projectName_test)

app.yaml

If the app is handling a task in order to populate data for a backend service that does not process user network requests than making sure only one instance is running is important. Otherwise multiple instances can populate the backend with duplicate data. (see Configuring your App with app.yaml)

manual_scaling:instances: 1

Generate Deployable JAR — Step 3 of 4

Packaging the application in a JAR

File > Project Structure > Project Settings > Artifacts > + > Jar > From modules with dependencies…

Module: projectModuleName_main
MainClass:YourMainClassWithMainMethod
Manifest: src/main/manifestName
Output directory: projectName/out/artifacts/jarName_jar

Build and run JAR

Build > Build Artifacts… > Action > Build
Run Jar: Run > Edit Configurations: + and select JAR Application (Before launch: + > Build Artifacts > artifactName)

Configure And Deploy to GCP — Step 4 of 4

Deployment Configuration

Deployment archive: yourPathRoute/projectJarName.jar
Project: select from GCP
app.yaml: select from file path
Run!

Debugging

It may fail the first time if Billing is not set up > click the link in the output and enable Billing.
(7/25/2018) There is a bug in the deploy process with AppEngine if there is a .json file used in the app. AppEngine will not process .json files included in the .jar despite the .jar reading the files successfully when being ran on it’s own in IntelliJ. I’ve outlined a workaround on Stackoverflow involving creating an object and converting it to json using the Gson library. This is relevant when needing to use .json files to authenticate services such as Firebase.

Managing The Server

Start/Stop server: You can control your server in the Google Cloud Console under AppEngine > Versions.

Redeploy future versions

Stop the current version of the server app under AppEngine > Versions.
Under Build > Build Artifacts… > Rebuild which will rebuild the .JAR.
Deploy with same AppEngine configuration above since it should be pointing to the same .JAR path.

Scaling

AppEngine will take care of scaling automatically so it is important to understand the pricing and monitor the app if you are doing more than just prototyping.

I’m Adam Hurwitz — hit the clapping hands icon and check out the rest of my writing if you enjoyed the above. Thanks for reading!

Resources

GCP

IntelliJ

JAR

Firebase

From Firestore to BigQuery with Firebase Functions

Online Communities