spark-as-service-using-embedded-server

This application comes as Spark2.1-REST-Service-Provider using an embedded, Reactive-Streams-based, fully asynchronous HTTP server.

1. Central Idea

I wanted to build an interactive REST api service on top of my ApacheSpark application which serves use-cases like:

- Load the trained model in SparkSession and quickly do the prediction for user given query._
- Have your big-data cached in cluster and provide user an endpoint to query it.
- Run some recurrent spark queries with varying parameters.

As you can see that the core of the application is not primarily a web-application OR browser-interaction but to have REST service performing big-data cluster-computation on ApacheSpark.

2. Akka-HTTP as apt-fit:

With Akka-Http, you normally don’t build your application on top of Akka HTTP, but you build your application on top of whatever makes sense and use Akka HTTP merely for the HTTP integration needs. So, I found Akka-HTTP to be right fit for the usecases mentioned above.

3. Architecture

3.1 To demo this, I've configured following four routes:

homepage - http://localhost:8001 - says "hello world"
version - http://localhost:8001/version - queries shared SparkSession and tells "spark version"
activeStreams - http://localhost:8001/activeStreams - tells how many spark streams are active currently
count - http://localhost:8001/count - random spark job to count number of elements in a sequence.

Following picture illustrates the routing of a HttpRequest:

4. Building

It uses Scala 2.11, Spark 2.1 and Akka-Http

mvn clean install

5. Execution

We can start our application as stand-alone jar like this:

mvn exec:java

5.1 cmd-line-args

Optionally, you can provide configuration params like spark-master, akka-port etc from command line. To see the list of configurable params, just type:

mvn exec:java -Dexec.args="--help" 
OR 
mvn exec:java -Dexec.args=“-h"

Help content will look something like this:
This application comes as Spark2.1-REST-Service-Provider using an embedded,
Reactive-Streams-based, fully asynchronous HTTP server (i.e., using akka-http).
So, this application needs config params like AkkaWebPort to bind to, SparkMaster
and SparkAppName

Usage: spark-submit spark-as-service-using-embedded-server.jar [options]
  Options:
  -h, --help
  -m, --master <master_url>                    spark://host:port, mesos://host:port, yarn, or local. Default: local
  -n, --name <name>                            A name of your application. Default: SparkAsRestService
  -p, --akkaHttpPort <portnumber>              Port where akka-http is binded. Default: 8001

5.2 Tweak Default cmd-line args

There are 2 ways to change the default param values:

Update src/main/resources/application.conf file directly. Build and then Run
mvn exec:java -Dexec.args="--master <master> --name <spark-app-name> --akkaHttpPort <port-to-which-akka-should-listen-to>"

6. References

Akka

spoddutur / spark-as-service-using-embedded-server

readme