pyspark - How to set up a local development environment for Scala Spark ETL to run in AWS Glue?

Question

I'd like to be able to write Scala in my local IDE and then deploy it to AWS Glue as part of a build process. But I'm having trouble finding the libraries required to build the GlueApp skeleton generated by AWS.

The aws-java-sdk-glue doesn't contain the classes imported, and I can't find those libraries anywhere else. Though they must exist somewhere, but perhaps they are just a Java/Scala port of this library: aws-glue-libs

The template scala code from AWS:

import com.amazonaws.services.glue.GlueContext
import com.amazonaws.services.glue.MappingSpec
import com.amazonaws.services.glue.errors.CallSite
import com.amazonaws.services.glue.util.GlueArgParser
import com.amazonaws.services.glue.util.Job
import com.amazonaws.services.glue.util.JsonOptions
import org.apache.spark.SparkContext
import scala.collection.JavaConverters._

object GlueApp {
  def main(sysArgs: Array[String]) {
    val spark: SparkContext = new SparkContext()
    val glueContext: GlueContext = new GlueContext(spark)
    // @params: [JOB_NAME]
    val args = GlueArgParser.getResolvedOptions(sysArgs, Seq("JOB_NAME").toArray)
    Job.init(args("JOB_NAME"), glueContext, args.asJava)
    // @type: DataSource
    // @args: [database = "raw-tickers-oregon", table_name = "spark_delivery_2_1", transformation_ctx = "datasource0"]
    // @return: datasource0
    // @inputs: []
    val datasource0 = glueContext.getCatalogSource(database = "raw-tickers-oregon", tableName = "spark_delivery_2_1", redshiftTmpDir = "", transformationContext = "datasource0").getDynamicFrame()
    // @type: ApplyMapping
    // @args: [mapping = [("exchangeid", "int", "exchangeid", "int"), ("data", "struct", "data", "struct")], transformation_ctx = "applymapping1"]
    // @return: applymapping1
    // @inputs: [frame = datasource0]
    val applymapping1 = datasource0.applyMapping(mappings = Seq(("exchangeid", "int", "exchangeid", "int"), ("data", "struct", "data", "struct")), caseSensitive = false, transformationContext = "applymapping1")
    // @type: DataSink
    // @args: [connection_type = "s3", connection_options = {"path": "s3://spark-ticker-oregon/target", "compression": "gzip"}, format = "json", transformation_ctx = "datasink2"]
    // @return: datasink2
    // @inputs: [frame = applymapping1]
    val datasink2 = glueContext.getSinkWithFormat(connectionType = "s3", options = JsonOptions("""{"path": "s3://spark-ticker-oregon/target", "compression": "gzip"}"""), transformationContext = "datasink2", format = "json").writeDynamicFrame(applymapping1)
    Job.commit()
  }
}

And the build.sbt I have started putting together for a local build:

name := "aws-glue-scala"

version := "0.1"

scalaVersion := "2.11.12"

updateOptions := updateOptions.value.withCachedResolution(true)

libraryDependencies += "org.apache.spark" %% "spark-core" % "2.2.1"

The documentation for AWS Glue Scala API seems to outline similar functionality as is available in the AWS Glue Python library. So perhaps all that is required is to download and build the PySpark AWS Glue library and add it on the classpath? Perhaps possible since the Glue python library uses Py4J.

Answer 1

@Frederic gave a very helpful hint to get the dependency from s3://aws-glue-jes-prod-us-east-1-assets/etl/jars/glue-assembly.jar.

Unfortunately that version of glue-assembly.jar is already outdated and brings spark in versoin 2.1. It's fine if you're using backward compatible features, but if you rely on latest spark version (and possibly latest glue features) you can get the appropriate jar from a Glue dev-endpoint under /usr/share/aws/glue/etl/jars/glue-assembly.jar.

Provided you have a dev-endpoint named my-dev-endpoint you can copy the current jar from it:

export DEV_ENDPOINT_HOST=`aws glue get-dev-endpoint --endpoint-name my-dev-endpoint --query 'DevEndpoint.PublicAddress' --output text`

scp -i dev-endpoint-private-key 
glue@$DEV_ENDPOINT_HOST:/usr/share/aws/glue/etl/jars/glue-assembly.jar .