While you need Hadoop libraries you do not have to have Hadoop installed to run locally and write to S3. I just happened to try this out with writing a Parquet output based on Avro schema and generated SpecificRecord to S3. I am running a version of the following code locally through SBT and Intellij Idea. Needed parts:
1) Have the following file specifying the needed Hadoop properties (Note: defining the AWS access key/secret key is not recommended. Better to run on an EC2 instance that has proper IAM role to read/write to your S3 bucket. But needed for local for testing)
<configuration>
<property>
<name>fs.s3.impl</name>
<value>org.apache.hadoop.fs.s3a.S3AFileSystem</value>
</property>
<!-- Comma separated list of local directories used to buffer
large results prior to transmitting them to S3. -->
<property>
<name>fs.s3a.buffer.dir</name>
<value>/tmp</value>
</property>
<!-- set your AWS ID using key defined in org.apache.hadoop.fs.s3a.Constants -->
<property>
<name>fs.s3a.access.key</name>
<value>YOUR_ACCESS_KEY</value>
</property>
<!-- set your AWS access key -->
<property>
<name>fs.s3a.secret.key</name>
<value>YOUR_SECRET_KEY</value>
</property>
</configuration>
2) Imports:
import com.uebercomputing.eventrecord.EventOnlyRecord
import org.apache.flink.api.scala.hadoop.mapreduce.HadoopOutputFormat
import org.apache.flink.api.scala.{ExecutionEnvironment, _}
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat
import org.apache.hadoop.conf.{Configuration => HadoopConfiguration}
import org.apache.hadoop.fs.Path
import org.apache.hadoop.mapreduce.Job
import org.apache.parquet.avro.AvroParquetOutputFormat
3) Flink code uses HadoopOutputFormat with above configuration:
val events: DataSet[(Void, EventOnlyRecord)] = ...
val hadoopConfig = getHadoopConfiguration(hadoopConfigFile)
val outputFormat = new AvroParquetOutputFormat[EventOnlyRecord]
val outputJob = Job.getInstance
//Note: AvroParquetOutputFormat extends FileOutputFormat[Void,T]
//so key is Void, value of type T - EventOnlyRecord in this case
val hadoopOutputFormat = new HadoopOutputFormat[Void, EventOnlyRecord](
outputFormat,
outputJob
)
val outputConfig = outputJob.getConfiguration
outputConfig.addResource(hadoopConfig)
val outputPath = new Path("s3://<bucket>/<dir-prefix>")
FileOutputFormat.setOutputPath(outputJob, outputPath)
AvroParquetOutputFormat.setSchema(outputJob, EventOnlyRecord.getClassSchema)
events.output(hadoopOutputFormat)
env.execute
...
def getHadoopConfiguration(hadoodConfigPath: String): HadoopConfiguration = {
val hadoopConfig = new HadoopConfiguration()
hadoopConfig.addResource(new Path(hadoodConfigPath))
hadoopConfig
}
4) Build dependencies and versions used:
val awsSdkVersion = "1.7.4"
val hadoopVersion = "2.7.3"
val flinkVersion = "1.1.4"
val flinkDependencies = Seq(
("org.apache.flink" %% "flink-scala" % flinkVersion),
("org.apache.flink" %% "flink-hadoop-compatibility" % flinkVersion)
)
val providedFlinkDependencies = flinkDependencies.map(_ % "provided")
val serializationDependencies = Seq(
("org.apache.avro" % "avro" % "1.7.7"),
("org.apache.avro" % "avro-mapred" % "1.7.7").classifier("hadoop2"),
("org.apache.parquet" % "parquet-avro" % "1.8.1")
)
val s3Dependencies = Seq(
("com.amazonaws" % "aws-java-sdk" % awsSdkVersion),
("org.apache.hadoop" % "hadoop-aws" % hadoopVersion)
)
Edit for using writeAsText to S3:
1) Create a Hadoop configuration directory (will reference this as hadoop-conf-dir) with a file core-site.xml in it.
For example:
mkdir /home/<user>/hadoop-config
cd /home/<user>/hadoop-config
vi core-site.xml
#content of core-site.xml
<configuration>
<property>
<name>fs.s3.impl</name>
<value>org.apache.hadoop.fs.s3a.S3AFileSystem</value>
</property>
<!-- Comma separated list of local directories used to buffer
large results prior to transmitting them to S3. -->
<property>
<name>fs.s3a.buffer.dir</name>
<value>/tmp</value>
</property>
<!-- set your AWS ID using key defined in org.apache.hadoop.fs.s3a.Constants -->
<property>
<name>fs.s3a.access.key</name>
<value>YOUR_ACCESS_KEY</value>
</property>
<!-- set your AWS access key -->
<property>
<name>fs.s3a.secret.key</name>
<value>YOUR_SECRET_KEY</value>
</property>
</configuration>
2) Create a directory (will reference this as flink-conf-dir) with a file flink-conf.yaml in it.
For example:
mkdir /home/<user>/flink-config
cd /home/<user>/flink-config
vi flink-conf.yaml
//content of flink-conf.yaml - continuing earlier example
fs.hdfs.hadoopconf: /home/<user>/hadoop-config
3) Edit your IntelliJ Run configuration used to run your S3 Flink job - Run - Edit configurations - and add the following environment variable:
FLINK_CONF_DIR and set it to your flink-conf-dir
Continuing the example above:
FLINK_CONF_DIR=/home/<user>/flink-config
4) Run the code with that environment variable set:
events.writeAsText("s3://<bucket>/<prefix-dir>")
env.execute
