spark-submit: Passing java properties file gives FileNotFoundException

Question

I have a spark job that I am running in yarn cluster mode. I need to pass my application specific java properties file. I am doing this java FileInputStream .
But I am facing FileNotFoundException for /home/aiman/config/my-file.properties

My java properties file is: /home/aiman/config/my-file.properties

mongo_server=my.mongo.server
mongo_port=1530
mongo_user=mongoadmin
mongo_password=readONLYpass
mongo_db=testdb

My spark-submit looks like:

spark-submit --master yarn --deploy-mode cluster --class mongo.MongoRead --jars /home/aiman/ojdbc-7.jar /home/aiman/app/jars/MongoRead-0.1-jar-with-dependencies.jar /home/aiman/config/my-file.properties

Code snippet:

public static void main(final String[] args) throws Exception
    {       
        if(args.length<1){
            System.out.println("Please provide properties file path");
            System.exit(1);
        }

        System.out.println("Mongo Import Begins...");

        Properties prop = new Properties();
        InputStream in = null;
        try{
            in = new FileInputStream(args[0]);
            prop.load(in);
        }
        catch(Exception e){
            e.printStackTrace();
            System.exit(1);
        }

        /*Reading the properties*/
        String mongoServer = prop.getProperty("mongo_server");
        String mongoPort = prop.getProperty("mongo_port");
        String mongoUser = prop.getProperty("mongo_user");
        ....
        ...

        String mongoAddress = "mongodb://"+mongoUser+":"+mongoPassword+"@"+mongoServer+":"+mongoPort+"/"+mongoDb+"."+tableNm;

        SparkSession spark = SparkSession.builder()
            .appName("MongoSparkRecordReader")
            .config("spark.mongodb.input.uri", mongoAddress)
            .config("spark.mongodb.output.uri", mongoAddress)
            .getOrCreate();

        JavaSparkContext jsc = new JavaSparkContext(spark.sparkContext());
        ...
        ..

}

Please suggest what is going wrong. Am I reading the properties file in a wrong manner, that it isn't finding the file?

Answer 1

Running this in --deploy-mode client resolves the issue. The Driver is getting launched where the properties file is present. So file is getting read and not getting any Exception.

Answer 2

You should pass the local file via the "--file /home/aiman/config/my-file.properties" spark-submit's argument, which will distribute the file to the containers launched by YARN that will find the "my-file.properties" in their JVM's classpath. You should then be able to read it as a JVM's local resource.