2

I am new to AWS and i have learnt and developed code in spark -scala .

My application basically merge two files in spark and created final output.

I read both files (MAIN files and INCR files )in spark from S3 bucket .

All are working fine and i am getting correct output also . But i dont know how to automate whole process to put in production .

Here are the steps the i am doing in order to get output .

STEP 1: Loading Main files (5K text files ).I am reading files from FTP in EC2 and then uploading in the S3 bucket .

STEP 2: Loading INCR (incremental files) same way as i am loading MAIN files .

STEP 3: Creating EMR cluster manually from UI.

STEP 4: Opening Zeppelin note book and copy paste spark-scala script and run .

STEP 5: Again creating EC2 instance to read S3 bucket and send output files from S3 to FTP client .

I am using EC2 because in my case i dont have direct connect from S3 to FTP .We are in a process to get DIRECT CONNECT from AWS .

Please assist me how can i automate in a best way .

1

You have 2 options. If you also want to shutdown and start cluster then option 1, the AWS Data Pipeline is better:

1. AWS Data Pipeline DeveloperGuide: what-is-datapipeline

AWS Data Pipeline

2. Oozie Workflow

You could use Oozie. Here is an example how to automate Spark jobs in AWS with Oozie:

https://aws.amazon.com/de/blogs/big-data/use-apache-oozie-workflows-to-automate-apache-spark-jobs-and-more-on-amazon-emr/

If you have already the FTP connection for the data transfer on your EC2 instance then you can trigger a shell script from Oozie. Here is an example running-shell-script-with-oozie

  • But oozie will not help me automate my creating cluster and loading s3 files ?It will help me just to automate my job .. – SUDARSHAN Jan 16 '18 at 8:21
  • Ah, it was not clear that you want to initialise also the cluster. I though its running all the time. In this case, you could also use the AWS Data Pipeline in order to automate the movement and transformation of data. – Rene B. Jan 16 '18 at 9:04
  • so we can create cluster also from AWS data pipeline ? Also my spark job will run once in a week and also i have upload files once in a week only . – SUDARSHAN Jan 16 '18 at 10:54
  • yes, you can launch a cluster using the command line as described here: docs.aws.amazon.com/datapipeline/latest/DeveloperGuide/… – Rene B. Jan 16 '18 at 11:45
  • Just one more question ..I am using Data pipleine ...But curious to know the difference between Data pipeline and Cloud formation template ? – SUDARSHAN Apr 5 '18 at 15:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.