pig-dynamodb

This is a fork of mortar's code to support Tuples to String Sets.

An Apache Pig storage function for DynamoDB, by Mortar Data.

Features

• Automatic write throttling: Writes are automatically throttled and balanced across Hadoop tasks to match your DynamoDB table's write throughput.
• Automatic schema conversion: Whatever you pass to the storage function gets converted and stored.
• Retries, retries, retries: Automatic retries and logging when DynamoDB write limits are hit.

Building

Use maven to build the package:

mvn clean package

...and then upload the JAR in the target directory to S3, where you can register it from your pigscripts.

Storing Data to DynamoDB

For full instructions, see Storing Data into DynamoDB on the Mortar help site.

Here's a quick sample of how it works:

-- register the jar
REGISTER 's3://my-bucket/my-folder/pig-dynamodb-0.2-SNAPSHOT.jar'

-- Percentage of the table's write throughput to use
SET dynamodb.throughput.write.percent 1.0;

-- Disable Pig's MultiQuery optimization when using DynamoDBStorage
SET opt.multiquery false;

-- Load up some input data
input_data = LOAD '$INPUT_PATH'
    USING PigStorage()
    AS (...);

-- Select exactly the fields you want to store to dynamodb.
-- MUST include your DynamoDB table's primary key.
exact_fields_to_store = FOREACH input_data
    GENERATE my_field AS name_in_dynamodb_1,
             my_field_2 AS name_in_dynamodb_2,
             my_field_3 AS name_in_dynamodb_3;

-- Store the data to DynamoDB
STORE exact_fields_to_store
 INTO 's3://ignored-output-bucket/ignored'
USING com.mortardata.pig.storage.DynamoDBStorage(
	'my_dynamo_db_table', 
	'my_aws_access_key_id', 
	'my_aws_secret_access_key');

Options

• dynamodb.throughput.write.percent: Percentage of total write throughput on the table to use for this STORE. Range: [0.1 - 1.5]. Default: 0.5.
• dynamodb.retry.max_per_batch_write: Number of times to retry a failed write (due to throttling) before throwing an Exception. Default: 100.
• dynamodb.retry.wait.max: Maximum time, in milliseconds, to wait (during exponential backoff) between retries on a failed write. Default: 120000.