Skip to content

AWS S3 Storage Backend

delta-rs offers native support for using AWS S3 as an object storage backend.

You don’t need to install any extra dependencies to read/write Delta tables to S3 with engines that use delta-rs. You do need to configure your AWS access credentials correctly.

Note for boto3 users

Many Python engines use boto3 to connect to AWS. This library supports reading credentials automatically from your local .aws/config or .aws/creds file.

For example, if you’re running locally with the proper credentials in your local .aws/config or .aws/creds file then you can write a Parquet file to S3 like this with pandas:

    import pandas as pd
    df = pd.DataFrame({'x': [1, 2, 3]})
    df.to_parquet("s3://avriiil/parquet-test-pandas")

The delta-rs writer does not use boto3 and therefore does not support taking credentials from your .aws/config or .aws/creds file. If you’re used to working with writers from Python engines like Polars, pandas or Dask, this may mean a small change to your workflow.

Passing AWS Credentials

You can pass your AWS credentials explicitly by using:

  • the storage_optionskwarg
  • Environment variables
  • EC2 metadata if using EC2 instances
  • AWS Profiles

Example

Let's work through an example with Polars. The same logic applies to other Python engines like Pandas, Daft, Dask, etc.

Follow the steps below to use Delta Lake on S3 with Polars:

  1. Install Polars and deltalake. For example, using:

pip install polars deltalake

  1. Create a dataframe with some toy data.

df = pl.DataFrame({'x': [1, 2, 3]})

  1. Set your storage_options correctly.
storage_options = {
    "AWS_REGION":<region_name>,
    'AWS_ACCESS_KEY_ID': <key_id>,
    'AWS_SECRET_ACCESS_KEY': <access_key>,
    'AWS_S3_LOCKING_PROVIDER': 'dynamodb',
    'DELTA_DYNAMO_TABLE_NAME': 'delta_log',
}
  1. Write data to Delta table using the storage_options kwarg.
df.write_delta(
    "s3://bucket/delta_table",
    storage_options=storage_options,
)

Delta Lake on AWS S3: Safe Concurrent Writes

You need a locking provider to ensure safe concurrent writes when writing Delta tables to AWS S3. This is because AWS S3 does not guarantee mutual exclusion.

A locking provider guarantees that only one writer is able to create the same file. This prevents corrupted or conflicting data.

delta-rs uses DynamoDB to guarantee safe concurrent writes.

Run the code below in your terminal to create a DynamoDB table that will act as your locking provider.

    aws dynamodb create-table \
    --table-name delta_log \
    --attribute-definitions AttributeName=tablePath,AttributeType=S AttributeName=fileName,AttributeType=S \
    --key-schema AttributeName=tablePath,KeyType=HASH AttributeName=fileName,KeyType=RANGE \
    --provisioned-throughput ReadCapacityUnits=5,WriteCapacityUnits=5

If for some reason you don't want to use DynamoDB as your locking mechanism you can choose to set the AWS_S3_ALLOW_UNSAFE_RENAME variable to true in order to enable S3 unsafe writes.

Read more in the Usage section.

Delta Lake on AWS S3: Required permissions

You need to have permissions to get, put and delete objects in the S3 bucket you're storing your data in. Please note that you must be allowed to delete objects even if you're just appending to the Delta Lake, because there are temporary files into the log folder that are deleted after usage.

In AWS S3, you will need the following permissions:

  • s3:GetObject
  • s3:PutObject
  • s3:DeleteObject

In DynamoDB, you will need the following permissions:

  • dynamodb:GetItem
  • dynamodb:Query
  • dynamodb:PutItem
  • dynamodb:UpdateItem