Skip to main content
Version: 1.1.0 (latest)

AWS Athena / Glue Catalog

The Athena destination stores data as Parquet files in S3 buckets and creates external tables in AWS Athena. You can then query those tables with Athena SQL commands, which will scan the entire folder of Parquet files and return the results. This destination works very similarly to other SQL-based destinations, with the exception that the merge write disposition is not supported at this time. The dlt metadata will be stored in the same bucket as the Parquet files, but as iceberg tables. Athena also supports writing individual data tables as Iceberg tables, so they may be manipulated later. A common use case would be to strip GDPR data from them.

Install dlt with Athenaโ€‹

To install the dlt library with Athena dependencies:

pip install "dlt[athena]"

Setup guideโ€‹

1. Initialize the dlt projectโ€‹

Let's start by initializing a new dlt project as follows:

dlt init chess athena

๐Ÿ’ก This command will initialize your pipeline with chess as the source and AWS Athena as the destination using the filesystem staging destination.

2. Setup bucket storage and Athena credentialsโ€‹

First, install dependencies by running:

pip install -r requirements.txt

or with pip install "dlt[athena]", which will install s3fs, pyarrow, pyathena, and botocore packages.

caution

You may also install the dependencies independently. Try

pip install dlt
pip install s3fs
pip install pyarrow
pip install pyathena

so pip does not fail on backtracking.

To edit the dlt credentials file with your secret info, open .dlt/secrets.toml. You will need to provide a bucket_url, which holds the uploaded parquet files, a query_result_bucket, which Athena uses to write query results to, and credentials that have write and read access to these two buckets as well as the full Athena access AWS role.

The toml file looks like this:

[destination.filesystem]
bucket_url = "s3://[your_bucket_name]" # replace with your bucket name,

[destination.filesystem.credentials]
aws_access_key_id = "please set me up!" # copy the access key here
aws_secret_access_key = "please set me up!" # copy the secret access key here

[destination.athena]
query_result_bucket="s3://[results_bucket_name]" # replace with your query results bucket name

[destination.athena.credentials]
aws_access_key_id="please set me up!" # same as credentials for filesystem
aws_secret_access_key="please set me up!" # same as credentials for filesystem
region_name="please set me up!" # set your AWS region, for example "eu-central-1" for Frankfurt

If you have your credentials stored in ~/.aws/credentials, just remove the [destination.filesystem.credentials] and [destination.athena.credentials] sections above and dlt will fall back to your default profile in local credentials. If you want to switch the profile, pass the profile name as follows (here: dlt-ci-user):

[destination.filesystem.credentials]
profile_name="dlt-ci-user"

[destination.athena.credentials]
profile_name="dlt-ci-user"

Additional destination configurationโ€‹

You can provide an Athena workgroup like so:

[destination.athena]
athena_work_group="my_workgroup"

Write dispositionโ€‹

The athena destination handles the write dispositions as follows:

  • append - files belonging to such tables are added to the dataset folder.
  • replace - all files that belong to such tables are deleted from the dataset folder, and then the current set of files is added.
  • merge - falls back to append (unless you're using iceberg tables).

Data loadingโ€‹

Data loading occurs by storing parquet files in an S3 bucket and defining a schema on Athena. If you query data via SQL queries on Athena, the returned data is read by scanning your bucket and reading all relevant parquet files in there.

dlt internal tables are saved as Iceberg tables.

Data typesโ€‹

Athena tables store timestamps with millisecond precision, and with that precision, we generate parquet files. Keep in mind that Iceberg tables have microsecond precision.

Athena does not support JSON fields, so JSON is stored as a string.

โ—Athena does not support TIME columns in parquet files. dlt will fail such jobs permanently. Convert datetime.time objects to str or datetime.datetime to load them.

Table and column identifiersโ€‹

Athena uses case-insensitive identifiers and will lowercase all the identifiers that are stored in the INFORMATION SCHEMA. Do not use case-sensitive naming conventions. Letter casing will be removed anyway, and you risk generating identifier collisions, which are detected by dlt and will fail the load process.

Under the hood, Athena uses different SQL engines for DDL (catalog) and DML/Queries:

  • DDL uses HIVE escaping with ``````
  • Other queries use PRESTO and regular SQL escaping.

Staging supportโ€‹

Using a staging destination is mandatory when using the Athena destination. If you do not set staging to filesystem, dlt will automatically do this for you.

If you decide to change the filename layout from the default value, keep the following in mind so that Athena can reliably build your tables:

  • You need to provide the {table_name} placeholder, and this placeholder needs to be followed by a forward slash.
  • You need to provide the {file_id} placeholder, and it needs to be somewhere after the {table_name} placeholder.
  • {table_name} must be the first placeholder in the layout.

Additional destination optionsโ€‹

Iceberg data tablesโ€‹

You can save your tables as Iceberg tables to Athena. This will enable you, for example, to delete data from them later if you need to. To switch a resource to the Iceberg table format, supply the table_format argument like this:

@dlt.resource(table_format="iceberg")
def data() -> Iterable[TDataItem]:
...

For every table created as an Iceberg table, the Athena destination will create a regular Athena table in the staging dataset of both the filesystem and the Athena glue catalog, and then copy all data into the final Iceberg table that lives with the non-Iceberg tables in the same dataset on both the filesystem and the glue catalog. Switching from Iceberg to regular table or vice versa is not supported.

merge supportโ€‹

The merge write disposition is supported for Athena when using Iceberg tables.

Note that:

  1. There is a risk of tables ending up in an inconsistent state in case a pipeline run fails mid-flight because Athena doesn't support transactions, and dlt uses multiple DELETE/UPDATE/INSERT statements to implement merge.
  2. dlt creates additional helper tables called insert_<table name> and delete_<table name> in the staging schema to work around Athena's lack of temporary tables.

dbt supportโ€‹

Athena is supported via dbt-athena-community. Credentials are passed into aws_access_key_id and aws_secret_access_key of the generated dbt profile. Iceberg tables are supported, but you need to make sure that you materialize your models as Iceberg tables if your source table is Iceberg. We encountered problems with materializing date-time columns due to different precision on Iceberg (nanosecond) and regular Athena tables (millisecond). The Athena adapter requires that you set up region_name in the Athena configuration below. You can also set up the table catalog name to change the default: awsdatacatalog

[destination.athena]
aws_data_catalog="awsdatacatalog"

Syncing of dlt stateโ€‹

  • This destination fully supports dlt state sync.. The state is saved in Athena Iceberg tables in your S3 bucket.

Supported file formatsโ€‹

You can choose the following file formats:

Athena adapterโ€‹

You can use the athena_adapter to add partitioning to Athena tables. This is currently only supported for Iceberg tables.

Iceberg tables support a few transformation functions for partitioning. Info on all supported functions in the AWS documentation.

Use the athena_partition helper to generate the partitioning hints for these functions:

  • athena_partition.year(column_name: str): Partition by year of date/datetime column.
  • athena_partition.month(column_name: str): Partition by month of date/datetime column.
  • athena_partition.day(column_name: str): Partition by day of date/datetime column.
  • athena_partition.hour(column_name: str): Partition by hour of date/datetime column.
  • athena_partition.bucket(n: int, column_name: str): Partition by hashed value to n buckets
  • athena_partition.truncate(length: int, column_name: str): Partition by truncated value to length (or width for numbers)

Here is an example of how to use the adapter to partition a table:

from datetime import date

import dlt
from dlt.destinations.adapters import athena_partition, athena_adapter

data_items = [
(1, "A", date(2021, 1, 1)),
(2, "A", date(2021, 1, 2)),
(3, "A", date(2021, 1, 3)),
(4, "A", date(2021, 2, 1)),
(5, "A", date(2021, 2, 2)),
(6, "B", date(2021, 1, 1)),
(7, "B", date(2021, 1, 2)),
(8, "B", date(2021, 1, 3)),
(9, "B", date(2021, 2, 1)),
(10, "B", date(2021, 3, 2)),
]

@dlt.resource(table_format="iceberg")
def partitioned_data():
yield [{"id": i, "category": c, "created_at": d} for i, c, d in data_items]

# Add partitioning hints to the table
athena_adapter(
partitioned_table,
partition=[
# Partition per category and month
"category",
athena_partition.month("created_at"),
],
)


pipeline = dlt.pipeline("athena_example")
pipeline.run(partitioned_data)

Additional Setup guidesโ€‹

This demo works on codespaces. Codespaces is a development environment available for free to anyone with a Github account. You'll be asked to fork the demo repository and from there the README guides you with further steps.
The demo uses the Continue VSCode extension.

Off to codespaces!

DHelp

Ask a question

Welcome to "Codex Central", your next-gen help center, driven by OpenAI's GPT-4 model. It's more than just a forum or a FAQ hub โ€“ it's a dynamic knowledge base where coders can find AI-assisted solutions to their pressing problems. With GPT-4's powerful comprehension and predictive abilities, Codex Central provides instantaneous issue resolution, insightful debugging, and personalized guidance. Get your code running smoothly with the unparalleled support at Codex Central - coding help reimagined with AI prowess.