Ethereum ETL Airflow

Read this article: https://cloud.google.com/blog/products/data-analytics/ethereum-bigquery-how-we-built-dataset

Local Development Prerequisites

• direnv
• pyenv

We are using direnv to automatically set up and load the correct python version. We also create a venv in the root folder, that is automatically activated when entering the project folder.

Setting up Airflow DAGs using Google Cloud Composer

Create BigQuery Datasets

• Sign in to BigQuery https://bigquery.cloud.google.com/
• Create new datasets called crypto_ethereum, crypto_ethereum_raw, crypto_ethereum_temp

Create Google Cloud Storage bucket

• Create a new Google Storage bucket to store exported files https://console.cloud.google.com/storage/browser

Create Google Cloud Composer (version 2) environment

Create a new Cloud Composer environment:

export ENVIRONMENT_NAME=ethereum-etl-0

AIRFLOW_CONFIGS_ARR=(
    "celery-worker_concurrency=8"
    "scheduler-dag_dir_list_interval=300"
    "scheduler-min_file_process_interval=120"
)
export AIRFLOW_CONFIGS=$(IFS=, ; echo "${AIRFLOW_CONFIGS_ARR[*]}")

gcloud composer environments create \
    $ENVIRONMENT_NAME \
    --location=us-central1 \
    --image-version=composer-2.1.14-airflow-2.5.1 \
    --environment-size=medium \
    --scheduler-cpu=2 \
    --scheduler-memory=13 \
    --scheduler-storage=1 \
    --scheduler-count=1 \
    --web-server-cpu=1 \
    --web-server-memory=2 \
    --web-server-storage=512MB \
    --worker-cpu=2 \
    --worker-memory=13 \
    --worker-storage=10 \
    --min-workers=1 \
    --max-workers=8 \
    --airflow-configs=$AIRFLOW_CONFIGS

gcloud composer environments update \
    $ENVIRONMENT_NAME \
    --location=us-central1 \
    --update-pypi-packages-from-file=requirements_airflow.txt

Create variables in Airflow (Admin > Variables in the UI):

Variable	Description
ethereum_output_bucket	GCS bucket to store exported files
ethereum_provider_uris	Comma separated URIs of Ethereum nodes
ethereum_destination_dataset_project_id	Project ID of BigQuery datasets
notification_emails	email for notifications

Check other variables in dags/ethereumetl_airflow/variables.py.

Updating package requirements

Suggested package requirements for Composer are stored in requirements_airflow.txt.

You can update the Composer environment using the following script:

ENVIRONMENT_NAME="ethereum-etl-0"
LOCAL_REQUIREMENTS_PATH="$(mktemp)"

# grep pattern removes comments and whitespace:
cat "./requirements_airflow.txt" | grep -o '^[^#| ]*' > "$LOCAL_REQUIREMENTS_PATH"

gcloud composer environments update \
  "$ENVIRONMENT_NAME" \
  --location="us-central1" \
  --update-pypi-packages-from-file="$LOCAL_REQUIREMENTS_PATH"

Note: Composer can be very pedantic about conflicts in additional packages. You may have to fix dependency conflicts where you had no issues testing locally (when updating dependencies, Composer does something "cleverer" than just pip install -r requirements.txt). This is why eth-hash is currently pinned in requirements_airflow.txt. Typically we have found that pinning eth-hash and/or eth-rlp may make things work, though Your Mileage May Vary.

See this issue for further ideas on how to unblock problems you may encounter.

Upload DAGs

> ./upload_dags.sh <airflow_bucket>

Running Tests

pip install \
    -r requirements_test.txt \
    -r requirements_local.txt \
    -r requirements_airflow.txt
pytest -vv -s

Running locally

A docker compose definition has been provided to easily spin up a local Airflow instance.

To build the required image:

docker compose build

To start Airflow:

docker compose up airflow

The instance requires the CLOUDSDK_CORE_PROJECT environment variable to be set in most cases. Airflow Variables can be defined in variables.json.

Creating Table Definition Files for Parsing Events and Function Calls

Read this article: https://medium.com/@medvedev1088/query-ens-and-0x-events-with-sql-in-google-bigquery-4d197206e644

Debugging Table Defenition Files

A utility script for debugging and verifying contract parsing in Ethereum data processing pipelines is available. You can simply run

python3 generate_parse_sql.py <path_to_table_definition_file> <date>

This will output some example SQL that can be used to debug if the generated json files from the contract parser are correct.

NOTE: certain files may not have the contract_address field specified as a valid address ERC20Pool_event_TransferLP but use a select statement on another table instead. For these you can simply pass the contract address yourself like below:

python3 generate_parse_sql.py <path_to_table_definition_file> <date> --contract_address <contract_address>

More Information

You can follow the instructions here for Polygon DAGs https://github.com/blockchain-etl/polygon-etl. The architecture there is very similar to Ethereum so in most case substituting polygon for ethereum will work. Contributions to this README file for porting documentation from Polygon to Ethereum are welcome.