Datadog connection

Submit metrics, events, and query monitors from a task over a managed Leoflow Connection. DatadogHook authenticates with an API key + app key against the Datadog site (US, EU, …).

Declare the provider

# leoflow.yaml
dag_id: datadog_metric
connectors:
  - datadog

URI shape

datadog:?__extra__=<json>

There is no host and no password — every field (API host, API key, app key, source type) lives in Extra under __extra__. This is the extra-only shape.

Fields the UI asks for

Extra example:

{
  "api_host": "https://api.datadoghq.eu",
  "api_key": "...",
  "app_key": "...",
  "source_type_name": "airflow"
}

Example DAG

# dag.py
from airflow.sdk import DAG, task


@task
def send():
    from airflow.providers.datadog.hooks.datadog import DatadogHook

    hook = DatadogHook(datadog_conn_id="datadog_default")
    hook.send_metric(
        metric_name="leoflow.dag.runs",
        datapoint=1,
        tags=["dag:datadog_metric"],
    )


with DAG("datadog_metric", schedule=None, catchup=False, tags=["example"]):
    send()

# leoflow.yaml
schema_version: "1.0"
dag_id: datadog_metric
python_version: "3.12"
connectors:
  - datadog
connections:
  - datadog_default

Security notes

• API key vs app key: the API key authorizes submission; the app key authorizes reads/queries. Scope the app key narrowly.
• Pick the right site: api_host must match your Datadog org’s region (datadoghq.com, datadoghq.eu, …) or calls 403.
• Never log AIRFLOW_CONN_DATADOG_DEFAULT; the keys ride in __extra__.

Related

• ADR 0019 — secret encryption at rest.
• ADR 0021 — agent secret delivery (AIRFLOW_CONN_<CONN_ID>).
• TestDatadogConnectionURIShapeIntegration — chain-of-custody delivery test.