Skip to main content
Version: 1.26.0

Python

Overview​

The Python client is the basis of existing OpenLineage integrations such as Airflow and dbt.

The client enables the creation of lineage metadata events with Python code. The core data structures currently offered by the client are the RunEvent, RunState, Run, Job, Dataset, and Transport classes. These either configure or collect data for the emission of lineage events.

You can use the client to create your own custom integrations.

Installation​

Download the package using pip with

pip install openlineage-python

To install the package from source, use

python -m pip install .

Configuration​

We recommend configuring the client with an openlineage.yml file that contains all the details of how to connect to your OpenLineage backend.

You can make this file available to the client in three ways (the list also presents precedence of the configuration):

  1. Set an OPENLINEAGE_CONFIG environment variable to a file path: OPENLINEAGE_CONFIG=path/to/openlineage.yml.
  2. Place an openlineage.yml file in the current working directory (the absolute path of the directory where your script or process is currently running).
  3. Place an openlineage.yml file under .openlineage/ in the user's home directory (~/.openlineage/openlineage.yml).

In openlineage.yml, use a standard Transport interface to specify the transport type (http, console, kafka, file, or custom) and authorization parameters. See the example config file for each transport type.

If there is no config file found, the OpenLineage client looks at environment variables for HTTP transport.

At the end, if no configuration is found, ConsoleTransport is used, the events are printed in the console.

Environment Variables​

The following environment variables are available to use:

NameDescriptionExampleSince
OPENLINEAGE_CONFIGThe path to the YAML configuration filepath/to/openlineage.yml
OPENLINEAGE_CLIENT_LOGGINGLogging level of OpenLineage client and its child modulesDEBUG
OPENLINEAGE_DISABLEDWhen true, OpenLineage will not emit events (default: false)false0.9.0
OPENLINEAGE_URLThe URL to send lineage events to (also see OPENLINEAGE_ENDPOINT)https://myapp.com
OPENLINEAGE_ENDPOINTEndpoint to which events are sent (default: api/v1/lineage)api/v2/events
OPENLINEAGE_API_KEYToken included in the Authentication HTTP header as the Bearersecret_token_123

If you are using Airflow integration, there are additional environment variables available.

Dynamic configuration with environment variables​

You can also configure the client with dynamic environment variables. Environment variables that configure the OpenLineage client follow a specific pattern. All variables that affect the client configuration start with the prefix OPENLINEAGE__, followed by nested keys separated by double underscores (__).

Key Features​
  1. Prefix Requirement: All environment variables must begin with OPENLINEAGE__.
  2. Sections Separation: Configuration sections are separated using double underscores __ to form the hierarchy.
  3. Lowercase Conversion: Environment variable values are automatically converted to lowercase.
  4. JSON String Support: You can pass a JSON string at any level of the configuration hierarchy, which will be merged into the final configuration structure.
  5. Hyphen Restriction: Since environment variable names cannot contain - (hyphen), if a name strictly requires a hyphen, use a JSON string as the value of the environment variable.
  6. Precedence Rules:
  • Top-level keys have precedence and will not be overwritten by more nested entries.
  • For example, OPENLINEAGE__TRANSPORT='{..}' will not have its keys overwritten by OPENLINEAGE__TRANSPORT__AUTH__KEY='key'.
Dynamic Alias for Transport Variables​

To facilitate easier management of environment variables, aliases are dynamically created for certain variables like OPENLINEAGE_URL. If OPENLINEAGE_URL is set, it automatically translates into specific transport configurations that can be used with Composite transport with default_http as the name of the HTTP transport.

Alias rules are following:

  • If environment variable OPENLINEAGE_URL="http://example.com" is set, it would insert following environment variables:
OPENLINEAGE__TRANSPORT__TRANSPORTS__DEFAULT_HTTP__TYPE="http"
OPENLINEAGE__TRANSPORT__TRANSPORTS__DEFAULT_HTTP__URL="http://example.com"
  • Similarly if environment variable OPENLINEAGE_API_KEY="random_key" is set, it will be translated to:
OPENLINEAGE__TRANSPORT__TRANSPORTS__DEFAULT_HTTP__AUTH='{"type": "api_key", "apiKey": "random_key"}'

qually with environment variable OPENLINEAGE_ENDPOINT="api/v1/lineage", that translates to:

OPENLINEAGE__TRANSPORT__TRANSPORTS__DEFAULT_HTTP__ENDPOINT="api/v1/lineage"
  • If one does not want to use aliased HTTP transport in Composite Transport, they can set OPENLINEAGE__TRANSPORT__TRANSPORTS__DEFAULT_HTTP to {}.

Examples​

Setting following environment variables:

OPENLINEAGE__TRANSPORT__TYPE=http
OPENLINEAGE__TRANSPORT__URL=http://localhost:5050
OPENLINEAGE__TRANSPORT__ENDPOINT=/api/v1/lineage
OPENLINEAGE__TRANSPORT__AUTH='{"type":"api_key", "apiKey":"random_token"}'
OPENLINEAGE__TRANSPORT__COMPRESSION=gzip

is equivalent to passing following YAML configuration:

transport:
  type: http
  url: http://localhost:5050
  endpoint: api/v1/lineage
  auth:
    type: api_key
    apiKey: random_token
  compression: gzip

HTTP transport configuration with environment variables​

For backwards compatibility, the simplest HTTP transport configuration, with only a subset of its config, can be done with environment variables (all other transport types are only configurable with YAML file). This setup can be done with the following environment variables:

  • OPENLINEAGE_URL (required)
  • OPENLINEAGE_ENDPOINT (optional, default: api/v1/lineage)
  • OPENLINEAGE_API_KEY (optional).

Built-in Transport Types​

HTTP​

Allows sending events to HTTP endpoint, using requests.

Configuration​

  • type - string, must be "http". Required.
  • url - string, base url for HTTP requests. Required.
  • endpoint - string specifying the endpoint to which events are sent, appended to url. Optional, default: api/v1/lineage.
  • timeout - float specifying timeout (in seconds) value used while connecting to server. Optional, default: 5.
  • verify - boolean specifying whether the client should verify TLS certificates from the backend. Optional, default: true.
  • auth - dictionary specifying authentication options. Optional, by default no authorization is used. If set, requires the type property.
    • type - string specifying the "api_key" or the fully qualified class name of your TokenProvider. Required if auth is provided.
    • apiKey - string setting the Authentication HTTP header as the Bearer. Required if type is api_key.
  • compression - string, name of algorithm used by HTTP client to compress request body. Optional, default value null, allowed values: gzip. Added in v1.13.0.
  • custom_headers - dictionary of additional headers to be sent with each request. Optional, default: {}.

Behavior​

Events are serialized to JSON, and then are send as HTTP POST request with Content-Type: application/json.

Examples​

transport:
  type: http
  url: https://backend:5000
  endpoint: api/v1/lineage
  timeout: 5
  verify: false
  auth:
    type: api_key
    apiKey: f048521b-dfe8-47cd-9c65-0cb07d57591e
  compression: gzip

Console​

This straightforward transport emits OpenLineage events directly to the console through a logger. No additional configuration is required.

Configuration​

  • type - string, must be "console". Required.

Behavior​

Events are serialized to JSON. Then each event is logged with INFO level to logger with name openlineage.client.transport.console.

Notes​

Be cautious when using the DEBUG log level, as it might result in double-logging due to the OpenLineageClient also logging.

Examples​

transport:
  type: console

Kafka​

Kafka transport requires confluent-kafka package to be additionally installed. It can be installed also by specifying kafka client extension: pip install openlineage-python[kafka]

Configuration​

  • type - string, must be "kafka". Required.

  • topic - string specifying the topic on what events will be sent. Required.

  • config - a dictionary containing a Kafka producer config as in Kafka producer config. Required.

  • flush - boolean specifying whether Kafka should flush after each event. Optional, default: true.

  • messageKey - string, key for all Kafka messages produced by transport. Optional, default value described below. Added in v1.13.0.

    Default values for messageKey are:

    • run:{parentJob.namespace}/{parentJob.name} - for RunEvent with parent facet
    • run:{job.namespace}/{job.name} - for RunEvent
    • job:{job.namespace}/{job.name} - for JobEvent
    • dataset:{dataset.namespace}/{dataset.name} - for DatasetEvent

Behavior​

  • Events are serialized to JSON, and then dispatched to the Kafka topic.
  • If flush is true, messages will be flushed to the topic after each event being sent.

Notes​

It is recommended to provide messageKey if Job hierarchy is used. It can be any string, but it should be the same for all jobs in hierarchy, like Airflow task -> Spark application -> Spark task runs.

Using with Airflow integration​

There's a caveat for using KafkaTransport with Airflow integration. In this integration, a Kafka producer needs to be created for each OpenLineage event. It happens due to the Airflow execution and plugin model, which requires us to send messages from worker processes. These are created dynamically for each task execution.

Examples​

transport:
  type: kafka
  topic: my_topic
  config:
    bootstrap.servers: localhost:9092,another.host:9092
    acks: all
    retries: 3
  flush: true
  messageKey: some-value

File​

Designed mainly for integration testing, the FileTransport emits OpenLineage events to a given file(s).

Configuration​

  • type - string, must be "file". Required.
  • log_file_path - string specifying the path of the file or file prefix (when append is true). Required.
  • append - boolean, see Behavior section below. Optional, default: false.

Behavior​

  • If the target file is absent, it's created.
  • If append is true, each event will be appended to a single file log_file_path, separated by newlines.
  • If append is false, each event will be written to as separated file with name {log_file_path}-{datetime}.

Examples​

transport:
  type: file
  log_file_path: /path/to/your/file
  append: false

Composite​

The CompositeTransport is designed to combine multiple transports, allowing event emission to several destinations. This is useful when events need to be sent to multiple targets, such as a logging system and an API endpoint. The events are delivered sequentially - one after another in a defined order.

Configuration​

  • type - string, must be "composite". Required.
  • transports - a list or a map of transport configurations. Required.
  • continue_on_failure - boolean flag, determines if the process should continue even when one of the transports fails. Default is false.

Behavior​

  • The configured transports will be initialized and used in sequence to emit OpenLineage events.
  • If continue_on_failure is set to false, a failure in one transport will stop the event emission process, and an exception will be raised.
  • If continue_on_failure is true, the failure will be logged, but the remaining transports will still attempt to send the event.

Notes for Multiple Transports​

The composite transport can be used with any OpenLineage transport (e.g. HttpTransport, KafkaTransport, etc).

The transports configuration can be provided in two formats:

  1. A list of transport configurations, where each transport may optionally include a name field.
  2. A map of transport configurations, where the key acts as the name for each transport. The map format is particularly useful for configurations set via environment variables.
Why are transport names used?​

Transport names are not required for basic functionality. Their primary purpose is to enable configuration of composite transports via environment variables, which is only supported when names are defined.

Examples​

transport:
  type: composite
  continueOnFailure: true
  transports:
    - type: http
      url: http://example.com/api
      name: my_http
    - type: http
      url: http://localhost:5000
      endpoint: /api/v1/lineage

Custom Transport Type​

To implement a custom transport, follow the instructions in transport.py.

The type property (required) must be a fully qualified class name that can be imported.

Environment Variables Run Facet​

To include specific environment variables in OpenLineage events, the OpenLineageClient can add them as a facet called EnvironmentVariablesRunFacet. This feature allows you to specify which environment variables should be collected and attached to each emitted event.

To enable this, configure the environment_variables option within the facets section of your OpenLineageClient configuration.

facets:
  environment_variables:
    - VAR1
    - VAR2

Getting Started​

To try out the client, follow the steps below to install and explore OpenLineage, Marquez (the reference implementation of OpenLineage), and the client itself. Then, the instructions will show you how to use these tools to add a run event and datasets to an existing namespace.

Prerequisites​

  • Docker 17.05+
  • Docker Compose 1.29.1+
  • Git (preinstalled on most versions of MacOS; verify your version with git version)
  • 4 GB of available memory (the minimum for Docker — more is strongly recommended)

Install OpenLineage and Marquez​

Clone the Marquez Github repository:

git clone https://github.com/MarquezProject/marquez.git

Install the Python client​

pip install openlineage-python

Start Docker and Marquez​

Start Docker Desktop Run Marquez with preloaded data:

cd marquez
./docker/up.sh --seed

Marquez should be up and running at http://localhost:3000.

Take a moment to explore Marquez to get a sense of how metadata is displayed in the UI. Namespaces – the global contexts for runs and datasets – can be found in the top right corner, and icons for jobs and runs can be found in a tray along the left side.

Next, configure OpenLineage and add a script to your project that will generate a new job and new datasets within an existing namespace (here we’re using the food_delivery namespace that got passed to Marquez with the –seed argument we used earlier).

Create a directory for your script:

..
mkdir python_scripts && cd python_scripts

In the python_scripts directory, create a Python script (we used the name generate_events.py for ours) and an openlineage.yml file.

In openlineage.yml, define a transport type and URL to tell OpenLineage where and how to send metadata:

transport:
  type: http
  url: http://localhost:5000

In generate_events.py, import the Python client and the methods needed to create a job and datasets. Also required (to create a run): the datetime and uuid packages:

from openlineage.client import OpenLineageClient
from openlineage.client.event_v2 import (
    Dataset,
    InputDataset,
    Job,
    OutputDataset,
    Run,
    RunEvent,
    RunState,
)
from openlineage.client.uuid import generate_new_uuid
from datetime import datetime

Then, in the same file, initialize the Python client:

client = OpenLineageClient.from_environment()

It is also possible to specify parameters such as URL for client to connect to, without using environment variables or openlineage.yaml file, by directly setting it up when instantiating OpenLineageClient:

client = OpenLineageClient(url="http://localhost:5000")

For more details about options to setup OpenLineageClient such as API tokens or HTTP transport settings, please refer to the following example

Specify the producer of the new lineage metadata with a string:

producer = "OpenLineage.io/website/blog"

Now you can create some basic dataset objects. These require a namespace and name:

inventory = Dataset(namespace="food_delivery", name="public.inventory")
menus = Dataset(namespace="food_delivery", name="public.menus_1")
orders = Dataset(namespace="food_delivery", name="public.orders_1")

You can also create a job object (we’ve borrowed this one from the existing food_delivery namespace):

job = Job(namespace="food_delivery", name="example.order_data")

To create a run object you’ll need to specify a unique ID:

run = Run(runId=str(generate_new_uuid()))

a START run event:

client.emit(
	RunEvent(
		eventType=RunState.START,
		eventTime=datetime.now().isoformat(),
		run=run, 
        job=job, 
        producer=producer,
	)
)

and, finally, a COMPLETE run event:

client.emit(
	RunEvent(
		eventType=RunState.COMPLETE,
		eventTime=datetime.now().isoformat(),
		run=run, job=job, producer=producer,
		inputs=[inventory],
		outputs=[menus, orders],
	)
)

Now you have a complete script for creating datasets and a run event! Execute it in the terminal to send the metadata to Marquez:

python3 generate_scripts.py

Marquez will update itself automatically, so the new job and datasets should now be visible in the UI. Clicking on the jobs icon (the icon with the three interlocking gears), will make the example.order_data job appear in the list of jobs:

the Marquez jobs list

When you click on the job, you will see a new map displaying the job, input and outputs we created with our script:

the Marquez graph

Full Example Source Code​

#!/usr/bin/env python3
from datetime import datetime, timedelta, timezone
from random import random

from openlineage.client.client import OpenLineageClient, OpenLineageClientOptions
from openlineage.client.event_v2 import (
    Dataset,
    InputDataset,
    Job,
    OutputDataset,
    Run,
    RunEvent,
    RunState,
)
from openlineage.client.facet_v2 import (
    nominal_time_run,
    schema_dataset,
    source_code_location_job,
    sql_job,
)
from openlineage.client.uuid import generate_new_uuid

PRODUCER = "https://github.com/openlineage-user"
namespace = "python_client"
dag_name = "user_trends"

# update to your host
url = "http://mymarquez.host:5000"
api_key = "1234567890ckcu028rzu5l"

client = OpenLineageClient(
    url=url,
    # optional api key in case marquez requires it. When running marquez in
    # your local environment, you usually do not need this.
    options=OpenLineageClientOptions(api_key=api_key),
)

# If you want to log to a file instead of Marquez
# from openlineage.client import OpenLineageClient
# from openlineage.client.transport.file import FileConfig, FileTransport
# 
# file_config = FileConfig(
#     log_file_path="ol.json",
#     append=True,
# )
# 
# client = OpenLineageClient(transport=FileTransport(file_config))


# generates job facet
def job(job_name, sql, location):
    facets = {"sql": sql_job.SQLJobFacet(query=sql)}
    if location != None:
        facets.update(
            {
                "sourceCodeLocation": source_code_location_job.SourceCodeLocationJobFacet(
                    "git", location
                )
            }
        )
    return Job(namespace=namespace, name=job_name, facets=facets)


# geneartes run racet
def run(run_id, hour):
    return Run(
        runId=run_id,
        facets={
            "nominalTime": nominal_time_run.NominalTimeRunFacet(
                nominalStartTime=f"2022-04-14T{twoDigits(hour)}:12:00Z",
                # nominalEndTime=None
            )
        },
    )


# generates dataset
def dataset(name, schema=None, ns=namespace):
    if schema == None:
        facets = {}
    else:
        facets = {"schema": schema}
    return Dataset(namespace=ns, name=name, facets=facets)


# generates output dataset
def outputDataset(dataset, stats):
    output_facets = {"stats": stats, "outputStatistics": stats}
    return OutputDataset(dataset.namespace,
                         dataset.name,
                         facets=dataset.facets,
                         outputFacets=output_facets)


# generates input dataset
def inputDataset(dataset, dq):
    input_facets = {
        "dataQuality": dq,
    }
    return InputDataset(dataset.namespace, dataset.name,
                        facets=dataset.facets,
                        inputFacets=input_facets)


def twoDigits(n):
    if n < 10:
        result = f"0{n}"
    elif n < 100:
        result = f"{n}"
    else:
        raise f"error: {n}"
    return result


now = datetime.now(timezone.utc)


# generates run Event
def runEvents(job_name, sql, inputs, outputs, hour, min, location, duration):
    run_id = str(generate_new_uuid())
    myjob = job(job_name, sql, location)
    myrun = run(run_id, hour)
    started_at = now + timedelta(hours=hour, minutes=min, seconds=20 + round(random() * 10))
    ended_at = started_at + timedelta(minutes=duration, seconds=20 + round(random() * 10))
    return (
        RunEvent(
            eventType=RunState.START,
            eventTime=started_at.isoformat(),
            run=myrun,
            job=myjob,
            producer=PRODUCER,
            inputs=inputs,
            outputs=outputs,
        ),
        RunEvent(
            eventType=RunState.COMPLETE,
            eventTime=ended_at.isoformat(),
            run=myrun,
            job=myjob,
            producer=PRODUCER,
            inputs=inputs,
            outputs=outputs,
        ),
    )


# add run event to the events list
def addRunEvents(events, job_name, sql, inputs, outputs, hour, minutes, location=None, duration=2):
    (start, complete) = runEvents(job_name, sql, inputs, outputs, hour, minutes, location, duration)
    events.append(start)
    events.append(complete)


events = []

# create dataset data
for i in range(0, 5):
    user_counts = dataset("tmp_demo.user_counts")
    user_history = dataset(
        "temp_demo.user_history",
        schema_dataset.SchemaDatasetFacet(
            fields=[
                schema_dataset.SchemaDatasetFacetFields(
                    name="id", type="BIGINT", description="the user id"
                ),
                schema_dataset.SchemaDatasetFacetFields(
                    name="email_domain", type="VARCHAR", description="the user id"
                ),
                schema_dataset.SchemaDatasetFacetFields(
                    name="status", type="BIGINT", description="the user id"
                ),
                schema_dataset.SchemaDatasetFacetFields(
                    name="created_at",
                    type="DATETIME",
                    description="date and time of creation of the user",
                ),
                schema_dataset.SchemaDatasetFacetFields(
                    name="updated_at",
                    type="DATETIME",
                    description="the last time this row was updated",
                ),
                schema_dataset.SchemaDatasetFacetFields(
                    name="fetch_time_utc",
                    type="DATETIME",
                    description="the time the data was fetched",
                ),
                schema_dataset.SchemaDatasetFacetFields(
                    name="load_filename",
                    type="VARCHAR",
                    description="the original file this data was ingested from",
                ),
                schema_dataset.SchemaDatasetFacetFields(
                    name="load_filerow",
                    type="INT",
                    description="the row number in the original file",
                ),
                schema_dataset.SchemaDatasetFacetFields(
                    name="load_timestamp",
                    type="DATETIME",
                    description="the time the data was ingested",
                ),
            ]
        ),
        "snowflake://",
    )

    create_user_counts_sql = """CREATE OR REPLACE TABLE TMP_DEMO.USER_COUNTS AS (
            SELECT DATE_TRUNC(DAY, created_at) date, COUNT(id) as user_count
            FROM TMP_DEMO.USER_HISTORY
            GROUP BY date
            )"""

    # location of the source code
    location = "https://github.com/some/airflow/dags/example/user_trends.py"

    # run simulating Airflow DAG with snowflake operator
    addRunEvents(
        events,
        dag_name + ".create_user_counts",
        create_user_counts_sql,
        [user_history],
        [user_counts],
        i,
        11,
        location,
    )


for event in events:
    from openlineage.client.serde import Serde

    print(event)
    print(Serde.to_json(event))
    # time.sleep(1)
    client.emit(event)

The resulting lineage events received by Marquez would look like this.

the Marquez graph