Concepts

The Airflow platform is a tool for describing, executing, and monitoring workflows.

Core Ideas

DAGs

In Airflow, a DAG – or a Directed Acyclic Graph – is a collection of all the tasks you want to run, organized in a way that reflects their relationships and dependencies.

A DAG is defined in a Python script, which represents the DAGs structure (tasks and their dependencies) as code.

For example, a simple DAG could consist of three tasks: A, B, and C. It could say that A has to run successfully before B can run, but C can run anytime. It could say that task A times out after 5 minutes, and B can be restarted up to 5 times in case it fails. It might also say that the workflow will run every night at 10pm, but shouldn’t start until a certain date.

In this way, a DAG describes how you want to carry out your workflow; but notice that we haven’t said anything about what we actually want to do! A, B, and C could be anything. Maybe A prepares data for B to analyze while C sends an email. Or perhaps A monitors your location so B can open your garage door while C turns on your house lights. The important thing is that the DAG isn’t concerned with what its constituent tasks do; its job is to make sure that whatever they do happens at the right time, or in the right order, or with the right handling of any unexpected issues.

DAGs are defined in standard Python files that are placed in Airflow’s DAG_FOLDER. Airflow will execute the code in each file to dynamically build the DAG objects. You can have as many DAGs as you want, each describing an arbitrary number of tasks. In general, each one should correspond to a single logical workflow.

Note

When searching for DAGs, Airflow only considers python files that contain the strings “airflow” and “DAG” by default. To consider all python files instead, disable the DAG_DISCOVERY_SAFE_MODE configuration flag.

Scope

Airflow will load any DAG object it can import from a DAGfile. Critically, that means the DAG must appear in globals(). Consider the following two DAGs. Only dag_1 will be loaded; the other one only appears in a local scope.

dag_1 = DAG('this_dag_will_be_discovered')

def my_function():
    dag_2 = DAG('but_this_dag_will_not')

my_function()

Sometimes this can be put to good use. For example, a common pattern with SubDagOperator is to define the subdag inside a function so that Airflow doesn’t try to load it as a standalone DAG.

Default Arguments

If a dictionary of default_args is passed to a DAG, it will apply them to any of its operators. This makes it easy to apply a common parameter to many operators without having to type it many times.

default_args = {
    'start_date': datetime(2016, 1, 1),
    'owner': 'airflow'
}

dag = DAG('my_dag', default_args=default_args)
op = DummyOperator(task_id='dummy', dag=dag)
print(op.owner) # Airflow

Context Manager

Added in Airflow 1.8

DAGs can be used as context managers to automatically assign new operators to that DAG.

with DAG('my_dag', start_date=datetime(2016, 1, 1)) as dag:
    op = DummyOperator('op')

op.dag is dag # True

Functional DAGs

DAGs can be defined using functional abstractions. Outputs and inputs are sent between tasks using XCom values. In addition, you can wrap functions as tasks using the task decorator. Airflow will also automatically add dependencies between tasks to ensure that XCom messages are available when operators are executed.

Example DAG with functional abstraction

with DAG(
    'send_server_ip', default_args=default_args, schedule_interval=None
) as dag:

  # Using default connection as it's set to httpbin.org by default
  get_ip = SimpleHttpOperator(
      task_id='get_ip', endpoint='get', method='GET', xcom_push=True
  )

  @dag.task(multiple_outputs=True)
  def prepare_email(raw_json: str) -> Dict[str, str]:
    external_ip = json.loads(raw_json)['origin']
    return {
      'subject':f'Server connected from {external_ip}',
      'body': f'Seems like today your server executing Airflow is connected from the external IP {external_ip}<br>'
    }

  email_info = prepare_email(get_ip.output)

  send_email = EmailOperator(
      task_id='send_email',
      to='example@example.com',
      subject=email_info['subject'],
      html_content=email_info['body']
  )

DAG Runs

A DAG run is an instantiation of a DAG, containing task instances that run for a specific execution_date.

A DAG run is usually created by the Airflow scheduler, but can also be created by an external trigger. Multiple DAG runs may be running at once for a particular DAG, each of them having a different execution_date. For example, we might currently have two DAG runs that are in progress for 2016-01-01 and 2016-01-02 respectively.

execution_date

The execution_date is the logical date and time which the DAG Run, and its task instances, are running for.

This allows task instances to process data for the desired logical date & time. While a task_instance or DAG run might have a actual start date of now, their logical date might be 3 months ago because we are busy reloading something.

In the prior example the execution_date was 2016-01-01 for the first DAG Run and 2016-01-02 for the second.

A DAG run and all task instances created within it are instanced with the same execution_date, so that logically you can think of a DAG run as simulating the DAG running all of its tasks at some previous date & time specified by the execution_date.

Tasks

A Task defines a unit of work within a DAG; it is represented as a node in the DAG graph, and it is written in Python.

Each task is an implementation of an Operator, for example a PythonOperator to execute some Python code, or a BashOperator to run a Bash command.

The task implements an operator by defining specific values for that operator, such as a Python callable in the case of PythonOperator or a Bash command in the case of BashOperator.

Relations between Tasks

Consider the following DAG with two tasks. Each task is a node in our DAG, and there is a dependency from task_1 to task_2:

with DAG('my_dag', start_date=datetime(2016, 1, 1)) as dag:
    task_1 = DummyOperator('task_1')
    task_2 = DummyOperator('task_2')
    task_1 >> task_2 # Define dependencies

We can say that task_1 is upstream of task_2, and conversely task_2 is downstream of task_1. When a DAG Run is created, task_1 will start running and task_2 waits for task_1 to complete successfully before it may start.

Python task decorator

Airflow task decorator converts any Python function to an Airflow operator. The decorated function can be called once to set the arguments and key arguments for operator execution.

with DAG('my_dag', start_date=datetime(2020, 5, 15)) as dag:
    @dag.task
    def hello_world():
        print('hello world!')


    # Also...
    from airflow.decorators import task


    @task
    def hello_name(name: str):
        print(f'hello {name}!')


    hello_name('Airflow users')

Task decorator captures returned values and sends them to the XCom backend. By default, returned value is saved as a single XCom value. You can set multiple_outputs key argument to True to unroll dictionaries, lists or tuples into seprate XCom values. This can be used with regular operators to create functional DAGs.

Calling a decorated function returns an XComArg instance. You can use it to set templated fields on downstream operators.

You can call a decorated function more than once in a DAG. The decorated function will automatically generate a unique task_id for each generated operator.

with DAG('my_dag', start_date=datetime(2020, 5, 15)) as dag:

  @dag.task
  def update_user(user_id: int):
    ...

  # Avoid generating this list dynamically to keep DAG topology stable between DAG runs
  for user_id in user_ids:
    update_user(user_id)

  # This will generate an operator for each user_id

Task ids are generated by appending a number at the end of the original task id. For the above example, the DAG will have the following task ids: [update_user, update_user__1, update_user__2, ... update_user__n].

Task Instances

A task instance represents a specific run of a task and is characterized as the combination of a DAG, a task, and a point in time (execution_date). Task instances also have an indicative state, which could be “running”, “success”, “failed”, “skipped”, “up for retry”, etc.

Tasks are defined in DAGs, and both are written in Python code to define what you want to do. Task Instances belong to DAG Runs, have an associated execution_date, and are instantiated, runnable entities.

Relations between Task Instances

Again consider the following tasks, defined for some DAG:

with DAG('my_dag', start_date=datetime(2016, 1, 1)) as dag:
    task_1 = DummyOperator('task_1')
    task_2 = DummyOperator('task_2')
    task_1 >> task_2 # Define dependencies

When we enable this DAG, the scheduler creates several DAG Runs - one with execution_date of 2016-01-01, one with execution_date of 2016-01-02, and so on up to the current date.

Each DAG Run will contain a task_1 Task Instance and a task_2 Task instance. Both Task Instances will have execution_date equal to the DAG Run’s execution_date, and each task_2 will be downstream of (depends on) its task_1.

We can also say that task_1 for 2016-01-01 is the previous task instance of the task_1 for 2016-01-02. Or that the DAG Run for 2016-01-01 is the previous DAG Run to the DAG Run of 2016-01-02. Here, previous refers to the logical past/prior execution_date, that runs independently of other runs, and upstream refers to a dependency within the same run and having the same execution_date.

Note

The Airflow documentation sometimes refers to previous instead of upstream in places, and vice-versa. If you find any occurrences of this, please help us improve by contributing some corrections!

Task Lifecycle

A task goes through various stages from start to completion. In the Airflow UI (graph and tree views), these stages are displayed by a color representing each stage:

_images/task_stages.png

The complete lifecycle of the task looks like this:

_images/task_lifecycle_diagram.png

The happy flow consists of the following stages:

  1. No status (scheduler created empty task instance)

  2. Scheduled (scheduler determined task instance needs to run)

  3. Queued (scheduler sent task to executor to run on the queue)

  4. Running (worker picked up a task and is now running it)

  5. Success (task completed)8

There is also visual difference between scheduled and manually triggered DAGs/tasks:

_images/task_manual_vs_scheduled.png

The DAGs/tasks with a black border are scheduled runs, whereas the non-bordered DAGs/tasks are manually triggered, i.e. by airflow dags trigger.

Operators

While DAGs describe how to run a workflow, Operators determine what actually gets done by a task.

An operator describes a single task in a workflow. Operators are usually (but not always) atomic, meaning they can stand on their own and don’t need to share resources with any other operators. The DAG will make sure that operators run in the correct order; other than those dependencies, operators generally run independently. In fact, they may run on two completely different machines.

This is a subtle but very important point: in general, if two operators need to share information, like a filename or small amount of data, you should consider combining them into a single operator. If it absolutely can’t be avoided, Airflow does have a feature for operator cross-communication called XCom that is described in the section XComs

Airflow provides operators for many common tasks, including:

In addition to these basic building blocks, there are many more specific operators: DockerOperator, HiveOperator, S3FileTransformOperator, PrestoToMySqlOperator, SlackAPIOperator… you get the idea!

Operators are only loaded by Airflow if they are assigned to a DAG.

DAG Assignment

Added in Airflow 1.8

Operators do not have to be assigned to DAGs immediately (previously dag was a required argument). However, once an operator is assigned to a DAG, it can not be transferred or unassigned. DAG assignment can be done explicitly when the operator is created, through deferred assignment, or even inferred from other operators.

dag = DAG('my_dag', start_date=datetime(2016, 1, 1))

# sets the DAG explicitly
explicit_op = DummyOperator(task_id='op1', dag=dag)

# deferred DAG assignment
deferred_op = DummyOperator(task_id='op2')
deferred_op.dag = dag

# inferred DAG assignment (linked operators must be in the same DAG)
inferred_op = DummyOperator(task_id='op3')
inferred_op.set_upstream(deferred_op)

Bitshift Composition

Added in Airflow 1.8

We recommend you setting operator relationships with bitshift operators rather than set_upstream() and set_downstream().

Traditionally, operator relationships are set with the set_upstream() and set_downstream() methods. In Airflow 1.8, this can be done with the Python bitshift operators >> and <<. The following four statements are all functionally equivalent:

op1 >> op2
op1.set_downstream(op2)

op2 << op1
op2.set_upstream(op1)

When using the bitshift to compose operators, the relationship is set in the direction that the bitshift operator points. For example, op1 >> op2 means that op1 runs first and op2 runs second. Multiple operators can be composed – keep in mind the chain is executed left-to-right and the rightmost object is always returned. For example:

op1 >> op2 >> op3 << op4

is equivalent to:

op1.set_downstream(op2)
op2.set_downstream(op3)
op3.set_upstream(op4)

We can put this all together to build a simple pipeline:

with DAG('my_dag', start_date=datetime(2016, 1, 1)) as dag:
    (
        DummyOperator(task_id='dummy_1')
        >> BashOperator(
            task_id='bash_1',
            bash_command='echo "HELLO!"')
        >> PythonOperator(
            task_id='python_1',
            python_callable=lambda: print("GOODBYE!"))
    )

Bitshift can also be used with lists. For example:

op1 >> [op2, op3] >> op4

is equivalent to:

op1 >> op2 >> op4
op1 >> op3 >> op4

and equivalent to:

op1.set_downstream([op2, op3])
op4.set_upstream([op2, op3])

Relationship Builders

Moved in Airflow 2.0

In Airflow 2.0 those two methods moved from airflow.utils.helpers to airflow.models.baseoperator.

chain and cross_downstream function provide easier ways to set relationships between operators in specific situation.

When setting a relationship between two lists, if we want all operators in one list to be upstream to all operators in the other, we cannot use a single bitshift composition. Instead we have to split one of the lists:

[op1, op2, op3] >> op4
[op1, op2, op3] >> op5
[op1, op2, op3] >> op6

cross_downstream could handle list relationships easier.

cross_downstream([op1, op2, op3], [op4, op5, op6])

When setting single direction relationships to many operators, we could concat them with bitshift composition.

op1 >> op2 >> op3 >> op4 >> op5

This can be accomplished using chain

chain(op1, op2, op3, op4, op5)

even without operator’s name

chain([DummyOperator(task_id='op' + i, dag=dag) for i in range(1, 6)])

chain can handle a list of operators

chain(op1, [op2, op3], op4)

is equivalent to:

op1 >> [op2, op3] >> op4

When chain sets relationships between two lists of operators, they must have the same size.

chain(op1, [op2, op3], [op4, op5], op6)

is equivalent to:

op1 >> [op2, op3]
op2 >> op4
op3 >> op5
[op4, op5] >> op6

Workflows

You’re now familiar with the core building blocks of Airflow. Some of the concepts may sound very similar, but the vocabulary can be conceptualized like this:

  • DAG: The work (tasks), and the order in which work should take place (dependencies), written in Python.

  • DAG Run: An instance of a DAG for a particular logical date and time.

  • Operator: A class that acts as a template for carrying out some work.

  • Task: Defines work by implementing an operator, written in Python.

  • Task Instance: An instance of a task - that has been assigned to a DAG and has a state associated with a specific DAG run (i.e for a specific execution_date).

  • execution_date: The logical date and time for a DAG Run and its Task Instances.

By combining DAGs and Operators to create TaskInstances, you can build complex workflows.

Additional Functionality

In addition to the core Airflow objects, there are a number of more complex features that enable behaviors like limiting simultaneous access to resources, cross-communication, conditional execution, and more.

Hooks

Hooks are interfaces to external platforms and databases like Hive, S3, MySQL, Postgres, HDFS, and Pig. Hooks implement a common interface when possible, and act as a building block for operators. They also use the airflow.models.connection.Connection model to retrieve hostnames and authentication information. Hooks keep authentication code and information out of pipelines, centralized in the metadata database.

Hooks are also very useful on their own to use in Python scripts, Airflow airflow.operators.PythonOperator, and in interactive environments like iPython or Jupyter Notebook.

Pools

Some systems can get overwhelmed when too many processes hit them at the same time. Airflow pools can be used to limit the execution parallelism on arbitrary sets of tasks. The list of pools is managed in the UI (Menu -> Admin -> Pools) by giving the pools a name and assigning it a number of worker slots. Tasks can then be associated with one of the existing pools by using the pool parameter when creating tasks (i.e., instantiating operators).

aggregate_db_message_job = BashOperator(
    task_id='aggregate_db_message_job',
    execution_timeout=timedelta(hours=3),
    pool='ep_data_pipeline_db_msg_agg',
    bash_command=aggregate_db_message_job_cmd,
    dag=dag)
aggregate_db_message_job.set_upstream(wait_for_empty_queue)

The pool parameter can be used in conjunction with priority_weight to define priorities in the queue, and which tasks get executed first as slots open up in the pool. The default priority_weight is 1, and can be bumped to any number. When sorting the queue to evaluate which task should be executed next, we use the priority_weight, summed up with all of the priority_weight values from tasks downstream from this task. You can use this to bump a specific important task and the whole path to that task gets prioritized accordingly.

Tasks will be scheduled as usual while the slots fill up. Once capacity is reached, runnable tasks get queued and their state will show as such in the UI. As slots free up, queued tasks start running based on the priority_weight (of the task and its descendants).

Pools are not thread-safe , in case of more than one scheduler in localExecutor Mode you can’t ensure the non-scheduling of task even if the pool is full.

Note that if tasks are not given a pool, they are assigned to a default pool default_pool. default_pool is initialized with 128 slots and can changed through the UI or CLI (though it cannot be removed).

To combine Pools with SubDAGs see the SubDAGs section.

Connections

The information needed to connect to external systems is stored in the Airflow metastore database and can be managed in the UI (Menu -> Admin -> Connections). A conn_id is defined there, and hostname / login / password / schema information attached to it. Airflow pipelines retrieve centrally-managed connections information by specifying the relevant conn_id.

You may add more than one connection with the same conn_id. When there is more than one connection with the same conn_id, the get_connection() method on BaseHook will choose one connection randomly. This can be be used to provide basic load balancing and fault tolerance, when used in conjunction with retries.

Airflow also provides a mechanism to store connections outside the database, e.g. in environment variables. Additional sources may be enabled, e.g. AWS SSM Parameter Store, or you may roll your own secrets backend.

Many hooks have a default conn_id, where operators using that hook do not need to supply an explicit connection ID. For example, the default conn_id for the PostgresHook is postgres_default.

See Managing Connections for details on creating and managing connections.

Queues

When using the CeleryExecutor, the Celery queues that tasks are sent to can be specified. queue is an attribute of BaseOperator, so any task can be assigned to any queue. The default queue for the environment is defined in the airflow.cfg’s celery -> default_queue. This defines the queue that tasks get assigned to when not specified, as well as which queue Airflow workers listen to when started.

Workers can listen to one or multiple queues of tasks. When a worker is started (using the command airflow celery worker), a set of comma-delimited queue names can be specified (e.g. airflow celery worker -q spark). This worker will then only pick up tasks wired to the specified queue(s).

This can be useful if you need specialized workers, either from a resource perspective (for say very lightweight tasks where one worker could take thousands of tasks without a problem), or from an environment perspective (you want a worker running from within the Spark cluster itself because it needs a very specific environment and security rights).

XComs

XComs let tasks exchange messages, allowing more nuanced forms of control and shared state. The name is an abbreviation of “cross-communication”. XComs are principally defined by a key, value, and timestamp, but also track attributes like the task/DAG that created the XCom and when it should become visible. Any object that can be pickled can be used as an XCom value, so users should make sure to use objects of appropriate size.

XComs can be “pushed” (sent) or “pulled” (received). When a task pushes an XCom, it makes it generally available to other tasks. Tasks can push XComs at any time by calling the xcom_push() method. In addition, if a task returns a value (either from its Operator’s execute() method, or from a PythonOperator’s python_callable function), then an XCom containing that value is automatically pushed.

Tasks call xcom_pull() to retrieve XComs, optionally applying filters based on criteria like key, source task_ids, and source dag_id. By default, xcom_pull() filters for the keys that are automatically given to XComs when they are pushed by being returned from execute functions (as opposed to XComs that are pushed manually).

If xcom_pull is passed a single string for task_ids, then the most recent XCom value from that task is returned; if a list of task_ids is passed, then a corresponding list of XCom values is returned.

# inside a PythonOperator called 'pushing_task'
def push_function():
    return value

# inside another PythonOperator
def pull_function(task_instance):
    value = task_instance.xcom_pull(task_ids='pushing_task')

When specifying arguments that are part of the context, they will be automatically passed to the function.

It is also possible to pull XCom directly in a template, here’s an example of what this may look like:

SELECT * FROM {{ task_instance.xcom_pull(task_ids='foo', key='table_name') }}

Note that XComs are similar to Variables, but are specifically designed for inter-task communication rather than global settings.

Custom XCom backend

It is possible to change XCom behaviour os serialization and deserialization of tasks’ result. To do this one have to change xcom_backend parameter in Airflow config. Provided value should point to a class that is subclass of BaseXCom. To alter the serialaization / deserialization mechanism the custom class should override serialize_value and deserialize_value methods.

Variables

Variables are a generic way to store and retrieve arbitrary content or settings as a simple key value store within Airflow. Variables can be listed, created, updated and deleted from the UI (Admin -> Variables), code or CLI. In addition, json settings files can be bulk uploaded through the UI. While your pipeline code definition and most of your constants and variables should be defined in code and stored in source control, it can be useful to have some variables or configuration items accessible and modifiable through the UI.

from airflow.models import Variable
foo = Variable.get("foo")
bar = Variable.get("bar", deserialize_json=True)
baz = Variable.get("baz", default_var=None)

The second call assumes json content and will be deserialized into bar. Note that Variable is a sqlalchemy model and can be used as such. The third call uses the default_var parameter with the value None, which either returns an existing value or None if the variable isn’t defined. The get function will throw a KeyError if the variable doesn’t exist and no default is provided.

You can use a variable from a jinja template with the syntax :

echo {{ var.value.<variable_name> }}

or if you need to deserialize a json object from the variable :

echo {{ var.json.<variable_name> }}

Storing Variables in Environment Variables

New in version 1.10.10.

Airflow Variables can also be created and managed using Environment Variables. The environment variable naming convention is AIRFLOW_VAR_{VARIABLE_NAME}, all uppercase. So if your variable key is FOO then the variable name should be AIRFLOW_VAR_FOO.

For example,

export AIRFLOW_VAR_FOO=BAR

# To use JSON, store them as JSON strings
export AIRFLOW_VAR_FOO_BAZ='{"hello":"world"}'

You can use them in your DAGs as:

from airflow.models import Variable
foo = Variable.get("foo")
foo_json = Variable.get("foo_baz", deserialize_json=True)

Note

Single underscores surround VAR. This is in contrast with the way airflow.cfg parameters are stored, where double underscores surround the config section name. Variables set using Environment Variables would not appear in the Airflow UI but you will be able to use it in your DAG file.

Branching

Sometimes you need a workflow to branch, or only go down a certain path based on an arbitrary condition which is typically related to something that happened in an upstream task. One way to do this is by using the BranchPythonOperator.

The BranchPythonOperator is much like the PythonOperator except that it expects a python_callable that returns a task_id (or list of task_ids). The task_id returned is followed, and all of the other paths are skipped. The task_id returned by the Python function has to reference a task directly downstream from the BranchPythonOperator task.

Note that when a path is a downstream task of the returned task (list), it will not be skipped:

_images/branch_note.png

Paths of the branching task are branch_a, join and branch_b. Since join is a downstream task of branch_a, it will be excluded from the skipped tasks when branch_a is returned by the Python callable.

The BranchPythonOperator can also be used with XComs allowing branching context to dynamically decide what branch to follow based on upstream tasks. For example:

def branch_func(ti):
    xcom_value = int(ti.xcom_pull(task_ids='start_task'))
    if xcom_value >= 5:
        return 'continue_task'
    else:
        return 'stop_task'

start_op = BashOperator(
    task_id='start_task',
    bash_command="echo 5",
    xcom_push=True,
    dag=dag)

branch_op = BranchPythonOperator(
    task_id='branch_task',
    python_callable=branch_func,
    dag=dag)

continue_op = DummyOperator(task_id='continue_task', dag=dag)
stop_op = DummyOperator(task_id='stop_task', dag=dag)

start_op >> branch_op >> [continue_op, stop_op]

If you wish to implement your own operators with branching functionality, you can inherit from BaseBranchOperator, which behaves similarly to BranchPythonOperator but expects you to provide an implementation of the method choose_branch. As with the callable for BranchPythonOperator, this method should return the ID of a downstream task, or a list of task IDs, which will be run, and all others will be skipped.

class MyBranchOperator(BaseBranchOperator):
    def choose_branch(self, context):
        """
        Run an extra branch on the first day of the month
        """
        if context['execution_date'].day == 1:
            return ['daily_task_id', 'monthly_task_id']
        else:
            return 'daily_task_id'

SubDAGs

SubDAGs are perfect for repeating patterns. Defining a function that returns a DAG object is a nice design pattern when using Airflow.

Airbnb uses the stage-check-exchange pattern when loading data. Data is staged in a temporary table, after which data quality checks are performed against that table. Once the checks all pass the partition is moved into the production table.

As another example, consider the following DAG:

_images/subdag_before.png

We can combine all of the parallel task-* operators into a single SubDAG, so that the resulting DAG resembles the following:

_images/subdag_after.png

Note that SubDAG operators should contain a factory method that returns a DAG object. This will prevent the SubDAG from being treated like a separate DAG in the main UI. For example:

airflow/example_dags/subdags/subdag.pyView Source

from airflow import DAG
from airflow.operators.dummy_operator import DummyOperator


def subdag(parent_dag_name, child_dag_name, args):
    """
    Generate a DAG to be used as a subdag.

    :param str parent_dag_name: Id of the parent DAG
    :param str child_dag_name: Id of the child DAG
    :param dict args: Default arguments to provide to the subdag
    :return: DAG to use as a subdag
    :rtype: airflow.models.DAG
    """
    dag_subdag = DAG(
        dag_id='%s.%s' % (parent_dag_name, child_dag_name),
        default_args=args,
        schedule_interval="@daily",
    )

    for i in range(5):
        DummyOperator(
            task_id='%s-task-%s' % (child_dag_name, i + 1),
            default_args=args,
            dag=dag_subdag,
        )

    return dag_subdag

This SubDAG can then be referenced in your main DAG file:

airflow/example_dags/example_subdag_operator.pyView Source

from airflow import DAG
from airflow.example_dags.subdags.subdag import subdag
from airflow.operators.dummy_operator import DummyOperator
from airflow.operators.subdag_operator import SubDagOperator
from airflow.utils.dates import days_ago

DAG_NAME = 'example_subdag_operator'

args = {
    'owner': 'airflow',
    'start_date': days_ago(2),
}

dag = DAG(
    dag_id=DAG_NAME,
    default_args=args,
    schedule_interval="@once",
    tags=['example']
)

start = DummyOperator(
    task_id='start',
    dag=dag,
)

section_1 = SubDagOperator(
    task_id='section-1',
    subdag=subdag(DAG_NAME, 'section-1', args),
    dag=dag,
)

some_other_task = DummyOperator(
    task_id='some-other-task',
    dag=dag,
)

section_2 = SubDagOperator(
    task_id='section-2',
    subdag=subdag(DAG_NAME, 'section-2', args),
    dag=dag,
)

end = DummyOperator(
    task_id='end',
    dag=dag,
)

start >> section_1 >> some_other_task >> section_2 >> end

You can zoom into a SubDagOperator from the graph view of the main DAG to show the tasks contained within the SubDAG:

_images/subdag_zoom.png

Some other tips when using SubDAGs:

  • by convention, a SubDAG’s dag_id should be prefixed by its parent and a dot. As in parent.child

  • share arguments between the main DAG and the SubDAG by passing arguments to the SubDAG operator (as demonstrated above)

  • SubDAGs must have a schedule and be enabled. If the SubDAG’s schedule is set to None or @once, the SubDAG will succeed without having done anything

  • clearing a SubDagOperator also clears the state of the tasks within

  • marking success on a SubDagOperator does not affect the state of the tasks within

  • refrain from using depends_on_past=True in tasks within the SubDAG as this can be confusing

  • it is possible to specify an executor for the SubDAG. It is common to use the SequentialExecutor if you want to run the SubDAG in-process and effectively limit its parallelism to one. Using LocalExecutor can be problematic as it may over-subscribe your worker, running multiple tasks in a single slot

See airflow/example_dags for a demonstration.

Note that airflow pool is not honored by SubDagOperator. Hence resources could be consumed by SubdagOperators.

SLAs

Service Level Agreements, or time by which a task or DAG should have succeeded, can be set at a task level as a timedelta. If one or many instances have not succeeded by that time, an alert email is sent detailing the list of tasks that missed their SLA. The event is also recorded in the database and made available in the web UI under Browse->SLA Misses where events can be analyzed and documented.

SLAs can be configured for scheduled tasks by using the sla parameter. In addition to sending alerts to the addresses specified in a task’s email parameter, the sla_miss_callback specifies an additional Callable object to be invoked when the SLA is not met.

If you don’t want to check SLAs, you can disable globally (all the DAGs) by setting check_slas=False under [core] section in airflow.cfg file:

[core]
check_slas = False

Note

For information on the email configuration, see Email Configuration

Trigger Rules

Though the normal workflow behavior is to trigger tasks when all their directly upstream tasks have succeeded, Airflow allows for more complex dependency settings.

All operators have a trigger_rule argument which defines the rule by which the generated task get triggered. The default value for trigger_rule is all_success and can be defined as “trigger this task when all directly upstream tasks have succeeded”. All other rules described here are based on direct parent tasks and are values that can be passed to any operator while creating tasks:

  • all_success: (default) all parents have succeeded

  • all_failed: all parents are in a failed or upstream_failed state

  • all_done: all parents are done with their execution

  • one_failed: fires as soon as at least one parent has failed, it does not wait for all parents to be done

  • one_success: fires as soon as at least one parent succeeds, it does not wait for all parents to be done

  • none_failed: all parents have not failed (failed or upstream_failed) i.e. all parents have succeeded or been skipped

  • none_failed_or_skipped: all parents have not failed (failed or upstream_failed) and at least one parent has succeeded.

  • none_skipped: no parent is in a skipped state, i.e. all parents are in a success, failed, or upstream_failed state

  • dummy: dependencies are just for show, trigger at will

Note that these can be used in conjunction with depends_on_past (boolean) that, when set to True, keeps a task from getting triggered if the previous schedule for the task hasn’t succeeded.

One must be aware of the interaction between trigger rules and skipped tasks in schedule level. Skipped tasks will cascade through trigger rules all_success and all_failed but not all_done, one_failed, one_success, none_failed, none_failed_or_skipped, none_skipped and dummy.

For example, consider the following DAG:

#dags/branch_without_trigger.py
import datetime as dt

from airflow.models import DAG
from airflow.operators.dummy_operator import DummyOperator
from airflow.operators.python import BranchPythonOperator

dag = DAG(
    dag_id='branch_without_trigger',
    schedule_interval='@once',
    start_date=dt.datetime(2019, 2, 28)
)

run_this_first = DummyOperator(task_id='run_this_first', dag=dag)
branching = BranchPythonOperator(
    task_id='branching', dag=dag,
    python_callable=lambda: 'branch_a'
)

branch_a = DummyOperator(task_id='branch_a', dag=dag)
follow_branch_a = DummyOperator(task_id='follow_branch_a', dag=dag)

branch_false = DummyOperator(task_id='branch_false', dag=dag)

join = DummyOperator(task_id='join', dag=dag)

run_this_first >> branching
branching >> branch_a >> follow_branch_a >> join
branching >> branch_false >> join

In the case of this DAG, join is downstream of follow_branch_a and branch_false. The join task will show up as skipped because its trigger_rule is set to all_success by default and skipped tasks will cascade through all_success.

_images/branch_without_trigger.png

By setting trigger_rule to none_failed_or_skipped in join task,

#dags/branch_with_trigger.py
...
join = DummyOperator(task_id='join', dag=dag, trigger_rule='none_failed_or_skipped')
...

The join task will be triggered as soon as branch_false has been skipped (a valid completion state) and follow_branch_a has succeeded. Because skipped tasks will not cascade through none_failed_or_skipped.

_images/branch_with_trigger.png

Latest Run Only

Standard workflow behavior involves running a series of tasks for a particular date/time range. Some workflows, however, perform tasks that are independent of run time but need to be run on a schedule, much like a standard cron job. In these cases, backfills or running jobs missed during a pause just wastes CPU cycles.

For situations like this, you can use the LatestOnlyOperator to skip tasks that are not being run during the most recent scheduled run for a DAG. The LatestOnlyOperator skips all direct downstream tasks, if the time right now is not between its execution_time and the next scheduled execution_time or the DagRun has been externally triggered.

For example, consider the following DAG:

airflow/example_dags/example_latest_only_with_trigger.pyView Source

import datetime as dt

from airflow import DAG
from airflow.operators.dummy_operator import DummyOperator
from airflow.operators.latest_only_operator import LatestOnlyOperator
from airflow.utils.dates import days_ago
from airflow.utils.trigger_rule import TriggerRule

dag = DAG(
    dag_id='latest_only_with_trigger',
    schedule_interval=dt.timedelta(hours=4),
    start_date=days_ago(2),
    tags=['example']
)

latest_only = LatestOnlyOperator(task_id='latest_only', dag=dag)
task1 = DummyOperator(task_id='task1', dag=dag)
task2 = DummyOperator(task_id='task2', dag=dag)
task3 = DummyOperator(task_id='task3', dag=dag)
task4 = DummyOperator(task_id='task4', dag=dag, trigger_rule=TriggerRule.ALL_DONE)

latest_only >> task1 >> [task3, task4]
task2 >> [task3, task4]

In the case of this DAG, the task task1 is directly downstream of latest_only and will be skipped for all runs except the latest. task2 is entirely independent of latest_only and will run in all scheduled periods. task3 is downstream of task1 and task2 and because of the default trigger_rule being all_success will receive a cascaded skip from task1. task4 is downstream of task1 and task2, but it will not be skipped, since its trigger_rule is set to all_done.

_images/latest_only_with_trigger.png

Zombies & Undeads

Task instances die all the time, usually as part of their normal life cycle, but sometimes unexpectedly.

Zombie tasks are characterized by the absence of an heartbeat (emitted by the job periodically) and a running status in the database. They can occur when a worker node can’t reach the database, when Airflow processes are killed externally, or when a node gets rebooted for instance. Zombie killing is performed periodically by the scheduler’s process.

Undead processes are characterized by the existence of a process and a matching heartbeat, but Airflow isn’t aware of this task as running in the database. This mismatch typically occurs as the state of the database is altered, most likely by deleting rows in the “Task Instances” view in the UI. Tasks are instructed to verify their state as part of the heartbeat routine, and terminate themselves upon figuring out that they are in this “undead” state.

Cluster Policy

In case you want to apply cluster-wide mutations to the Airflow tasks, you can either mutate the task right after the DAG is loaded or mutate the task instance before task execution.

Mutate tasks after DAG loaded

To mutate the task right after the DAG is parsed, you can define a policy function in airflow_local_settings.py that mutates the task based on other task or DAG attributes (through task.dag). It receives a single argument as a reference to the task object and you can alter its attributes.

For example, this function could apply a specific queue property when using a specific operator, or enforce a task timeout policy, making sure that no tasks run for more than 48 hours. Here’s an example of what this may look like inside your airflow_local_settings.py:

def policy(task):
    if task.__class__.__name__ == 'HivePartitionSensor':
        task.queue = "sensor_queue"
    if task.timeout > timedelta(hours=48):
        task.timeout = timedelta(hours=48)

Please note, cluster policy will have precedence over task attributes defined in DAG meaning if task.sla is defined in dag and also mutated via cluster policy then later will have precedence.

Mutate task instances before task execution

To mutate the task instance before the task execution, you can define a task_instance_mutation_hook function in airflow_local_settings.py that mutates the task instance.

For example, this function re-routes the task to execute in a different queue during retries:

def task_instance_mutation_hook(ti):
    if ti.try_number >= 1:
        ti.queue = 'retry_queue'

Where to put airflow_local_settings.py?

Add a airflow_local_settings.py file to your $PYTHONPATH or to $AIRFLOW_HOME/config folder.

Documentation & Notes

It’s possible to add documentation or notes to your DAGs & task objects that become visible in the web interface (“Graph View” & “Tree View” for DAGs, “Task Details” for tasks). There are a set of special task attributes that get rendered as rich content if defined:

attribute

rendered to

doc

monospace

doc_json

json

doc_yaml

yaml

doc_md

markdown

doc_rst

reStructuredText

Please note that for DAGs, doc_md is the only attribute interpreted.

This is especially useful if your tasks are built dynamically from configuration files, it allows you to expose the configuration that led to the related tasks in Airflow.

"""
### My great DAG
"""

dag = DAG('my_dag', default_args=default_args)
dag.doc_md = __doc__

t = BashOperator("foo", dag=dag)
t.doc_md = """\
#Title"
Here's a [url](www.airbnb.com)
"""

This content will get rendered as markdown respectively in the “Graph View” and “Task Details” pages.

Jinja Templating

Airflow leverages the power of Jinja Templating and this can be a powerful tool to use in combination with macros (see the Macros reference section).

For example, say you want to pass the execution date as an environment variable to a Bash script using the BashOperator.

# The execution date as YYYY-MM-DD
date = "{{ ds }}"
t = BashOperator(
    task_id='test_env',
    bash_command='/tmp/test.sh ',
    dag=dag,
    env={'EXECUTION_DATE': date})

Here, {{ ds }} is a macro, and because the env parameter of the BashOperator is templated with Jinja, the execution date will be available as an environment variable named EXECUTION_DATE in your Bash script.

You can use Jinja templating with every parameter that is marked as “templated” in the documentation. Template substitution occurs just before the pre_execute function of your operator is called.

You can also use Jinja templating with nested fields, as long as these nested fields are marked as templated in the structure they belong to: fields registered in template_fields property will be submitted to template substitution, like the path field in the example below:

class MyDataReader:
  template_fields = ['path']

  def __init__(self, my_path):
    self.path = my_path

  # [additional code here...]

t = PythonOperator(
    task_id='transform_data',
    python_callable=transform_data
    op_args=[
      MyDataReader('/tmp/{{ ds }}/my_file')
    ],
    dag=dag)

Note

template_fields property can equally be a class variable or an instance variable.

Deep nested fields can also be substituted, as long as all intermediate fields are marked as template fields:

class MyDataTransformer:
  template_fields = ['reader']

  def __init__(self, my_reader):
    self.reader = my_reader

  # [additional code here...]

class MyDataReader:
  template_fields = ['path']

  def __init__(self, my_path):
    self.path = my_path

  # [additional code here...]

t = PythonOperator(
    task_id='transform_data',
    python_callable=transform_data
    op_args=[
      MyDataTransformer(MyDataReader('/tmp/{{ ds }}/my_file'))
    ],
    dag=dag)

You can pass custom options to the Jinja Environment when creating your DAG. One common usage is to avoid Jinja from dropping a trailing newline from a template string:

my_dag = DAG(dag_id='my-dag',
             jinja_environment_kwargs={
                  'keep_trailing_newline': True,
                  # some other jinja2 Environment options here
             })

See Jinja documentation to find all available options.

Exceptions

Airflow defines a number of exceptions; most of these are used internally, but a few are relevant to authors of custom operators or python callables called from PythonOperator tasks. Normally any exception raised from an execute method or python callable will either cause a task instance to fail if it is not configured to retry or has reached its limit on retry attempts, or to be marked as “up for retry”. A few exceptions can be used when different behavior is desired:

  • AirflowSkipException can be raised to set the state of the current task instance to “skipped”

  • AirflowFailException can be raised to set the state of the current task to “failed” regardless of whether there are any retry attempts remaining.

This example illustrates some possibilities

from airflow.exceptions import AirflowFailException, AirflowSkipException

def fetch_data():
    try:
        data = get_some_data(get_api_key())
        if not data:
            # Set state to skipped and do not retry
            # Downstream task behavior will be determined by trigger rules
            raise AirflowSkipException("No data available.")
    except Unauthorized:
        # If we retry, our api key will still be bad, so don't waste time retrying!
        # Set state to failed and move on
        raise AirflowFailException("Our api key is bad!")
    except TransientError:
        print("Looks like there was a blip.")
        # Raise the exception and let the task retry unless max attempts were reached
        raise
    handle(data)

task = PythonOperator(task_id="fetch_data", python_callable=fetch_data, retries=10)

Packaged DAGs

While often you will specify DAGs in a single .py file it might sometimes be required to combine a DAG and its dependencies. For example, you might want to combine several DAGs together to version them together or you might want to manage them together or you might need an extra module that is not available by default on the system you are running Airflow on. To allow this you can create a zip file that contains the DAG(s) in the root of the zip file and have the extra modules unpacked in directories.

For instance you can create a zip file that looks like this:

my_dag1.py
my_dag2.py
package1/__init__.py
package1/functions.py

Airflow will scan the zip file and try to load my_dag1.py and my_dag2.py. It will not go into subdirectories as these are considered to be potential packages.

In case you would like to add module dependencies to your DAG you basically would do the same, but then it is more suitable to use a virtualenv and pip.

virtualenv zip_dag
source zip_dag/bin/activate

mkdir zip_dag_contents
cd zip_dag_contents

pip install --install-option="--install-lib=$PWD" my_useful_package
cp ~/my_dag.py .

zip -r zip_dag.zip *

Note

the zip file will be inserted at the beginning of module search list (sys.path) and as such it will be available to any other code that resides within the same interpreter.

Note

packaged dags cannot be used with pickling turned on.

Note

packaged dags cannot contain dynamic libraries (eg. libz.so) these need to be available on the system if a module needs those. In other words only pure python modules can be packaged.

.airflowignore

A .airflowignore file specifies the directories or files in DAG_FOLDER or PLUGINS_FOLDER that Airflow should intentionally ignore. Each line in .airflowignore specifies a regular expression pattern, and directories or files whose names (not DAG id) match any of the patterns would be ignored (under the hood,``re.findall()`` is used to match the pattern). Overall it works like a .gitignore file. Use the # character to indicate a comment; all characters on a line following a # will be ignored.

.airflowignore file should be put in your DAG_FOLDER. For example, you can prepare a .airflowignore file with contents

project_a
tenant_[\d]

Then files like project_a_dag_1.py, TESTING_project_a.py, tenant_1.py, project_a/dag_1.py, and tenant_1/dag_1.py in your DAG_FOLDER would be ignored (If a directory’s name matches any of the patterns, this directory and all its subfolders would not be scanned by Airflow at all. This improves efficiency of DAG finding).

The scope of a .airflowignore file is the directory it is in plus all its subfolders. You can also prepare .airflowignore file for a subfolder in DAG_FOLDER and it would only be applicable for that subfolder.