Implementation:Apache Airflow BaseExecutor Interface

Knowledge Sources	Apache Airflow
Domains	Execution, Python_API
Last Updated	2026-02-08 00:00 GMT

Overview

Concrete tool for dispatching task execution provided by the BaseExecutor, LocalExecutor, and ExecutorLoader classes.

Description

BaseExecutor defines the abstract interface all executors must implement. LocalExecutor extends it with multiprocessing-based execution using SimpleQueue for task distribution. ExecutorLoader is a utility class that dynamically resolves and loads the configured executor by name.

Usage

Configure the executor in airflow.cfg under [core].executor. Custom executors must subclass BaseExecutor and implement the required methods.

Code Reference

Source Location

Repository: Apache Airflow
File: airflow-core/src/airflow/executors/base_executor.py
Lines: L136-626

Signature

class BaseExecutor(LoggingMixin):
    def __init__(
        self,
        parallelism: int = PARALLELISM,
        team_name: str | None = None,
    ):
        ...

    # Key methods:
    def execute_async(self, key, command, queue=None, executor_config=None): ...
    def sync(self): ...
    def heartbeat(self): ...
    def trigger_tasks(self, open_slots): ...
    def end(self): ...

LocalExecutor:

class LocalExecutor(BaseExecutor):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        # Uses multiprocessing.SimpleQueue for task distribution
        # Spawns worker processes up to parallelism limit

ExecutorLoader:

class ExecutorLoader:
    # Maps executor names to classes
    # Supports: LocalExecutor, CeleryExecutor, KubernetesExecutor, etc.

    @classmethod
    def _get_executor_names(cls, validate_teams: bool = True) -> list[ExecutorName]: ...

    @classmethod
    def load_executor(cls, executor_name: str) -> BaseExecutor: ...

Import

from airflow.executors.base_executor import BaseExecutor
from airflow.executors.local_executor import LocalExecutor
from airflow.executors.executor_loader import ExecutorLoader

I/O Contract

Inputs

Name	Type	Required	Description
parallelism	int	No	Max concurrent tasks (default from config)
team_name	str or None	No	Multi-team executor support
TaskInstances	Scheduled TIs	Yes	Tasks queued by the scheduler

Outputs

Name	Type	Description
event_buffer	dict[TaskInstanceKey, EventBufferValueType]	Completion states for finished tasks
running	set[TaskInstanceKey]	Currently executing task set

Usage Examples

Configuration

# airflow.cfg
# [core]
# executor = LocalExecutor

# Custom executor
class MyExecutor(BaseExecutor):
    def execute_async(self, key, command, queue=None, executor_config=None):
        # Custom execution logic
        ...

    def sync(self):
        # Check running task status
        ...

Related Pages

Implements Principle

Principle:Apache_Airflow_Executor_Dispatch

Requires Environment

Environment:Apache_Airflow_Python_Runtime_Environment

Uses Heuristic

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment