Apache Airflow
Apache Airflow (or simply Airflow) is a platform to programmatically author, schedule, and monitor workflows.
When workflows are defined as code, they become more maintainable, versionable, testable, and collaborative.
Use Airflow to author workflows as directed acyclic graphs (DAGs) of tasks. The Airflow scheduler executes your tasks on an array of workers while following the specified dependencies. Rich command line utilities make performing complex surgeries on DAGs a snap. The rich user interface makes it easy to visualize pipelines running in production, monitor progress, and troubleshoot issues when needed.
Table of contents
- Project Focus
- Principles
- Requirements
- Getting started
- Installing from PyPI
- Official source code
- Convenience packages
- User Interface
- Semantic versioning
- Version Life Cycle
- Support for Python and Kubernetes versions
- Base OS support for reference Airflow images
- Approach to dependencies of Airflow
- Release process for Providers
- Contributing
- Who uses Apache Airflow?
- Who Maintains Apache Airflow?
- Can I use the Apache Airflow logo in my presentation?
- Airflow merchandise
- Links
- Sponsors
Project Focus
Airflow works best with workflows that are mostly static and slowly changing. When the DAG structure is similar from one run to the next, it clarifies the unit of work and continuity. Other similar projects include Luigi, Oozie and Azkaban.
Airflow is commonly used to process data, but has the opinion that tasks should ideally be idempotent (i.e., results of the task will be the same, and will not create duplicated data in a destination system), and should not pass large quantities of data from one task to the next (though tasks can pass metadata using Airflow's XCom feature). For high-volume, data-intensive tasks, a best practice is to delegate to external services specializing in that type of work.
Airflow is not a streaming solution, but it is often used to process real-time data, pulling data off streams in batches.
Principles
- Dynamic: Airflow pipelines are configuration as code (Python), allowing for dynamic pipeline generation. This allows for writing code that instantiates pipelines dynamically.
- Extensible: Easily define your own operators, executors and extend the library so that it fits the level of abstraction that suits your environment.
- Elegant: Airflow pipelines are lean and explicit. Parameterizing your scripts is built into the core of Airflow using the powerful Jinja templating engine.
- Scalable: Airflow has a modular architecture and uses a message queue to orchestrate an arbitrary number of workers.
Requirements
Apache Airflow is tested with:
Main version (dev) | Stable version (2.5.3) | |
---|---|---|
Python | 3.7, 3.8, 3.9, 3.10 | 3.7, 3.8, 3.9, 3.10 |
Platform | AMD64/ARM64(*) | AMD64/ARM64(*) |
Kubernetes | 1.23, 1.24, 1.25, 1.26 | 1.21, 1.22, 1.23, 1.24, 1.25 |
PostgreSQL | 11, 12, 13, 14, 15 | 11, 12, 13, 14, 15 |
MySQL | 5.7, 8 | 5.7, 8 |
SQLite | 3.15.0+ | 3.15.0+ |
MSSQL | 2017(*), 2019(*) | 2017(*), 2019(*) |
* Experimental
Note: MySQL 5.x versions are unable to or have limitations with running multiple schedulers -- please see the Scheduler docs. MariaDB is not tested/recommended.
Note: SQLite is used in Airflow tests. Do not use it in production. We recommend using the latest stable version of SQLite for local development.
Note: Airflow currently can be run on POSIX-compliant Operating Systems. For development it is regularly
tested on fairly modern Linux Distros and recent versions of MacOS.
On Windows you can run it via WSL2 (Windows Subsystem for Linux 2) or via Linux Containers.
The work to add Windows support is tracked via #10388 but
it is not a high priority. You should only use Linux-based distros as "Production" execution environment
as this is the only environment that is supported. The only distro that is used in our CI tests and that
is used in the Community managed DockerHub image is
Debian Bullseye
.
Getting started
Visit the official Airflow website documentation (latest stable release) for help with installing Airflow, getting started, or walking through a more complete tutorial.
Note: If you're looking for documentation for the main branch (latest development branch): you can find it on s.apache.org/airflow-docs.
For more information on Airflow Improvement Proposals (AIPs), visit the Airflow Wiki.
Documentation for dependent projects like provider packages, Docker image, Helm Chart, you'll find it in the documentation index.
Installing from PyPI
We publish Apache Airflow as apache-airflow
package in PyPI. Installing it however might be sometimes tricky
because Airflow is a bit of both a library and application. Libraries usually keep their dependencies open, and
applications usually pin them, but we should do neither and both simultaneously. We decided to keep
our dependencies as open as possible (in setup.py
) so users can install different versions of libraries
if needed. This means that pip install apache-airflow
will not work from time to time or will
produce unusable Airflow installation.
To have repeatable installation, however, we keep a set of "known-to-be-working" constraint
files in the orphan constraints-main
and constraints-2-0
branches. We keep those "known-to-be-working"
constraints files separately per major/minor Python version.
You can use them as constraint files when installing Airflow from PyPI. Note that you have to specify
correct Airflow tag/version/branch and Python versions in the URL.
- Installing just Airflow:
Note: Only
pip
installation is currently officially supported.
While it is possible to install Airflow with tools like Poetry or
pip-tools, they do not share the same workflow as
pip
- especially when it comes to constraint vs. requirements management.
Installing via Poetry
or pip-tools
is not currently supported.
If you wish to install Airflow using those tools, you should use the constraint files and convert them to the appropriate format and workflow that your tool requires.
pip install 'apache-airflow==2.5.3' \
--constraint "https://raw.githubusercontent.com/apache/airflow/constraints-2.5.3/constraints-3.7.txt"
- Installing with extras (i.e., postgres, google)
pip install 'apache-airflow[postgres,google]==2.5.3' \
--constraint "https://raw.githubusercontent.com/apache/airflow/constraints-2.5.3/constraints-3.7.txt"
For information on installing provider packages, check providers.
Official source code
Apache Airflow is an Apache Software Foundation (ASF) project, and our official source code releases:
- Follow the ASF Release Policy
- Can be downloaded from the ASF Distribution Directory
- Are cryptographically signed by the release manager
- Are officially voted on by the PMC members during the Release Approval Process
Following the ASF rules, the source packages released must be sufficient for a user to build and test the release provided they have access to the appropriate platform and tools.
Convenience packages
There are other ways of installing and using Airflow. Those are "convenience" methods - they are
not "official releases" as stated by the ASF Release Policy
, but they can be used by the users
who do not want to build the software themselves.
Those are - in the order of most common ways people install Airflow:
- PyPI releases to install Airflow using standard
pip
tool - Docker Images to install airflow via
docker
tool, use them in Kubernetes, Helm Charts,docker-compose
,docker swarm
, etc. You can read more about using, customising, and extending the images in the Latest docs, and learn details on the internals in the IMAGES.rst document. - Tags in GitHub to retrieve the git project sources that were used to generate official source packages via git
All those artifacts are not official releases, but they are prepared using officially released sources. Some of those artifacts are "development" or "pre-release" ones, and they are clearly marked as such following the ASF Policy.
User Interface
-
DAGs: Overview of all DAGs in your environment.
-
Grid: Grid representation of a DAG that spans across time.
-
Graph: Visualization of a DAG's dependencies and their current status for a specific run.
-
Task Duration: Total time spent on different tasks over time.
-
Gantt: Duration and overlap of a DAG.
-
Code: Quick way to view source code of a DAG.
Semantic versioning
As of Airflow 2.0.0, we support a strict SemVer approach for all packages released.
There are few specific rules that we agreed to that define details of versioning of the different packages:
- Airflow: SemVer rules apply to core airflow only (excludes any changes to providers). Changing limits for versions of Airflow dependencies is not a breaking change on its own.
- Airflow Providers: SemVer rules apply to changes in the particular provider's code only.
SemVer MAJOR and MINOR versions for the packages are independent of the Airflow version.
For example,
google 4.1.0
andamazon 3.0.3
providers can happily be installed withAirflow 2.1.2
. If there are limits of cross-dependencies between providers and Airflow packages, they are present in providers asinstall_requires
limitations. We aim to keep backwards compatibility of providers with all previously released Airflow 2 versions but there will sometimes be breaking changes that might make some, or all providers, have minimum Airflow version specified. Change of that minimum supported Airflow version is a breaking change for provider because installing the new provider might automatically upgrade Airflow (which might be an undesired side effect of upgrading provider). - Airflow Helm Chart: SemVer rules apply to changes in the chart only. SemVer MAJOR and MINOR versions for the chart are independent from the Airflow version. We aim to keep backwards compatibility of the Helm Chart with all released Airflow 2 versions, but some new features might only work starting from specific Airflow releases. We might however limit the Helm Chart to depend on minimal Airflow version.
- Airflow API clients: SemVer MAJOR and MINOR versions follow MAJOR and MINOR versions of Airflow. The first MAJOR or MINOR X.Y.0 release of Airflow should always be followed by X.Y.0 release of all clients. An airflow PATCH X.Y.Z release can be followed by a PATCH release of API clients, only if this PATCH is relevant to the clients. The clients then can release their own PATCH releases with bugfixes, independently of Airflow PATCH releases. As a consequence, each API client will have its own PATCH version that may or may not be in sync with the Airflow PATCH version. For a specific MAJOR/MINOR Airflow version, users should favor the latest PATCH version of clients independently of their Airflow PATCH version.
Version Life Cycle
Apache Airflow version life cycle:
Version | Current Patch/Minor | State | First Release | Limited Support | EOL/Terminated |
---|---|---|---|---|---|
2 | 2.5.3 | Supported | Dec 17, 2020 | TBD | TBD |
1.10 | 1.10.15 | EOL | Aug 27, 2018 | Dec 17, 2020 | June 17, 2021 |
1.9 | 1.9.0 | EOL | Jan 03, 2018 | Aug 27, 2018 | Aug 27, 2018 |
1.8 | 1.8.2 | EOL | Mar 19, 2017 | Jan 03, 2018 | Jan 03, 2018 |
1.7 | 1.7.1.2 | EOL | Mar 28, 2016 | Mar 19, 2017 | Mar 19, 2017 |
Limited support versions will be supported with security and critical bug fix only. EOL versions will not get any fixes nor support. We always recommend that all users run the latest available minor release for whatever major version is in use. We highly recommend upgrading to the latest Airflow major release at the earliest convenient time and before the EOL date.
Support for Python and Kubernetes versions
As of Airflow 2.0, we agreed to certain rules we follow for Python and Kubernetes support. They are based on the official release schedule of Python and Kubernetes, nicely summarized in the Python Developer's Guide and Kubernetes version skew policy.
-
We drop support for Python and Kubernetes versions when they reach EOL. Except for Kubernetes, a version stays supported by Airflow if two major cloud providers still provide support for it. We drop support for those EOL versions in main right after EOL date, and it is effectively removed when we release the first new MINOR (Or MAJOR if there is no new MINOR version) of Airflow. For example, for Python 3.7 it means that we will drop support in main right after 27.06.2023, and the first MAJOR or MINOR version of Airflow released after will not have it.
-
The "oldest" supported version of Python/Kubernetes is the default one until we decide to switch to later version. "Default" is only meaningful in terms of "smoke tests" in CI PRs, which are run using this default version and the default reference image available. Currently
apache/airflow:latest
andapache/airflow:2.5.3
images are Python 3.7 images. This means that default reference image will become the default at the time when we start preparing for dropping 3.7 support which is few months before the end of life for Python 3.7. -
We support a new version of Python/Kubernetes in main after they are officially released, as soon as we make them work in our CI pipeline (which might not be immediate due to dependencies catching up with new versions of Python mostly) we release new images/support in Airflow based on the working CI setup.
Base OS support for reference Airflow images
The Airflow Community provides conveniently packaged container images that are published whenever we publish an Apache Airflow release. Those images contain:
- Base OS with necessary packages to install Airflow (stable Debian OS)
- Base Python installation in versions supported at the time of release for the MINOR version of Airflow released (so there could be different versions for 2.3 and 2.2 line for example)
- Libraries required to connect to supported Databases (again the set of databases supported depends on the MINOR version of Airflow.
- Predefined set of popular providers (for details see the Dockerfile).
- Possibility of building your own, custom image where the user can choose their own set of providers and libraries (see Building the image)
- In the future Airflow might also support a "slim" version without providers nor database clients installed
The version of the base OS image is the stable version of Debian. Airflow supports using all currently active
stable versions - as soon as all Airflow dependencies support building, and we set up the CI pipeline for
building and testing the OS version. Approximately 6 months before the end-of-life of a previous stable
version of the OS, Airflow switches the images released to use the latest supported version of the OS.
For example since Debian Buster
end-of-life was August 2022, Airflow switched the images in main
branch
to use Debian Bullseye
in February/March 2022. The version was used in the next MINOR release after
the switch happened. In case of the Bullseye switch - 2.3.0 version used Debian Bullseye
.
The images released in the previous MINOR version continue to use the version that all other releases
for the MINOR version used.
Support for Debian Buster
image was dropped in August 2022 completely and everyone is expected to
stop building their images using Debian Buster
.
Users will continue to be able to build their images using stable Debian releases until the end of life and
building and verifying of the images happens in our CI but no unit tests were executed using this image in
the main
branch.
Approach to dependencies of Airflow
Airflow has a lot of dependencies - direct and transitive, also Airflow is both - library and application,
therefore our policies to dependencies has to include both - stability of installation of application,
but also ability to install newer version of dependencies for those users who develop DAGs. We developed
the approach where constraints
are used to make sure airflow can be installed in a repeatable way, while
we do not limit our users to upgrade most of the dependencies. As a result we decided not to upper-bound
version of Airflow dependencies by default, unless we have good reasons to believe upper-bounding them is
needed because of importance of the dependency as well as risk it involves to upgrade specific dependency.
We also upper-bound the dependencies that we know cause problems.
The constraint mechanism of ours takes care about finding and upgrading all the non-upper bound dependencies
automatically (providing that all the tests pass). Our main
build failures will indicate in case there
are versions of dependencies that break our tests - indicating that we should either upper-bind them or
that we should fix our code/tests to account for the upstream changes from those dependencies.
Whenever we upper-bound such a dependency, we should always comment why we are doing it - i.e. we should have a good reason why dependency is upper-bound. And we should also mention what is the condition to remove the binding.
Approach for dependencies for Airflow Core
Those extras
and providers
dependencies are maintained in setup.cfg
.
There are few dependencies that we decided are important enough to upper-bound them by default, as they are known to follow predictable versioning scheme, and we know that new versions of those are very likely to bring breaking changes. We commit to regularly review and attempt to upgrade to the newer versions of the dependencies as they are released, but this is manual process.
The important dependencies are:
SQLAlchemy
: upper-bound to specific MINOR version (SQLAlchemy is known to remove deprecations and introduce breaking changes especially that support for different Databases varies and changes at various speed (example: SQLAlchemy 1.4 broke MSSQL integration for Airflow)Alembic
: it is important to handle our migrations in predictable and performant way. It is developed together with SQLAlchemy. Our experience with Alembic is that it very stable in MINOR versionFlask
: We are using Flask as the back-bone of our web UI and API. We know major version of Flask are very likely to introduce breaking changes across those so limiting it to MAJOR version makes sensewerkzeug
: the library is known to cause problems in new versions. It is tightly coupled with Flask libraries, and we should update them togethercelery
: Celery is crucial component of Airflow as it used for CeleryExecutor (and similar). Celery follows SemVer, so we should upper-bound it to the next MAJOR version. Also when we bump the upper version of the library, we should make sure Celery Provider minimum Airflow version is updated).kubernetes
: Kubernetes is a crucial component of Airflow as it is used for the KubernetesExecutor (and similar). Kubernetes Python library follows SemVer, so we should upper-bound it to the next MAJOR version. Also when we bump the upper version of the library, we should make sure Kubernetes Provider minimum Airflow version is updated.
Approach for dependencies in Airflow Providers and extras
Those extras
and providers
dependencies are maintained in provider.yaml
of each provider.
By default, we should not upper-bound dependencies for providers, however each provider's maintainer might decide to add additional limits (and justify them with comment)
Release process for Providers
Minimum supported version of Airflow
Providers released by the community (with roughly monthly cadence) have
limitation of a minimum supported version of Airflow. The minimum version of
Airflow is the MINOR
version (2.2, 2.3 etc.) indicating that the providers
might use features that appeared in this release. The default support timespan
for the minimum version of Airflow (there could be justified exceptions) is
that we increase the minimum Airflow version, when 12 months passed since the
first release for the MINOR version of Airflow.
For example this means that by default we upgrade the minimum version of Airflow supported by providers
to 2.4.0 in the first Provider's release after 30th of April 2023. The 30th of April 2022 is the date when the
first PATCHLEVEL
of 2.3 (2.3.0) has been released.
When we increase the minimum Airflow version, this is not a reason to bump MAJOR
version of the providers
(unless there are other breaking changes in the provider). The reason for that is that people who use
older version of Airflow will not be able to use that provider (so it is not a breaking change for them)
and for people who are using supported version of Airflow this is not a breaking change on its own - they
will be able to use the new version without breaking their workflows. When we upgraded min-version to
2.2+, our approach was different but as of 2.3+ upgrade (November 2022) we only bump MINOR
version of the
provider when we increase minimum Airflow version.
Mixed governance model
Providers are often connected with some stakeholders that are vitally interested in maintaining backwards compatibilities in their integrations (for example cloud providers, or specific service providers). But, we are also bound with the Apache Software Foundation release policy which describes who releases, and how to release the ASF software. The provider's governance model is something we name "mixed governance" - where we follow the release policies, while the burden of maintaining and testing the cherry-picked versions is on those who commit to perform the cherry-picks and make PRs to older branches.
The "mixed governance" (optional, per-provider) means that:
- The Airflow Community and release manager decide when to release those providers. This is fully managed by the community and the usual release-management process following the Apache Software Foundation release policy
- The contributors (who might or might not be direct stakeholders in the provider) will carry the burden of cherry-picking and testing the older versions of providers.
- There is no "selection" and acceptance process to determine which version of the provider is released. It is determined by the actions of contributors raising the PR with cherry-picked changes and it follows the usual PR review process where maintainer approves (or not) and merges (or not) such PR. Simply speaking - the completed action of cherry-picking and testing the older version of the provider make it eligible to be released. Unless there is someone who volunteers and perform the cherry-picking and testing, the provider is not released.
- Branches to raise PR against are created when a contributor commits to perform the cherry-picking (as a comment in PR to cherry-pick for example)
Usually, community effort is focused on the most recent version of each provider. The community approach is that we should rather aggressively remove deprecations in "major" versions of the providers - whenever there is an opportunity to increase major version of a provider, we attempt to remove all deprecations. However, sometimes there is a contributor (who might or might not represent stakeholder), willing to make their effort on cherry-picking and testing the non-breaking changes to a selected, previous major branch of the provider. This results in releasing at most two versions of a provider at a time:
- potentially breaking "latest" major version
- selected past major version with non-breaking changes applied by the contributor
Cherry-picking such changes follows the same process for releasing Airflow patch-level releases for a previous minor Airflow version. Usually such cherry-picking is done when there is an important bugfix and the latest version contains breaking changes that are not coupled with the bugfix. Releasing them together in the latest version of the provider effectively couples them, and therefore they're released separately. The cherry-picked changes have to be merged by the committer following the usual rules of the community.
There is no obligation to cherry-pick and release older versions of the providers. The community continues to release such older versions of the providers for as long as there is an effort of the contributors to perform the cherry-picks and carry-on testing of the older provider version.
The availability of stakeholder that can manage "service-oriented" maintenance and agrees to such a responsibility, will also drive our willingness to accept future, new providers to become community managed.
Suspending releases for providers
In case a provider is found to require old dependencies that are not compatible with upcoming versions of the Apache Airflow or with newer dependencies required by other providers, the provider's release process can be suspended.
This means:
- The provider's status is set to "suspended"
- No new releases of the provider will be made until the problem with dependencies is solved
- Sources of the provider remain in the repository for now (in the future we might add process to remove them)
- No new changes will be accepted for the provider (other than the ones that fix the dependencies)
- The provider will be removed from the list of Apache Airflow extras in the next Airflow release (including patch-level release if it is possible/easy to cherry-pick the suspension change)
- Tests of the provider will not be run on our CI (in main branch)
- Dependencies of the provider will not be installed in our main branch CI image nor included in constraints
- We can still decide to apply security fixes to released providers - by adding fixes to the main branch but cherry-picking, testing and releasing them in the patch-level branch of the provider similar to the mixed governance model described above.
The suspension may be triggered by any committer after the following criteria are met:
- The maintainers of dependencies of the provider are notified about the issue and are given a reasonable time to resolve it (at least 1 week)
- Other options to resolve the issue have been exhausted and there are good reasons for upgrading the old dependencies in question
- Explanation why we need to suspend the provider is stated in a public discussion in the devlist. Followed by LAZY CONSENSUS or VOTE (with the majority of the voters agreeing that we should suspend the provider)
The suspension will be lifted when the dependencies of the provider are made compatible with the Apache Airflow and with other providers.
Contributing
Want to help build Apache Airflow? Check out our contributing documentation.
Official Docker (container) images for Apache Airflow are described in IMAGES.rst.
Who uses Apache Airflow?
More than 400 organizations are using Apache Airflow in the wild.
Who Maintains Apache Airflow?
Airflow is the work of the community, but the core committers/maintainers are responsible for reviewing and merging PRs as well as steering conversations around new feature requests. If you would like to become a maintainer, please review the Apache Airflow committer requirements.
Can I use the Apache Airflow logo in my presentation?
Yes! Be sure to abide by the Apache Foundation trademark policies and the Apache Airflow Brandbook. The most up to date logos are found in this repo and on the Apache Software Foundation website.
Airflow merchandise
If you would love to have Apache Airflow stickers, t-shirt, etc. then check out Redbubble Shop.
Links
Sponsors
The CI infrastructure for Apache Airflow has been sponsored by: