The regular Breeze development tasks are available as top-level commands. Those tasks are most often used during the development, that's why they are available without any sub-command. More advanced commands are separated to sub-commands.
- Entering Breeze shell
- Selecting Backend
- Port Forwarding
- Building the documentation
- Running static checks
- Selecting files to run static checks on
- Starting Airflow
- Launching multiple terminals in the same environment
- Compiling www assets
- Breeze cleanup
- Database volumes in Breeze
- Running arbitrary commands in container
- Running Breeze with Metrics
- Stopping the environment
- Using local virtualenv environment in Your Host IDE
- Auto-generating migration files
This is the most often used feature of breeze. It simply allows to enter the shell inside the Breeze development environment (inside the Breeze container).
You can use additional breeze flags to choose your environment. You can specify a Python
version to use, and backend (the meta-data database). Thanks to that, with Breeze, you can recreate the same
environments as we have in matrix builds in the CI. See next chapter for backend selection.
For example, you can choose to run Python 3.8 tests with MySQL as backend and with mysql version 8 as follows:
breeze --python 3.8 --backend mysql --mysql-version 8.0Note
Note for Windows WSL2 users
You may find error messages:
Current context is now "..."
protocol not available
Error 1 returnedTry adding --builder=default to your command. For example:
breeze --builder=default --python 3.8 --backend mysql --mysql-version 8.0The choices you make are persisted in the ./.build/ cache directory so that next time when you use the
breeze script, it could use the values that were used previously. This way you do not have to specify
them when you run the script. You can delete the .build/ directory in case you want to restore the
default settings.
You can also run breeze with SKIP_SAVING_CHOICES to non-empty value and breeze invocation will not save
used cache value to cache - this is useful when you run non-interactive scripts with breeze shell and
want to - for example - force Python version used only for that execution without changing the Python version
that user used last time.
You can see which value of the parameters that can be stored persistently in cache marked with >VALUE<
in the help of the commands (for example in output of breeze config --help).
When you run breeze commands, you can additionally select which backend you want to use. Currently Airflow supports Sqlite, MySQL and Postgres as backends - MySQL and Postgres are supported in various versions.
You can choose which backend to use by adding --backend flag and additionally you can select version
of the backend, if you want to start a different version of backend (for example for --backend postgres
you can specify --postgres-version 13 to start Postgres 13). The --help command in breeze commands
will show you which backends are supported and which versions are available for each backend.
The choice you made for backend and version are sticky - the last used selection is cached in the
.build folder and next time you run any of the breeze commands that use backend the will use the
last selected backend and version.
Note
You can also (temporarily for the time of running a single command) override the backend version
used via BACKEND_VERSION environment variable. This is used mostly in CI where we have common way of
running tests for all backends and we want to specify different parameters. In order to override the
backend version, it has to be a valid version for the backend you are using. For example if you set
BACKEND_VERSION to 13 and you are using --backend postgres, Postgres 13 will be used, but
if you set BACKEND_VERSION to 8.0 and you are using --backend postgres, the last used Postgres
version will be used.
Breeze will inform you at startup which backend and version it is using:
When you run Airflow Breeze, the following ports are automatically forwarded:
* 12322 -> forwarded to Airflow ssh server -> airflow:22 * 28080 -> forwarded to Airflow webserver -> airflow:8080 * 25555 -> forwarded to Flower dashboard -> airflow:5555 * 25433 -> forwarded to Postgres database -> postgres:5432 * 23306 -> forwarded to MySQL database -> mysql:3306 * 26379 -> forwarded to Redis broker -> redis:6379
You can connect to these ports/databases using:
* ssh connection for remote debugging: ssh -p 12322 [email protected] pw: airflow * Webserver: http://127.0.0.1:28080 * Flower: http://127.0.0.1:25555 * Postgres: jdbc:postgresql://127.0.0.1:25433/airflow?user=postgres&password=airflow * Mysql: jdbc:mysql://127.0.0.1:23306/airflow?user=root * Redis: redis://127.0.0.1:26379/0
If you do not use start-airflow command, you can start the webserver manually with
the airflow webserver command if you want to run it. You can use tmux to multiply terminals.
You may need to create a user prior to running the webserver in order to log in.
This can be done with the following command:
airflow users create --role Admin --username admin --password admin --email [email protected] --firstname foo --lastname barFor databases, you need to run airflow db reset at least once (or run some tests) after you started
Airflow Breeze to get the database/tables created. You can connect to databases with IDE or any other
database client:
You can change the used host port numbers by setting appropriate environment variables:
SSH_PORTWEBSERVER_HOST_PORTPOSTGRES_HOST_PORTMYSQL_HOST_PORTMSSQL_HOST_PORTFLOWER_HOST_PORTREDIS_HOST_PORT
If you set these variables, next time when you enter the environment the new ports should be in effect.
To build documentation in Breeze, use the build-docs command:
breeze build-docsResults of the build can be found in the docs/_build folder.
The documentation build consists of three steps:
- verifying consistency of indexes
- building documentation
- spell checking
You can choose only one stage of the two by providing --spellcheck-only or --docs-only after
extra -- flag.
breeze build-docs --spellcheck-onlyThis process can take some time, so in order to make it shorter you can filter by package, using package
short provider id (might be multiple of them).
breeze build-docs <provider id> <provider id>or you can use package filter. The filters are glob pattern matching full package names and can be used to select more than one package with single filter.
breeze build-docs --package-filter apache-airflow-providers-*Often errors during documentation generation come from the docstrings of auto-api generated classes.
During the docs building auto-api generated files are stored in the docs/_api folder. This helps you
easily identify the location the problems with documentation originated from.
These are all available flags of build-docs command:
While you can use full name of doc package starting with apache-airflow-providers- in package filter,
You can use shorthand version - just take the remaining part and replace every dash("-") with
a dot(".").
Example:
If the provider name is apache-airflow-providers-cncf-kubernetes, it will be cncf.kubernetes.
Note: For building docs for apache-airflow-providers index, use apache-airflow-providers
as the short hand operator.
You can run static checks via Breeze. You can also run them via pre-commit command but with auto-completion Breeze makes it easier to run selective static checks. If you press <TAB> after the static-check and if you have auto-complete setup you should see auto-completable list of all checks available.
For example, this following command:
breeze static-checks --type mypy-airflowwill run mypy check for currently staged files inside airflow/ excluding providers.
Pre-commits run by default on staged changes that you have locally changed. It will run it on all the
files you run git add on and it will ignore any changes that you have modified but not staged.
If you want to run it on all your modified files you should add them with git add command.
With --all-files you can run static checks on all files in the repository. This is useful when you
want to be sure they will not fail in CI, or when you just rebased your changes and want to
re-run latest pre-commits on your changes, but it can take a long time (few minutes) to wait for the result.
breeze static-checks --type mypy-airflow --all-filesThe above will run mypy check for all files.
You can limit that by selecting specific files you want to run static checks on. You can do that by
specifying (can be multiple times) --file flag.
breeze static-checks --type mypy-airflow --file airflow/utils/code_utils.py --file airflow/utils/timeout.pyThe above will run mypy check for those to files (note: autocomplete should work for the file selection).
However, often you do not remember files you modified and you want to run checks for files that belong
to specific commits you already have in your branch. You can use breeze static check to run the checks
only on changed files you have already committed to your branch - either for specific commit, for last
commit, for all changes in your branch since you branched off from main or for specific range
of commits you choose.
breeze static-checks --type mypy-airflow --last-commitThe above will run mypy check for all files in the last commit in your branch.
breeze static-checks --type mypy-airflow --only-my-changesThe above will run mypy check for all commits in your branch which were added since you branched off from main.
breeze static-checks --type mypy-airflow --commit-ref 639483d998ecac64d0fef7c5aa4634414065f690The above will run mypy check for all files in the 639483d998ecac64d0fef7c5aa4634414065f690 commit.
Any commit-ish reference from Git will work here (branch, tag, short/long hash etc.)
breeze static-checks --type identity --verbose --from-ref HEAD^^^^ --to-ref HEADThe above will run the check for the last 4 commits in your branch. You can use any commit-ish references
in --from-ref and --to-ref flags.
These are all available flags of static-checks command:
Note
When you run static checks, some of the artifacts (mypy_cache) is stored in docker-compose volume
so that it can speed up static checks execution significantly. However, sometimes, the cache might
get broken, in which case you should run breeze down to clean up the cache.
Note
You cannot change Python version for static checks that are run within Breeze containers.
The --python flag has no effect for them. They are always run with lowest supported Python version.
The main reason is to keep consistency in the results of static checks and to make sure that
our code is fine when running the lowest supported version.
For testing Airflow you often want to start multiple components (in multiple terminals). Breeze has
built-in start-airflow command that start breeze container, launches multiple terminals using tmux
and launches all Airflow necessary components in those terminals.
When you are starting airflow from local sources, www asset compilation is automatically executed before.
breeze --python 3.8 --backend mysql start-airflowYou can also use it to start different executor.
breeze start-airflow --executor CeleryExecutorYou can also use it to start any released version of Airflow from PyPI with the
--use-airflow-version flag - useful for testing and looking at issues raised for specific version.
breeze start-airflow --python 3.8 --backend mysql --use-airflow-version 2.7.0When you are installing version from PyPI, it's also possible to specify extras that should be used
when installing Airflow - you can provide several extras separated by coma - for example to install
providers together with Airflow that you are installing. For example when you are using celery executor
in Airflow 2.7.0+ you need to add celery extra.
breeze start-airflow --use-airflow-version 2.7.0 --executor CeleryExecutor --airflow-extras celeryThese are all available flags of start-airflow command:
Often if you want to run full airflow in the Breeze environment you need to launch multiple terminals and
run airflow webserver, airflow scheduler, airflow worker in separate terminals.
This can be achieved either via tmux or via exec-ing into the running container from the host. Tmux
is installed inside the container and you can launch it with tmux command. Tmux provides you with the
capability of creating multiple virtual terminals and multiplex between them. More about tmux can be
found at tmux GitHub wiki page . Tmux has several useful shortcuts
that allow you to split the terminals, open new tabs etc - it's pretty useful to learn it.
Another way is to exec into Breeze terminal from the host's terminal. Often you can
have multiple terminals in the host (Linux/MacOS/WSL2 on Windows) and you can simply use those terminals
to enter the running container. It's as easy as launching breeze exec while you already started the
Breeze environment. You will be dropped into bash and environment variables will be read in the same
way as when you enter the environment. You can do it multiple times and open as many terminals as you need.
These are all available flags of exec command:
Airflow webserver needs to prepare www assets - compiled with node and yarn. The compile-www-assets
command takes care about it. This is needed when you want to run webserver inside of the breeze.
Sometimes you need to cleanup your docker environment (and it is recommended you do that regularly). There are several reasons why you might want to do that.
Breeze uses docker images heavily and those images are rebuild periodically and might leave dangling, unused
images in docker cache. This might cause extra disk usage. Also running various docker compose commands
(for example running tests with breeze testing tests) might create additional docker networks that might
prevent new networks from being created. Those networks are not removed automatically by docker-compose.
Also Breeze uses it's own cache to keep information about all images.
All those unused images, networks and cache can be removed by running breeze cleanup command. By default
it will not remove the most recent images that you might need to run breeze commands, but you
can also remove those breeze images to clean-up everything by adding --all command (note that you will
need to build the images again from scratch - pulling from the registry might take a while).
Breeze will ask you to confirm each step, unless you specify --answer yes flag.
These are all available flags of cleanup command:
Breeze keeps data for all it's integration in named docker volumes. Each backend and integration
keeps data in their own volume. Those volumes are persisted until breeze down command.
You can also preserve the volumes by adding flag --preserve-volumes when you run the command.
Then, next time when you start Breeze, it will have the data pre-populated.
These are all available flags of down command:
More sophisticated usages of the breeze shell is using the breeze shell command - it has more parameters
and you can also use it to execute arbitrary commands inside the container.
breeze shell "ls -la"Those are all available flags of shell command:
You can launch an instance of Breeze pre-configured to emit StatsD metrics using
breeze start-airflow --integration statsd. This will launch an Airflow webserver
within the Breeze environment as well as containers running StatsD, Prometheus, and
Grafana. The integration configures the "Targets" in Prometheus, the "Datasources" in
Grafana, and includes a default dashboard in Grafana.
When you run Airflow Breeze with this integration, in addition to the standard ports (See "Port Forwarding" below), the following are also automatically forwarded:
- 29102 -> forwarded to StatsD Exporter -> breeze-statsd-exporter:9102
- 29090 -> forwarded to Prometheus -> breeze-prometheus:9090
- 23000 -> forwarded to Grafana -> breeze-grafana:3000
You can connect to these ports/databases using:
- StatsD Metrics: http://127.0.0.1:29102/metrics
- Prometheus Targets: http://127.0.0.1:29090/targets
- Grafana Dashboards: http://127.0.0.1:23000/dashboards
[Work in Progress] NOTE: This will launch the stack as described below but Airflow integration is still a Work in Progress. This should be considered experimental and likely to change by the time Airflow fully supports emitting metrics via OpenTelemetry.
You can launch an instance of Breeze pre-configured to emit OpenTelemetry metrics
using breeze start-airflow --integration otel. This will launch Airflow within
the Breeze environment as well as containers running OpenTelemetry-Collector,
Prometheus, and Grafana. The integration handles all configuration of the
"Targets" in Prometheus and the "Datasources" in Grafana, so it is ready to use.
When you run Airflow Breeze with this integration, in addition to the standard ports (See "Port Forwarding" below), the following are also automatically forwarded:
- 28889 -> forwarded to OpenTelemetry Collector -> breeze-otel-collector:8889
- 29090 -> forwarded to Prometheus -> breeze-prometheus:9090
- 23000 -> forwarded to Grafana -> breeze-grafana:3000
You can connect to these ports using:
- OpenTelemetry Collector: http://127.0.0.1:28889/metrics
- Prometheus Targets: http://127.0.0.1:29090/targets
- Grafana Dashboards: http://127.0.0.1:23000/dashboards
You can launch an instance of Breeze pre-configured to emit OpenLineage metrics using
breeze start-airflow --integration openlineage. This will launch an Airflow webserver
within the Breeze environment as well as containers running a [Marquez](https://marquezproject.ai/)
webserver and API server.
When you run Airflow Breeze with this integration, in addition to the standard ports (See "Port Forwarding" below), the following are also automatically forwarded:
- MARQUEZ_API_HOST_PORT (default 25000) -> forwarded to Marquez API -> marquez:5000
- MARQUEZ_API_ADMIN_HOST_PORT (default 25001) -> forwarded to Marquez Admin API -> marquez:5001
- MARQUEZ_HOST_PORT (default 23100) -> forwarded to Marquez -> marquez_web:3000
You can connect to these services using:
- Marquez Webserver: http://127.0.0.1:23100
- Marquez API: http://127.0.0.1:25000/api/v1
- Marquez Admin API: http://127.0.0.1:25001
Make sure to substitute the port numbers if you have customized them via the above env vars.
After starting up, the environment runs in the background and takes quite some memory which you might want to free for other things you are running on your host.
You can always stop it via:
breeze downThese are all available flags of down command:
You can set up your host IDE (for example, IntelliJ's PyCharm/Idea) to work with Breeze and benefit from all the features provided by your IDE, such as local and remote debugging, language auto-completion, documentation support, etc.
To use your host IDE with Breeze:
Create a local virtual environment:
You can use any of the following wrappers to create and manage your virtual environments: pyenv, pyenv-virtualenv, or virtualenvwrapper.
Use the right command to activate the virtualenv (
workonif you use virtualenvwrapper orpyenv activateif you use pyenv.Initialize the created local virtualenv:
./scripts/tools/initialize_virtualenv.pyWarning
Make sure that you use the right Python version in this command - matching the Python version you have in your local virtualenv. If you don't, you will get strange conflicts.
- Select the virtualenv you created as the project's default virtualenv in your IDE.
Note that you can also use the local virtualenv for Airflow development without Breeze. This is a lightweight solution that has its own limitations.
More details on using the local virtualenv are available in the Local Virtualenv.
After making changes in the ORM models, you need to generate migration files. You can do this by running the following command:
breeze generate-migration-file -m "Your migration message"This command will generate a migration file in the airflow/migrations/versions directory.
These are all available flags of generate-migration-file command:
Next step: Follow the Troubleshooting guide to troubleshoot your Breeze environment.