galaxy.webapps.galaxy package

Subpackages

Submodules

galaxy.webapps.galaxy.buildapp module

Provides factory methods to assemble the Galaxy web application

class galaxy.webapps.galaxy.buildapp.GalaxyWebApplication(galaxy_app: MinimalApp, session_cookie: str = 'galaxysession', name: str | None = None)[source]

Bases: WebApplication

injection_aware: bool = True

__init__(galaxy_app: MinimalApp, session_cookie: str = 'galaxysession', name: str | None = None) → None[source]: Create a new web application object. To actually connect some controllers use add_controller and add_route. Call finalize_config when all controllers and routes have been added and __call__ to handle a request (WSGI style).

galaxy.webapps.galaxy.buildapp.app_factory(*args, **kwargs)[source]: Return a wsgi application serving the root object

galaxy.webapps.galaxy.buildapp.app_pair(global_conf, load_app_kwds=None, wsgi_preflight=True, **kwargs)[source]

Return a wsgi application serving the root object and the Galaxy application.

When creating an app for asgi, set wsgi_preflight to False to allow FastAPI middleware to handle CORS options, etc..

galaxy.webapps.galaxy.buildapp.postfork_setup()[source]

galaxy.webapps.galaxy.buildapp.populate_api_routes(webapp, app)[source]

galaxy.webapps.galaxy.buildapp.wrap_in_middleware(app, global_conf, application_stack, **local_conf)[source]: Based on the configuration wrap app in a set of common and useful middleware.

galaxy.webapps.galaxy.fast_app module

galaxy.webapps.galaxy.fast_app.add_galaxy_middleware(app: FastAPI, gx_app)[source]

galaxy.webapps.galaxy.fast_app.initialize_fast_app(gx_wsgi_webapp, gx_app)[source]: Build the FastAPI app that fronts the Galaxy web server.

galaxy.webapps.galaxy.fast_factory module

Module containing factory class for building uvicorn app for Galaxy.

Information on uvicorn, its various settings, and how to invoke it can be found at https://www.uvicorn.org/.

Galaxy can be launched with uvicorn using the following invocation:

uvicorn --app-dir lib --factory galaxy.webapps.galaxy.fast_factory:factory

Use the environment variable GALAXY_CONFIG_FILE to specify a Galaxy configuration file. Galaxy configuration can be loading from a YAML or an .ini file (reads app:main currently but can be overridden with GALAXY_CONFIG_SECTION).

GALAXY_CONFIG_FILE=config/galaxy.yml uvicorn --app-dir lib --factory galaxy.webapps.galaxy.fast_factory:factory

Note

Information on additional ways to configure uvicorn can be found at https://www.uvicorn.org/.

Warning

If an ini file is supplied via GALAXY_CONFIG_FILE, the server properties such as host and port are not read from the file like older forms of configuring Galaxy.

Gunicorn is a server with more complex management options.

This factory function can be executed as a uvicorn worker managed with gunicorn with the following command-line.

gunicorn 'galaxy.webapps.galaxy.fast_factory:factory()' --env GALAXY_CONFIG_FILE=config/galaxy.ini --pythonpath lib -w 4 -k uvicorn.workers.UvicornWorker --config lib/galaxy/web_stack/gunicorn_config.py

class galaxy.webapps.galaxy.fast_factory.GalaxyWebApp(galaxy_app: UniverseApplication, wsgi_app: Any, asgi_app: FastAPI)[source]

Bases: object

The application objects assembled for a Galaxy web process.

galaxy_app: UniverseApplication

wsgi_app: Any

asgi_app: FastAPI

__init__(galaxy_app: UniverseApplication, wsgi_app: Any, asgi_app: FastAPI) → None

galaxy.webapps.galaxy.fast_factory.build_galaxy_web_app(galaxy_config: ~collections.abc.Mapping[str, ~typing.Any] | None = None, *, global_conf: ~collections.abc.Mapping[str, ~typing.Any] | None = None, load_app_kwds: ~collections.abc.Mapping[str, ~typing.Any] | None = None, wsgi_preflight: bool = False, register_shutdown_at_exit: bool = True, init_fast_app: ~collections.abc.Callable[[~typing.Any, ~galaxy.app.UniverseApplication], ~fastapi.applications.FastAPI] = <function initialize_fast_app>) → GalaxyWebApp[source]

Build Galaxy’s application objects from programmatic configuration.

Unlike the uvicorn factory below, this entry point does not require a configuration file or environment-variable setup. It also returns the underlying Galaxy application so embedding callers can own its lifecycle. Application construction exceptions are allowed to propagate to the caller.

galaxy.webapps.galaxy.fast_factory.factory()[source]

galaxy.webapps.galaxy.workers module

Gunicorn worker class used to run the Galaxy ASGI application.

uvicorn’s gunicorn worker makes two choices that together turn an ordinary worker recycle (--max-requests) into an unexplained core dump:

Server.shutdown() waits indefinitely for in-flight requests, because timeout_graceful_shutdown is never set, and it stops notifying the arbiter while it waits (callback_notify only fires from main_loop, which has already exited). The arbiter eventually decides the worker is hung.
init_signals() resets SIGABRT to SIG_DFL, so the arbiter’s timeout kill dumps core instead of running gunicorn’s worker_abort hook, and skips the atexit handler that shuts Galaxy down.

This worker bounds the drain, puts the SIGABRT handler back, and reports what was still in flight whenever a worker is torn down the hard way. Requests that never finish by themselves, such as event streams, can be ended when the drain starts via galaxy.web_stack.on_drain_start().

galaxy.webapps.galaxy.workers.DRAIN_REPORT_INTERVAL = 10: How often to report on a drain that is still in progress.

galaxy.webapps.galaxy.workers.DRAIN_POLL_INTERVAL = 0.25: How often the drain reporter wakes up.

galaxy.webapps.galaxy.workers.DRAIN_DEADLINE_LEAD = 1.0: How far ahead of uvicorn’s own cancellation to report the requests it will cancel.

galaxy.webapps.galaxy.workers.STACK_FRAME_LIMIT = 30: Frames to keep per thread in the stack dump.

class galaxy.webapps.galaxy.workers.Worker(*args, **kwargs)[source]

Bases: UvicornWorker

Galaxy’s gunicorn worker. Referenced by name in gravity and startup scripts.

__init__(*args, **kwargs)[source]: This is called pre-fork so it shouldn’t do anything to the current process. If there’s a need to make process wide changes you’ll want to do that in self.init_process().

init_signals()[source]

async report_drain(server)[source]: Log what is holding a graceful shutdown open, while it is still open.

in_flight_summary()[source]

Compact description of what this worker is still serving.

Kept small enough to survive as a Sentry event message. Called from a signal handler on the abort path, so it must never raise.

in_flight_stacks()[source]

Full stacks for the request tasks and every thread.

The thread pass is the important one: Galaxy’s sync and legacy WSGI endpoints run in a threadpool, so a request task stack usually bottoms out at run_in_threadpool and the code that is actually stuck is only visible here.