Understanding Eventlet Hubs

A hub forms the basis of Eventlet’s event loop, which dispatches I/O events and schedules greenthreads. It is the existence of the hub that promotes coroutines (which can be tricky to program with) into greenthreads (which are easy).

Eventlet has multiple hub implementations, and when you start using it, it tries to select the best hub implementation for your system. The hubs that it supports are (in order of preference):

epolls

Linux.

poll

On platforms that support it

selects

Lowest-common-denominator, available everywhere.

pyevent

This is a libevent-based backend and is thus the fastest. It’s disabled by default, because it does not support native threads, but you can enable it yourself if your use case doesn’t require them. (You have to install pyevent, too.)

If the selected hub is not ideal for the application, another can be selected. You can make the selection either with the environment variable EVENTLET_HUB, or with use_hub.

eventlet.hubs.use_hub(hub=None)

Use this to control which hub Eventlet selects. Call it with the name of the desired hub module. Make sure to do this before the application starts doing any I/O! Calling use_hub completely eliminates the old hub, and any file descriptors or timers that it had been managing will be forgotten. Put the call as one of the first lines in the main module.:

""" This is the main module """
from eventlet import hubs
hubs.use_hub("pyevent")

Hubs are implemented as thread-local class instances. eventlet.hubs.use_hub() only operates on the current thread. When using multiple threads that each need their own hub, call eventlet.hubs.use_hub() at the beginning of each thread function that needs a specific hub. In practice, it may not be necessary to specify a hub in each thread; it works to use one special hub for the main thread, and let other threads use the default hub; this hybrid hub configuration will work fine.

It is also possible to use a third-party hub module in place of one of the built-in ones. Simply pass the module itself to eventlet.hubs.use_hub(). The task of writing such a hub is a little beyond the scope of this document, it’s probably a good idea to simply inspect the code of the existing hubs to see how they work.:

from eventlet import hubs
from mypackage import myhub
hubs.use_hub(myhub)

Supplying None as the argument to eventlet.hubs.use_hub() causes it to select the default hub.

How the Hubs Work

The hub has a main greenlet, MAINLOOP. When one of the running coroutines needs to do some I/O, it registers a listener with the hub (so that the hub knows when to wake it up again), and then switches to MAINLOOP (via get_hub().switch()). If there are other coroutines that are ready to run, MAINLOOP switches to them, and when they complete or need to do more I/O, they switch back to the MAINLOOP. In this manner, MAINLOOP ensures that every coroutine gets scheduled when it has some work to do.

MAINLOOP is launched only when the first I/O operation happens, and it is not the same greenlet that __main__ is running in. This lazy launching is why it’s not necessary to explicitly call a dispatch() method like other frameworks, which in turn means that code can start using Eventlet without needing to be substantially restructured.

More Hub-Related Functions

eventlet.hubs.get_hub()

Get the current event hub singleton object.

Note

This is considered an internal API, and it might change unexpectedly without being deprecated first.

eventlet.hubs.get_default_hub()

Select the default hub implementation based on what multiplexing libraries are installed. The order that the hubs are tried is:

• epoll

• kqueue

• poll

• select

It won’t automatically select the pyevent hub, because it’s not python-thread-safe.

Note

This is considered an internal API, and it might change unexpectedly without being deprecated first.

eventlet.hubs.trampoline(fd, read=None, write=None, timeout=None, timeout_exc=<class 'eventlet.timeout.Timeout'>, mark_as_closed=None)

Suspend the current coroutine until the given socket object or file descriptor is ready to read, ready to write, or the specified timeout elapses, depending on arguments specified.

To wait for fd to be ready to read, pass read =True; ready to write, pass write =True. To specify a timeout, pass the timeout argument in seconds.

If the specified timeout elapses before the socket is ready to read or write, timeout_exc will be raised instead of trampoline() returning normally.

Note

This is considered an internal API, and it might change unexpectedly without being deprecated first.