Overriding Workers

You can override the code run by each uWSGI worker thanks to the “worker” hook exposed to plugins.

Currently the python plugin is the only one exposing it:

; create a bunch of sockets
socket =
socket =
; spawn the master
master = true
; spawn 4 processes
processes = 4
; load a python script as the worker code
python-worker-override = aioserver.py

The python script has access to the uwsgi module so it can control/change its internals.

The following examples shows the use of aiohttp (requires python 3.5)

import asyncio
from aiohttp import web

import uwsgi
import socket
import sys
import signal

async def handle(request):
   name = request.match_info.get('name', "Anonymous")
   text = "Hello, " + name
   return web.Response(body=text.encode('utf-8'))

async def wshandler(request):
   ws = web.WebSocketResponse()
   await ws.prepare(request)

   async for msg in ws:
     if msg.tp == web.MsgType.text:
         ws.send_str("Hello, {}".format(msg.data))
     elif msg.tp == web.MsgType.binary:
     elif msg.tp == web.MsgType.close:

   return ws

 async def init(loop, fd):
    app = web.Application(loop=loop)
    app.router.add_route('GET', '/echo', wshandler)
    app.router.add_route('GET', '/{name}', handle)

    srv = await loop.create_server(app.make_handler(),
                                   sock=socket.fromfd(fd, socket.AF_INET, socket.SOCK_STREAM))
    print("asyncio server started on uWSGI {0}".format(uwsgi.version))
    return srv

def destroy():
   print("destroy worker {0}".format(uwsgi.worker_id()))

def graceful_reload():
   print("graceful reload for worker {0}".format(uwsgi.worker_id()))
   # TODO do somethign meaningful

loop = asyncio.get_event_loop()
loop.add_signal_handler(signal.SIGINT, destroy)
loop.add_signal_handler(signal.SIGHUP, graceful_reload)
# spawn a handler for every uWSGI socket
for fd in uwsgi.sockets:
   loop.run_until_complete(init(loop, fd))

In the example (taken from the official aiohttp docs) we see the uwsgi.sockets list (holding the list of uWSGI sockets file descriptors), and the override of SIGINT and SIGHUP to support reloading (SIGHUP should be adapted to support waiting for all the queued requests)

uwsgi.accepting() is called to notify the master that the worker is accepting requests, this is required for touch-chain-reload to work.

The script should be extended to call uwsgi.log(…) after every request and to (eventually) update some metrics