[Answer]-Is a single Python executable file shared by all clients when run on a server?

The execution flow should be the same when you run it on your local machine. Take the following code:-

import tornado.ioloop
import tornado.web

print 'This is executed only once and global'
name = 'myname'

class MainHandler(tornado.web.RequestHandler):
    print 'This is executed only once'
    def get(self):
        print 'This is executed for each requests'
        self.write("Hello, world %s " % name)

application = tornado.web.Application([
    (r"/", MainHandler),
])

if __name__ == "__main__":
    application.listen(8888)
    tornado.ioloop.IOLoop.instance().start()

When, we’ll get the following output:-

$ ./bin/python server.py 
This is executed only once and global
This is executed only once



This is executed for each requests
WARNING:tornado.access:404 GET /favicon.ico (::1) 0.52ms
This is executed for each requests
WARNING:tornado.access:404 GET /favicon.ico (::1) 0.25ms

This mean the code outside the class body will be executed every time the server start and any state will stay there until it being restarted. When the server was started, it also instantiate the class MainHandler and for each incoming requests, the get method on the instance get called. This mean the code in the class body will also being executed once and only the code in the method get will get a fresh state for each incoming requests. I assume ‘client’ in your question mean incoming web requests.

Usually in normal python wsgi application, the way to keep some shared state between function without explicitly passing parameters down the chain is to use the thread-local object. I don’t know much about tornado but from a brief read it seem doesn’t run inside thread so you have to consult the docs to find out the recommended way of sharing state between functions.