gevent, ways of triggering the newly-spawn tasks to run

Question

my project is using gevnet, which is new to me, to implement a thrift server.

I am reading the code and learning from its document. The code snippet below is in my project:

TSocket.socket = gevent.socket # I find this when comparing the production
                               # code and mine. It seems that this socket
                               # schedule the tasks when there are newly-spawn ones. 
                               # Could someone tell me more about it?

while True:
    try:
        client = self.serverTransport.accept()
        gevent.spawn(self._serve_client, client)
    except (SystemExit, KeyboardInterrupt):
        break
    except Exception as x:
        logging.exception(x)

After spawning, it directly finishes it work here.

But in my own implementation, which is for self-learning, i have to do the following:

while True:
    try:
        client = self.serverTransport.accept()
        gevent.spawn(self._serve_client, client)
        gevent.sleep(0)  # switch to the newly-spawn task.
                         # this makes it to process tasks one by one and
                         # this makes it non-concurrent
    except (SystemExit, KeyboardInterrupt):
        break
    except Exception as x:
        logging.exception(x)

In my production code, i haven't found any clue of how to trigger the tasks to run. So I am asking here for some light on the ways of running newly-spawn tasks in a server above.

Answer 1

In your code, you have to call gevent.sleep(0) because nothing else triggers the gevent loop. Having this sleep gives control back to gevent, which executes the spawned greenlet.

The line TSocket.socket = gevent.socket in the production code patches the default thrift socket implementation to use a gevent socket instead; gevent sockets run the gevent loop, so you really need this patch for your code to run.

Without the patch, TSocket.socket is blocking and kills concurrency (without OS-level threads).