[zeromq-dev] Clean shutdown of malamute client zactor in python binding?

Axel Voitier axel.voitier at gmail.com
Thu Feb 4 14:49:49 CET 2016


Python makes threading sequential, yes. But it can still chops execution of
a thread in the middle of a function. Such as a function writing to a
shared buffer to be flushed on a zmq socket later. That is how you can end
up with messages mixed up if two threads use the same function and/or write
to the same buffer. (Also happens with reads, of course).

At least with gevent/eventlet you do have sequential execution and you
control the yielding points.


Axel


2016-02-04 14:41 GMT+01:00 Aaron Sokoloski <asokoloski at gmail.com>:

> On 4 February 2016 at 05:45, Arnaud Loonstra <arnaud at sphaero.org> wrote:
>
>> As long as it is in Python it will be sequential anyway. In a real
>> concurrent setup you probably want n:1 topology in between the
>> webservers and the malamute instance and pass messages, right?
>>
>
> Well, yes, but this is a bit of an uncommon type of web server.  Internal
> use only, so not many clients, but pretty heavy weight processing.  I'm
> trying to avoid having lots of copies of the malamute client, because it
> needs to cache some pretty big state.  So instead the process will have a
> single client, and webserver threads will query the cache using inproc
> sockets or something like that.
>
> By the way, I think python isn't 100% sequential -- execution of python
> code is, due to the GIL, but c library code can actually execute in
> parallel.
>
> I might not have the same problem if using something like gevent or
> eventlet, but making czmq work with one of them is much more work than I
> can afford to get into right now.
>
>
>
> _______________________________________________
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20160204/62bb8519/attachment.htm>


More information about the zeromq-dev mailing list