I'd like to set up a 0MQ REQ-REP arrangement where there are many REP sockets connected to just one REQ socket. The actual number of REP sockets is unknown at design-time; however at run-time a list of the port addresses will be provided to the code running the REQ socket.
What I'd like is that, somehow, the code running the solitary REQ socket will loop through the list of port addresses and connect to the multiple REP sockets as they bind and become available. Subsequently the REQ socket would work with (poll??) the established connections to send and receive messages as necessary. It's the first time I've tried this. Looking at the zmq_poll reference in the manual http://api.zeromq.org/3-2:zmq-poll it's unclear to me whether I can handle variable number of connections in an iterative manner. For example in: http://zguide.zeromq.org/cpp:mspoller the poll set is hard coded. Another question is can a REQ socket handle multiple connections? How best could a message be routed to the desired destination REP socket? Is some more advanced pattern necessary here ROUTER, BROKER?? What I'm trying to achieve is a REQ-REP flow to act as a command/control layer which will coordinate a PUSH-PULL socket pair. There will be a run-time flexible number of PUSH sockets but always just one PULL socket. Is anyone able to offer any guidance to clear my muddy thoughts on how to make this work? Am coding in C++. With thanks, Riskybiz.
_______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
