push with bind about txzmq HOT 11 CLOSED

any update on this?

from txzmq.

You should do mode push in "connect" mode and pull in "bind" mode. Otherwise you end up with no clients on bind.

from txzmq.

That's not true at all... bind push and connect pull is a very widely used ZMQ system... in fact it's the third listed in the guide:

http://zguide.zeromq.org/page:all#Divide-and-Conquer

And a link to the code
http://zguide.zeromq.org/py:taskvent

You start all of your workers/pullers first regardless of whether or not the pusher is bind or connect.

from txzmq.

Sorry for not being clear enough. You can use push/pull with any combination of bind/connect, but for this particular example it doesn't work.

Sending (pushing) to 'bind' socket without pull consumers would result in EAGAIN from ZMQ, and txZMQ doesn't do queueing on its own. This is what happened in the issue description.

from txzmq.

If I connect with multiple (or one) pull consumers first and then bind with a push producing on an interval of one second, I still get EAGAIN from the example. So even though it's quite capable of consuming them quicker than they are produced, they won't be.

from txzmq.

You would get EAGAIN unless pull consumer is connected. Connection doesn't happen simultaneously.

I've pushed change to examples to skip such sends.

from txzmq.

I'm getting EAGAIN when a pull consumer is connected. I have 12 pull consumers connected and the interval is 1 second on the push. There's no way that one isn't connected at send. In bind-pull/connect-push, 1 consumer can handle sends 100x this (interval of 0.01 seconds)

from txzmq.

Please try adding try/except as in 39e841a, and see what happens. First send in this example happens not after 1 second, but immediately.

from txzmq.

Alright, so I just tested it as-is with 1 second interval (in the reactor.callLater call) and with a 0.01 interval. It seems like it successfully produced and consumed a call every time, but regardless of interval, the second logged call is always the "Skipping, no pull consumers..." which seems like it's not correct, since the first listed item and the third listed item are separated by 1x interval, not 2x like they should be if a call failed to find pullers. I'm just wondering what's causing that initial EAGAIN, which I think is erroneous.

from txzmq.

I don't see your code, so it's actually hard to tell what's going on. That EAGAIN is coming from ZMQ itself, not from the txZMQ wrapper. In synchronous version it would just block, txZMQ is async so it just errors back to the caller. You may decide whether you would like to queue outgoing messages, drop them or do anything else.

from txzmq.

I'm still using the push_pull.py example. I have a couple terminal windows open and I'm using typical command line arguments to set up one bind pusher and either one or a few connect pullers. If you do this, the second logged entry for the producer/pusher is always the EAGAIN fail message. I think this may be caused by the way that the example is setup, however. Give me a minute and I'll try a solution I think may work.

from txzmq.