[zeromq-dev] Detecting transport level failures

[Marc Allen]

I'm working on an embedded application that uses ZeroMQ to communicate between software components.  I'm trying to figure out how to detect certain types of failure modes on the sender/transport side.  Here's my scenario:

 *   Two TCP sockets, A is a push socket, B is a pull socket
 *   A tries to send a message to B using zmq_sendmsg(socket, msg, ZMQ_DONTWAIT)

My questions are:

 1.  If A's ZMQ socket cannot accept the message because queues/buffers are full or high watermarks are exceeded, will A know right away due to an error code?  I.e., EAGAIN?
 2.  If A's ZMQ socket accepts the message but is only able to partially send the message due to an unreliable connection to B, and the message is discarded due to expiration of ZMQ_SNDTIMEO or ZMQ_LINGER, is there any way for A to know?  I.e. a callback of some sort?  If this doesn't exist, how hard would it be to hack in, and where would you start?
 3.  When using zmq_poll and non-blocking I/O, is there a way to determine if send errors have occurred on a set of sockets?
 4.  Do the above answers change if using ipc:// instead of tcp:// ?

I'm mainly concerned with determining whether data transport to the endpoint was successful.  This is a a lot more complexity that could be discussed here, but this seems like a good starting point.

Thanks very much for your guidance.

-Marc

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20130613/c015afd4/attachment.htm>