> > be up as far as 0MQ is concerned, but unbeknown to it, the TCP socket
> > is dead or will be dead when it tries to use it next.
> That's what I mean - there are socket options for keepalive and
> reconnect timeouts (which also mean connect timeouts) - if you tweak
> those, the first connection pipe might get discarded after the timeout
> - needs testing though
Yeah. I tried tweeking all of these, but it seems that once the
message is queued on a connection via _send() the message never gets
moved to another connection regardless of any connect/reconnect
timeouts. (This is speculation on my part as I haven't look at the 0MQ
code to verify this behaviour - but it appears that way.)
In any event, I still need to deal with a server failure prior to it
sending back a REP and that's something that 0MQ can't handle
Is there an easy way to dump the internal state of each connection and
show the messages queued on it?
zeromq-dev mailing list