Thursday November 3, 2011

[Time] NameMessage
[00:56] tarcieri cremes: hmm, k
[00:58] tarcieri cremes: the only difference I see is you're synchronizing receiving. I'm having trouble doing concurrent sends
[00:58] tarcieri (concurrent sends synchronized with a mutex)
[01:05] tarcieri or in other words, it's the push socket that's having trouble
[01:58] sigmonsay Wow, this is very weird
[01:58] sigmonsay Assertion failed: new_sndbuf > old_sndbuf (mailbox.cpp:183)
[01:59] sigmonsay I'm using req/reply pattern and a request was likely interupted. Does this mean my servers state it entirely out of wack?
[02:06] mikko sigmonsay: are you using the socket from multiple threads?
[02:06] sigmonsay Not where this connection is occuring. The client is threaded, but I use a new socket. Only do I share the context
[02:06] mikko also, which version of zeromq is this?
[02:07] sigmonsay This is only under load also. libzmq1 2.1.7 from Martin Lucina
[02:08] mikko sigmonsay: this is fixed in 2.1.8 i believe
[02:08] sigmonsay Hmm
[02:08] mikko
[02:08] sigmonsay I see on the mailing list bits about raising tcp rmem and wmem. Prolly not a real fix?
[02:09] mikko sigmonsay: why not upgrade the version?
[02:09] sigmonsay I will do so, but just was using someone elses repo. I'll have to dig up where I got it and check if newer bits exist.
[02:09] sigmonsay This is my initial interaction with 0mq
[02:09] mikko which repo?
[02:09] mikko debian?
[02:10] sigmonsay ubuntu lucid, its a ppa I think
[02:10] sigmonsay
[02:11] mikko anyway, newer version should fix that issue
[02:11] mikko i gotta head to bed
[02:11] mikko good night
[02:11] sigmonsay thanks, good night
[11:01] CIA-79 jzmq: 03Gonzalo Diethelm 07master * r6ceaa43 10/ (src/ src/ src/org/zeromq/ Merge pull request #85 from aorzecho/master ...
[11:30] jivs
[14:06] cremes tarcieri: here's a small program that i think mimics your requirement
[14:06] cremes tarcieri:
[14:06] cremes tarcieri: it creates 1 push and 1 pull socket; it starts up a bunch of threads and pushes to that single socket
[14:06] cremes tarcieri: the push socket is protected with a mutex
[14:07] cremes i ran it with 10_000 messages and 500 threads; it exited cleanly and no errors were detected at runtime
[14:07] cremes feel free to adjust the parameters (and transport to inproc) to test for yourself
[14:07] cremes i'm going to be offline for the next 4 days; i may be able to respond to email via my iphone but don't count on it :)
[14:08] cremes good luck... i hope celluloid is a good netizen by the time i get back!
[14:10] cremes tarcieri: btw, the test with those params took about 2m20s to run on my box
[17:45] nerus Hi
[17:46] nerus I am trying to run a c++ version of a zeromq request-reply example
[17:47] nerus getting a really silly error, is there a faq/wiki page on how to troubleshoot?
[18:07] sustrik what erro?
[18:07] sustrik error?
[18:27] nerus sustrik_, sorry was trying out something
[18:28] nerus sustrik_, worker.cpp:1:20: fatal error: zmsg.hpp: No such file or directory
[18:28] nerus but the hpp file is already available in /usr/local/include
[18:29] sustrik zmsg.hpp?
[18:29] sustrik what's that?
[18:29] nerus sustrik_, oops
[18:31] nerus sustrik_, this is where it came from -
[18:31] nerus but I just assumed it was complaining about zmq.hpp
[18:31] nerus changed it to zmq and it works fine
[18:31] nerus sustrik_, thanks :)
[18:33] sustrik you are welcome
[18:44] nerus sustrik, finally figured it out -
[22:05] dirtmcgirt the the ZMQ_REP endpoint terminates, the zmq_recv on the ZMQ_REQ hangs - how do i get the error condition back?