IRC Log


Friday May 20, 2011

[Time] NameMessage
[00:07] jhawk28 has anyone had trouble with idle 0mq connections on the same computer?
[06:33] notostraca I have seen some complaints online people have had about 0MQ and I am wondering if one is still valid -- does 0MQ still use assert statements in production code?
[06:33] notostraca (these are pretty dated complaints IIRC)
[06:34] sustrik if you encounter an assert, it's a bug, report the issue on the bug tracker then
[06:35] notostraca good, thank you
[06:36] notostraca I've got a bunch of Sun Netra X1 machines from 2001 (4 of the little buggers are working), and I am in the process of installing libraries on them
[06:36] notostraca so I figured as long as communication over the network will be critical, might as well try something that does it really well -- 0MQ
[06:37] sustrik sure
[06:37] sustrik give it a try
[06:37] notostraca has anyone tried getting 0MQ to work with Hadoop?
[06:37] sustrik dunno
[06:37] sustrik try asking on the mailing list
[06:37] notostraca ok
[06:56] johnm1234 How can I get permission to enter new bugs? I've created an account for the new jira but going to Issues only gives me options to browse/search. thanks.
[06:58] johnm1234 And yes, I just found an assert() triggering in the core. :-)
[07:04] sustrik johnm1234: you have to create account in jira to be able to enter bug reports
[07:05] johnm1234 I have… it's johnm.
[07:05] johnm1234 I can login and see entries but I can't enter anything.
[07:07] sustrik yuck
[07:07] sustrik the issue tracker is being migrated from github to jira atm
[07:08] sustrik probably doesn't work well still
[07:08] johnm1234 Yeah, I saw that.
[07:08] sustrik can you send the bug report to the mailing list for now?
[07:08] johnm1234 When I created the account a couple of days ago, I could login but not see anything. :-)
[07:08] johnm1234 Sure.
[08:17] xyler Hi, I have just started using zeromq. I tried binding two pull sockets on different end points and connected to both of them using one push socket. Now, if a downstream node goes down while the upstream node is sending messages, I am losing messages. Is this the defined behavior or am I doing something wrong? Thank you.
[08:17] mikko sustrik: there? can you give me an example of issue where comments dont match?
[08:26] sustrik mikko: no idea
[08:26] sustrik mato has an example
[08:27] sustrik xyler: yes, it's nornal, the messages queued at the downstream node are lost when the node crashes
[08:28] xyler sustrik: Thanks for replying. So, for reliability do I need to send explicit acks or is there some other pattern that ensures reliability?
[08:29] sustrik explicit acks should solve the crashing downstream node problem
[08:29] pieter_hintjens sustrik: I'm away for the next hours, will try to be online randomly
[08:30] sustrik ah, you are in america already?
[08:30] sustrik sure, no problem
[08:30] pieter_hintjens no, I'm still in Brussels
[08:31] pieter_hintjens but have my usual friday meetings
[08:31] pieter_hintjens leaving for the states on Monday
[08:31] sustrik ok
[08:31] xyler sustrik: thanks.
[08:32] sustrik you are welcome
[12:37] mato mikko: yeah, the comments don't match up with the issues
[12:38] mato mikko: i'm not sure what happened; the user names etc. are all correct, it just looks like the comments are from some other issue
[12:38] mato mikko: take a look at the two I mentioned in my email
[13:43] mato mikko: ping?
[13:43] mato mikko: i can work with atlassian to fix the import, but i need at least the original issues data
[15:02] pieter_hintjens sustrik: again with the "Assertion failed: nbytes == sizeof (command_t) (mailbox.cpp:245)" on zeromq-dev
[15:02] pieter_hintjens can we catch this and report it in a more useful way?
[15:16] jhawk28 it looks like it was related to passing sockets between threads
[15:18] jhawk28 so I'm not sure its really a 0mq error, pieter
[15:19] jhawk28 may be more of a Java bindings error if it isn't passing around the sockets correctly
[15:20] pieter_hintjens jhawk28: sure, but this is the 2nd time we have this assert reported to the list, in a week
[15:20] pieter_hintjens both times, due to passing sockets between threads
[15:21] pieter_hintjens ok, I'm just going to add a warning to mailbox.cpp
[15:26] jhawk28 that would probably work
[15:27] jhawk28 do you happen to know where the HWM is used to limit messages in the mailbox?
[15:28] jhawk28 or is it elsewhere?
[15:31] pieter_hintjens jhawk28: nope, you can presumably find it in the code
[15:31] jhawk28 was looking...not great with C
[16:14] jhawk28 pieter_hintjens: ah, found it - its in pipe.cpp
[16:16] jhawk28 "bool zmq::writer_t::pipe_full ()"
[20:14] yy8 hi all,
[20:15] yy8 i have been looking for a efficient transport layer for quite a while and somebody recommend 0mq, i have a question about the implementation: is the current implementation use the TCP/IP layer?
[20:16] yy8 or it is using directly hardware NIC driver to implement the 0MQ library?
[20:18] jhawk28 TCP layer
[20:18] jhawk28 but, there are different transport protocols used such as PGM, EPGM, TCP, IPC, and INPROC
[20:19] yy8 i see, so it is live in user space
[20:19] yy8 ?
[20:19] ParadoxG Hi,
[20:19] ParadoxG I'm getting a seg fault on a message copy
[20:19] ParadoxG could anyone look and see whats happening: http://pastebin.com/Q53V3UHG
[20:21] jhawk28 yy8: yes
[20:21] yy8 jhawk28: thanks
[20:22] jhawk28 yy8: the guide is the best way to get a feel for 0mq - http://zguide.zeromq.org/page:all
[20:22] yy8 jhawk28: is there some consideration of moving the implementation to kernel space or as kernel model to further optimzie this
[20:23] yy8 thanks for the pointer, I read briefly of them, looks like most of them is on how to use 0mq
[20:25] jhawk28 yy8: that is in "initial stages" - see http://www.250bpm.com/hits for an idea on the direction
[20:26] yy8 jhawk28: thanks very much
[20:26] jhawk28 yy8: Architecture - http://www.zeromq.org/whitepapers:architecture and Roadmap for 3.0 - http://www.zeromq.org/docs:3-0
[20:26] yy8 is there some people use 0MQ for high performance computing, like compare it to MPI implemetnation?
[20:27] yy8 jhawk28: thanks. I am reading the architecture page too :)
[20:28] jhawk28 yy8: http://stackoverflow.com/questions/35490/spread-vs-mpi-vs-zeromq
[20:29] jhawk28 yy8: http://nichol.as/zeromq-an-introduction too (some mentions MPI and HPC)
[20:30] jhawk28 later, gotta run
[20:32] sustrik ParadoxG: the 'copy' is used uninitialised
[20:32] ParadoxG am I suppose to use zmg_msg_init on it before?
[20:32] ParadoxG I'm trying to write ZeroMQ bindings for Vala
[20:33] sustrik yes
[20:33] sustrik you have to init it before using it
[20:39] ParadoxG I added in the init, but still I get a seg fault: http://pastebin.com/neQALEYb
[20:43] blk ParadoxG: have you tried gdb? use zmq_msg_close to release a msg
[20:44] blk http://api.zeromq.org/2-1-1:zmq-close
[20:44] blk http://api.zeromq.org/2-1-1:zmq-msg-close
[20:52] ParadoxG changing zmq_close to zmq_msg_close worked, and works in Vala language now too :)
[20:52] ParadoxG thanks
[22:14] Seta00 libzmq.a is giant :/