Monday August 23, 2010

[Time] NameMessage
[04:49] sustrik benoitc: no
[06:21] shekispeaks hi
[06:22] shekispeaks I am looking to build a similar MQ as SQS using zeromq
[06:22] shekispeaks are there any example which are asynch message queues
[06:22] shekispeaks and probably also allow persistance
[06:35] sustrik shekispeaks: no persistence in 0mq
[06:36] shekispeaks ya but if I can push into a datastore like SQLite
[06:36] sustrik sure, why not
[06:36] shekispeaks apart from persistence
[06:36] shekispeaks I want to implement ansynchronus MQ
[06:36] sustrik that's what 0mq is
[06:36] shekispeaks so the Queue Program should store the messages till the consumer consumes them
[06:37] sustrik ah, take the queue device and add persistence there
[06:37] shekispeaks any examples which are doing the same
[06:37] sustrik src/queue.cpp
[06:37] sustrik that's the queue
[06:38] sustrik you'll have to plug SQLite there
[06:50] keffo why on earth would you jump to an sql relational database, for persistence of a message-queue
[06:50] keffo ?
[06:50] shekispeaks what else can be the options
[06:51] keffo a file? =)
[06:51] shekispeaks :D ya file
[06:52] keffo I'd just use lua though
[06:52] shekispeaks i am using python
[06:53] keffo probably not as efficient, but should work nicely
[06:53] keffo it all depends on how long you need to cache stuff, and how much
[07:19] shekispeaks can you point me to some similar examples preferably in python
[07:35] pieterh shekispeaks: i think it's a good idea, but we don't have examples yet
[07:35] pieterh it's one thing i wanted to make for the user guide
[07:35] pieterh it needs two parts
[07:35] pieterh reliable request-reply and then a queue device that sits on disk
[07:39] keffo pieterh, how is the reliable rep/req supposed to work, apart from the queue?
[07:39] pieterh well, you need a queue device in the center that can connect N clients to N services
[07:40] pieterh then you need a reliable way to get messages onto that queue, and off it
[07:40] pieterh you can start by taking apart the queue device itself
[07:40] pieterh and seeing how it connects XREP to XREQ
[07:41] pieterh you basically want to stick a persistent store in between those afaics
[07:41] pieterh so your chain from client to service is REQ--XREP--(queue)--XREQ--REP
[07:41] pieterh then you need to make the REQ--XREP and XREQ--REP parts reliable as well
[07:42] pieterh which also means some persistence (at least in memory) at each node
[07:42] pieterh as sustrik says, src/queue.cpp is the place to start
[07:43] pieterh kind of something i wanted to play with too...
[07:43] pieterh even using simple disk files would be doable IMO
[07:49] keffo metakit sounds perfect
[13:07] panki hello
[13:07] panki Martin, are you are here?
[14:58] cremes is the current master branch at zeromq/zeromq2 still api compatible with the 2.0.7 release?
[15:03] cremes the answer is "no"; zmq_poll has its contents commented out and has left an assertion that always triggers
[15:03] cremes anyone know what the last commit was where zmq_poll was supported in master?
[15:24] cremes i rolled back to commit 4777fe4010572d381a2ad8eb63df2fc5fb7e6642
[15:24] cremes i am seeing a problem where i get the following assertion: Assertion failed: !engine (session.cpp:287)
[15:26] cremes has anyone else seen that?
[15:26] cremes a search of the git issues doesn't provide any hits
[15:30] mato cremes: zeromq/zeromq2 master is api compatible with 2.0.7
[15:30] mato cremes: if you're looking at a commented out poll, my guess is you're using sustrik/zeromq2 master
[15:30] mato cremes: maybe you forgot to switch your url's?
[15:32] cremes mato: that's probably it; i'll reclone and try again
[15:43] cremes argh... autoconf issues on osx :(
[15:47] cremes "sudo port install libtool" fixed the issue
[16:39] sustrik cremes: the comment about building on OSX refers to build from package or build from git?
[16:40] cremes from git
[16:40] cremes the git sources don't include "configure" so it needs to be generated
[16:40] cremes using autoconf & friends
[16:43] sustrik ok, let me see
[16:44] sustrik what about putting the comment here:
[17:00] cremes sustrik: sure, i'll move it there
[17:03] sustrik ok
[17:06] cremes done
[17:07] sustrik thx