Friday May 14, 2010

[Time] NameMessage
[09:32] dodobas yello
[09:32] dodobas where can i find refernce for config files
[09:33] dodobas im trying to run zmq_queue
[09:38] dodobas can 0mq bu used like amqp broker, similar to rabbitmq
[09:39] guido_g not really, because there is no broker
[09:42] dodobas so, there is no amqp support?
[09:43] guido_g no
[09:43] dodobas im trying to find an replacement for rabbitmq, which is interfaced through py-amqp
[09:44] guido_g then 0mq is not an option
[09:44] dodobas and 0mq is refernced as an amqp message broker implementation
[09:44] dodobas too bad
[09:44] guido_g but if you're willing to leave amqp... :)
[09:44] guido_g 0mq is not related to amqp
[09:45] guido_g at least technically
[09:45] dodobas sorry, ill be back in an hour
[09:45] guido_g the refernce in question is wrong then
[10:37] dodobas guido_g: for instance
[10:49] guido_g so? it's plain wrong
[10:52] sjampoo 0MQ had some AMQP support in earlier versions but this has been dropped and the project has moved on
[10:53] sjampoo read some more info here:
[11:35] dodobas guido_g: i now understand :D
[11:35] guido_g fine
[11:36] dodobas but, if every article refrences 0mq, as AMQP, its hard not to be wrong :D
[11:37] guido_g huh? every article?
[11:37] dodobas 'almost'
[11:39] dodobas i've found this short message
[11:40] dodobas it doesn't really matter, thanks guido_g for clalification
[11:40] dodobas *clarification
[11:54] sustrik the AMQP issue is explained in FAQs
[11:55] sustrik not sure what else can be done...
[14:29] cremes sustrik: any plans for a flush command on sockets to remove all queued messages?
[14:30] cremes use case: an application-level request has timed out, so any late response needs to be discarded
[14:57] mike8901 Has anyone used a zeromq app with mpirun?
[14:58] mike8901 i.e. using an MPI cluster to bootstrap into zeromq
[15:08] sustrik cremes: yes, there's a plan for adding timeouts for req/rep scenario
[15:08] sustrik mike8901: i am not aware of it, but you may ask on the mailing list
[15:10] cremes how about a flush function for sockets?
[15:14] sustrik cremes: what's the connection between flush and timeouts?
[15:14] cremes i have a use case where i am using req/rep sockets as a command channel and pub/sub
[15:14] cremes sockets for data transfer
[15:15] cremes if the req/rep times out, i need to flush the sub socket if the data shows up late
[15:15] cremes make sense?
[15:15] sustrik flush = drop all the outstanding messages?
[15:15] cremes yes
[15:16] cremes since these sockets represent queues, shouldn't there be a way to flush them out?
[15:17] sustrik hm, you can read and drop the messages, no?
[15:18] cremes yes i can, but isn't the lib doing work behind the scenes to populate the zmq_msg_t? an explicit flush in the lib could avoid that, yes?
[15:18] sustrik no, the whole thing is zero-copy
[15:18] sustrik meaning that data are read to zmq_msg_t as they arrive from the network
[15:19] cremes oh, i see
[15:19] sustrik when you recv, the actual work is copying some 40 bytes or so
[15:19] sustrik almost nothing
[15:19] cremes then i can create my own flush at close to zero processing cost
[15:19] cremes never mind then!
[15:19] sustrik yes, even if it where implemented inside 0mq
[15:19] sustrik it would have to do the same thing
[15:20] sustrik (read & drop messages)
[19:10] CIA-15 pyzmq: 03Brian Granger 07master * r7d4c37b 10/ (6 files in 2 dirs): Fixing various bugs and updating to latest 0MQ master. -
[19:51] OldNugget meta-a /input jump_smart
[19:52] OldNugget oops... sorry
[22:14] CIA-15 pyzmq: 03Brian Granger 07testing * rc4bdbab 10/ (11 files in 4 dirs): Trying to get out of jail. -
[22:19] CIA-15 pyzmq: 03Brian Granger 07testing * r6441176 10/ zmq/tests/ : Finish cleaning up after merge. -
[22:20] CIA-15 pyzmq: 03Brian Granger 07master * rc4bdbab 10/ (11 files in 4 dirs): Trying to get out of jail. -
[22:20] CIA-15 pyzmq: 03Brian Granger 07master * r6441176 10/ zmq/tests/ : Finish cleaning up after merge. -