#rabbitmq IRC Log

Index

IRC Log for 2011-08-22

Timestamps are in GMT/BST.

[1:26] <eselle> Is there any way to force a new consumer tag every time and not reuse old ones?
[1:58] <munderwo> Hi all. We had a rabbitmq server go down, and not come back up. We've backed up the mnesisa directory. and are wondering how to recover the messages that are queued?
[3:00] <munderwo> hmm.. on close inspection it appears that all out messages are in the persister.LOG ..is there a way to get mnesia to read this?
[11:23] <TreeHutMonk> hi.
[11:24] <TreeHutMonk> I'm using the dotnet client for RabbitMQ. Is there any way to set it to attempt to reconnect if the underlying network connection disconnects?
[13:11] <snack__> I'm getting this on creating a queue: "NOT_FOUND - no queue 'qname' in vhost 'vhostname'"
[13:11] <snack__> isn't it supposed to create the queue?
[13:17] <TreeHutMonk> snack__: Are you declaring the queue?
[13:19] <snack__> with ruby amqp, I'm doing: queue = @channel.queue("qname", :auto_delete => false, :durable => true)
[13:20] <snack__> it works on other vhosts
[13:23] <antares_> snack__: take a look at the rabbitmq log
[13:23] <antares_> snack__: I have seen this today. It has nothing to do with any particular client.
[13:23] <antares_> looks like a rabbitmq issue. I re-created vhost and the problem went away.
[13:23] <antares_> it happened to me on binding
[13:24] <snack__> antares_: I'm trying tha
[13:24] <snack__> that*
[13:24] <antares_> there were no entries about queue.declare issues in the log
[13:24] <snack__> yup, worked
[13:24] <antares_> and I was using Java client most of the time
[13:25] <snack__> is there a bug ticket for this?
[13:26] <antares_> even worse: rabbitmq bug tracker is not open to the general public
[13:27] <antares_> snack__: by the way, one more thing I did was restarting the broker after I deleted vhost
[13:29] <snack__> ffs
[13:29] <snack__> antares_: I didn't have to do that
[13:33] <antares_> snack__: so it is ok now?
[13:34] <snack__> antares_: yeah for this vhost
[13:34] <snack__> but I run a SaaS service that has many vhosts
[13:34] <snack__> so I'm likely to find more vhosts with the same problem
[13:35] <antares_> snack__: I fear the same. Then please report it to rabbitmq-discuss.
[13:36] <snack__> word
[13:51] <jelveh> is there a simple way to get the pid of the rabbitmq process?
[13:58] <antares_> jelveh: $1.to_i if `rabbitmqctl status` =~ /\{pid,(\d+)\}/
[13:59] <antares_> jelveh: this is what we use in the amqp gem tests
[13:59] <jelveh> antares_: ahh cool, will make a bash thing out of that
[14:54] <rparra> Hi I am running a load/soak test in preparation of our deployment of RabbitMQ into production. I have the management-agent-plugin in stalled on all 3 nodes of a cluster. The management plugin is installed on the disk node, the other two are ram nodes. The load test simply "blasts" a stream of messages as fast as it can into rabbitmq (at a rate of about 1000 TPS). After about 10 minutes, the statistics gathering of the queues shows
[14:55] <rparra> However my test is still sending messages and receiving messages. It finishes just fine after an hour. (I'd like to bump this to 24 hours and 72 hours to verify rabbit can handle the consistent load / stress).
[14:56] <rparra> Anyone have any idea as to why the stastics just stop gathering ?? Is there some DB limit that needs to be increased to handle the load ?
[14:56] <antares_> rparra: I think half of your first message got lost
[14:56] <antares_> rparra: it cuts off at «the statistics gathering of the queues shows»
[14:56] <rparra> @antares_ lemme repost.
[14:56] <antares_> (for me)
[14:56] <rparra> Hi I am running a load/soak test in preparation of our deployment of RabbitMQ into production.
[14:56] <rparra> I have the management-agent-plugin in stalled on all 3 nodes of a cluster.
[14:57] <rparra> The management plugin is installed on the disk node, the other two are ram nodes.
[14:57] <rparra> The load test simply "blasts" a stream of messages as fast as it can into rabbitmq (at a rate of about 1000 TPS).
[14:57] <rparra> After about 10 minutes, the statistics gathering of the queues shows 0 msgs/sec and the channels show as "IDLE".
[14:57] <rparra> However my test is *STILL* sending messages and receiving messages.
[14:57] <rparra> It finishes just fine after an hour. (I'd like to bump this to 24 hours and 72 hours to verify rabbit can handle the consistent load / stress).
[14:57] <rparra> Anyone have any idea as to why the stastics just stop gathering ?? Is there some DB limit that needs to be increased to handle the load ?
[14:58] <rparra> antares_: is that better ?
[14:59] <antares_> rparra: yes
[14:59] <antares_> rparra: you are hitting TCP back pressure, I think
[14:59] <rparra> Would TCP back pressure impact msg delivery ?
[14:59] <antares_> if consumers cannot keep up and rabbitmq sees that messages pile up in queues, it will apply TCP back pressure until the queues catch up
[15:02] <swaj> Hey all. I've just installed RabbitMQ 2.5.1 on a windows-based server, and everything appears to be working just fine. I wanted to install the administration management plugin, and I've copied all the 6 *.ez files into the plugins folder and restarted rabbitMQ, but the interface won't come up. It's not even listening on the port.
[15:02] <swaj> anyone know how I could get the admin UI working? The log files don't show anything that looks out of place to me.
[15:03] <antares_> swaj: do the log files mention plugin activation?
[15:03] <swaj> they don't
[15:04] <antares_> swaj: it should say: 6 plugins activated (or something like that)
[15:04] <swaj> yeah, nothing like that is happening. I dropped the 6 files in the C:\program files (x86)\RabbitMQ Server\rabbitmq_server-2.5.1\plugins folder
[15:04] <swaj> and I made sure to "unblock" them with the windows properties window
[15:05] <balboah> Having a hard time gettint rabbitmqctl to run properly from a munin task, I get this error: http://pastie.org/2411547 but I don't really understasnd it
[15:05] <antares_> maybe another copy of rabbitmq was running and so it did not restart?
[15:06] <swaj> what's the proper way to restart? I did a "rabbitmq-service stop" and then "rabbitmq-service start && rabbitmqctl start_app"
[15:06] <antares_> swaj: I never used rabbitmq on windows so I am not sure
[15:06] <antares_> swaj: feel free to ask on rabbitmq-discuss (http://groups.google.com/group/rabbitmq-discuss)
[15:08] <rparra> antares_: I got disconnected.
[15:08] <antares_> rparra: ok. My reply was «if consumers cannot keep up and rabbitmq sees that messages pile up in queues, it will apply TCP back pressure until the queues catch up»
[15:09] <rparra> the stats says 0: ready deliv: 997 tps, ack rate: 997 tps
[15:09] <rparra> does this imply that consumers are keeping up ?
[15:10] <rparra> that first stat is ready: 0.
[15:12] <swaj> antares_: hah, I figured it out... you have to rabbitmq-service remove, then reinstall it
[15:12] <rparra> btw I have my own metrics gathering showing msgs are still being delivered @ ~1000tps... so it certainly sounds like some management statistics limit
[15:12] <swaj> apparently the plugins folder is only scanned once, when the service is installed.
[15:12] <antares_> rparra: the management plugin updates statistics every 5 seconds (or more)
[15:14] <swaj> woohoo, thanks for the help. My Admin UI is running :)
[15:20] <rparra> antares_: Yes. Every 5 seconds it (mgmt_plugin) updates just fine.
[15:20] <rparra> after about 10 mins or so... it seems like it "stops collecting statistics"
[15:39] <antares_> rparra: please report it to rabbitmq-discuss
[16:01] <rparra> antares_: Sure will. Thanks!
[20:05] <egradman> Hi. I'm using rabbitmq on ubuntu (vanilla install). My server has stops every day or so for unknown reasons.
[20:06] <egradman> I see a "Stop-and-halt request received; halting in 1000 milliseconds" in my previous log, but I can't track down the source of that stop-and-halt"
[20:06] <mvdir> egradman: Any errors in the logs?
[20:06] <mvdir> Ah.
[20:06] <antares_> egradman: please use http://www.rabbitmq.com/debian.html#apt
[20:06] <egradman> ah, excellent.
[20:07] <egradman> that's exactly what I wanted to hear. Install a better version and your problems will vanish :)
[20:07] <mvdir> I have a behavior I interpret as strange, where consumers will consume much faster when a queue is receiving many messages, rather than just a tricklel.
[20:09] <mvdir> ie, if a queue is receiving 2500 msg/s, it will publish at 3200.
[20:09] <mvdir> If It is receiving only 100 msgs/s, it will publish at only about 1200/s.
[20:09] <mvdir> Has anyone else observed this, or can they perhaps explain the behavior?
[20:10] <mvdir> Maybe the statistics collection is just bad...
[20:15] <captainramen> i have an architecture idea i wanted to run by you all..
[20:16] <captainramen> in the cloud. i was thinking of deploying one rabbitmq instance with each client facing webserver and one with each consumer (i.e. one consumer per machine)
[20:16] <captainramen> and then joining them all up in a cluster
[20:16] <captainramen> is this stupid?
[20:19] <antares_> captainramen: it depends on what you are trying to achieve with it :)
[20:27] <mvdir> Has anybody in here tried the riak persistence backend?
[20:38] <antares_> mvdir: there is a messages store that uses Riak?
[20:38] <antares_> or that exchange type that uses riak to back messages up?
[20:39] <mvdir> https://github.com/jbrisbin/rabbit_riak_queue
[20:39] <mvdir> Same guy.
[20:39] <mvdir> Never tried to deploy it or anything.
[20:40] <mvdir> I'm still teh erlang n00b.
[20:43] <mvdir> I think he abandoned it because there was no contract as to the data format that would be stored in the datastore.
[20:44] <mvdir> I think he wanted to peruse the queues while they were stored.
[20:45] <antares_> that's an excellent example for me to build a Redis one
[20:45] <antares_> we need it for travis-ci.org to get queue introspection
[21:33] <mvdir> Is there still no public rabbitmq bugtracker?
[21:46] <captainramen> doh
[22:20] <mvdir> I am looking at this:http://www.rabbitmq.com/tutorials/tutorial-two-java.html.
[22:21] <mvdir> Does only setting prefetch_count to 1 have the effect of equalizing the round-robin approach.
[22:21] <mvdir> Or would setting the prefetch_count parameter to any value have the same effect?
[22:22] <mvdir> ie. a prefetch count of 10 would have essentially the same effect.

These logs were automatically created by MyxoRoboto on irc.freenode.net using the Java IRC LogBot.