[rabbitmq-discuss] RabbitMQ configuration

Thu May 13 13:30:55 BST 2010

Hi,

Thanks for the reply.

Unfortunately it has to be windows as our company does not support non
windows.

As to our usage.

We're processing messages from a stock exchange and setup a routing key
containing of "level.isin"
where the level can be level 1 or level 2 depending on the type of
information and the isin is the stock.

"Clients" can then bind to a specific routing key to get just the
information they need.
I'm thus using RabbitMQ as a way to transfer data and demultiplex the
information that comes through.
I could use RabbitMQ just for the message transfer and then do the
demultiplexing myself but I thought I'd take a shortcut.

On the configuration. Is there some documentation on what or how RabbitMQ
can be configured.
Currently I can't find a configuration file anywhere so don't know where/how
to start.

Thanks again

Roaan

On Thu, May 13, 2010 at 11:45 AM, Matthew Sackman <matthew at lshift.net>wrote:

> Hi Roaan,
>
> On Thu, May 13, 2010 at 06:57:19AM +0200, Roaan Vos wrote:
> > We've been having problems with RabbitMQ just "hanging".
> > This is with the default installation, i.e no configuration.
>
> It's likely you've run out of memory and Rabbit's raised the
> channel.flow flag, preventing you from publishing more messages to it.
> Check the logs for memory high watermark warnings. This issue is fixed
> with the new persister branch as messages can be sent to disk and
> forgotten from RAM.
>
> > Can any one help with a configuration file or point me to where I can
> find
> > information on how to configure RabbitMQ.
>
> There's not a great deal you can do to fix this issue other than
> compiling from source with the new persister branch (bug21673), or
> installing much more memory.
>
> > We're running RabbitMQ 1.7.2 on a single node on Windows 2008 Server.
>
> Windows is a bad idea because there is no 64-bit Erlang available. As
> such, it doesn't matter how much RAM you have installed, only 2GB will
> be available to Rabbit. If you need to use Windows then your only
> solution will be to use the new persister branch (but compiling under
> Windows is involved to say the least - you'll need to start with
> cygwin). Really we'd recommend a Linux server here.
>
> That said, all my assumptions may be totally wrong - how many messages
> (and what size) are you buffering in Rabbit? Are you correctly acking
> messages when you're consuming them?
>
> > Also we're aiming on having about 10 thousand clients (connections) each
> > with their own queue. (Using Topics)
> > Is this feasable?
>
> Connections - yes. You will probably need to raise the process limit -
> about a million should be fine. Under non-Windows this would involve
> putting SERVER_START_ARGS="+P 1048576" into the rabbitmq.conf file
> (which normally lives under /etc/rabbitmq). However, because of lack of
> features of Windows BAT files, there is no such .conf file, so you need
> to set the environment variable RABBITMQ_SERVER_START_ARGS and set the
> value to "+P 1048576". That should do the trick.
>
> Topic exchanges are inefficient in our current implementation. It's not
> normally necessary to use them either - unless your domain of routing
> keys really is unbounded. Can you describe more about your use case and
> we should be able to advise you further.
>
> Matthew
>
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20100513/2760723e/attachment.htm 