qpid-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gordon Sim <g...@redhat.com>
Subject Re: QPID C++ Broker + persistent store
Date Mon, 23 Feb 2009 20:06:34 GMT
Kim van der Riet wrote:
> On Mon, 2009-02-23 at 09:24 -0800, ffrenchm wrote:
>> Is there any way to force the flow-to-disk option for all messages which are
>> putted in this queue ?
> Perhaps a terminology misunderstanding?
> Messages which are important and which must be recovered in the event of
> a broker failure must be marked durable, and be sent to a queue which
> has also been set to durable. These messages are always written to disk
> before they may be consumed, and will be recovered in the event of a
> broker failure. This is the normal mode to use for reliable messaging.
> Flow-to-disk is used to send messages to disk which are transient (ie do
> NOT need to be written to disk for reliability) when broker resources
> are running low. 

We use flow-to-disk somewhat loosely/incorrectly perhaps.

Currently where the queue policy is set to flow-to-disk (only available 
for durable queues) then the content of enqueued messages that would 
take the queue over the preconfigured limit (either in terms of the 
number of messages or their aggregate size), are released from memory 
and will be read from disk when they are delivered to clients.

This happens for either durable messages (which are ordinarily written 
to disk anyway) or transient messages (which are not).

As Kim points out though, the performance as perceived by subscribers in 
both cases will currently drop considerably if this behaviour is hit so 
it is primarily useful as a last resort to prevent excessive use of 
memory. For best performance you want to set the limit high enough to 
avoid this behaviour during normal operation.

Apache Qpid - AMQP Messaging Implementation
Project:      http://qpid.apache.org
Use/Interact: mailto:users-subscribe@qpid.apache.org

View raw message