[rabbitmq-discuss] Increasing base memory.

Matthew Sackman matthew at lshift.net
Wed Mar 17 12:44:43 GMT 2010


Hi,

On Wed, Mar 17, 2010 at 09:12:47AM -0300, Gustavo Aquino wrote:
> I would like to run RabbitMQ using 80Gb of memory, so I'm trying to
> configure it using the Rabbit guide where told to put simple line in
> rabbitmq.conf, ok I open my rabbitmq.conf and put this line and just
> changing 0.4 to 0.8 but it doesn't work, when I start Rabbit it don't
> recognize configuration.

The 0.4 is a scaling of the RAM that Rabbit detects in your machine.
Furthermore, as the documentation says, it is *not* a limit for how much
RAM rabbit will use. As I pointed out in my other email, Erlang is a
GC'd language. The 0.4 simply says that channel.flow will be invoked
once 0.4 of available RAM has been used by Rabbit. This does not
preclude Rabbit from using more than 0.4 - the GC events may cause up to
twice this to be used. In addition, it is beneficial to allow your OS to
have some RAM available for disk caches, thus allowing it to be able to
satisfy reads without having to go to disk.

> rabbitmq.conf
> 
> [{rabbit, [{vm_memory_high_watermark, 0.8}]}].
> NODENAME=rabbit at Nefertiti

The first line should be placed in rabbitmq.config. The second line is
valid for rabbitmq.conf. Note they are two very different files.
rabbitmq.config is an Erlang term file. rabbitmq.conf is sourced by the
various shell scripts thus is a location to configure shell variables.

Matthew




More information about the rabbitmq-discuss mailing list