[rabbitmq-discuss] problem with new persiter with 1000 topics/queues

Matthias Radestock matthias at rabbitmq.com
Wed Aug 25 06:05:04 BST 2010


Alex,

alex chen wrote:
> this could be the amqp_producer.c in rabbitmq-c/examples.  simply modify the 
> code to accept the topic name and message size from command line.

Please send us the modified code.

>> - how exactly you are invoking the producer and consumer ("1000  
>> amqp_consumers" - but with what args?, etc)
> 
> i use 3 bash shell scripts (bind.sh, producer.sh, consumer.sh):
> [...]

Thanks.

> the test steps are:
> 1) run bind.sh
> 2) run producer.sh

There must be a step 0) missing here. amqp_bind does not create any 
queues and will fail if the referenced queues do not exist. So how are 
the queues created? Please send us the code/scripts for that.

Also, are all the above running on the same machine as the broker?

> our recent test shows that if you run producer.sh and consumer.sh at the same 
> time (without 200 GB messages backed up), the broker process would also use 
> large amount of mem (around 6 GB).  this way you can reproduce the high mem 
> usage problem without waiting long time for the 200 GB messages backed up.

ok. That's useful to know.

Btw, what revision of the server are you running? 'hg id' will tell you.


Regards,

Matthias.


More information about the rabbitmq-discuss mailing list