[rabbitmq-discuss] RabbitMQ Server TCP buffer and network tweaks

[Donovan Levinson]

Dear AMQPers,

I am setting two AMQP message brokers on a LAN running the latest RabbitMQ 
Server 3.1.3.  They will be part of a high-availability cluster with full 
mirroring.  The cluster will receive about 430 messages per second (or 
280KB/s incoming data as shown by VM statistics).  Each server process will 
run on its own dedicated VM with 4 CPUs and 8GB of RAM.  Some of the 
publishers are on high-latency WAN connections, while the consumers are 
primarily located on the LAN.

I am considering the following changes:

echo 'net.core.wmem_max=1048576' >> /etc/sysctl.conf
echo 'net.core.rmem_max=1048576' >> /etc/sysctl.conf
echo 'net.ipv4.tcp_rmem= 10240 87380 1048576' >> /etc/sysctl.conf
echo 'net.ipv4.tcp_wmem= 10240 87380 1048576' >> /etc/sysctl.conf
echo 'net.ipv4.tcp_window_scaling = 1' >> /etc/sysctl.conf
echo 'net.ipv4.tcp_timestamps = 1' >> /etc/sysctl.conf
echo 'net.ipv4.tcp_sack = 1' >> /etc/sysctl.conf
echo 'net.ipv4.tcp_no_metrics_save = 1' >> /etc/sysctl.conf
echo 'net.core.netdev_max_backlog = 5000' >> /etc/sysctl.conf

Please let me know your thoughts on whether I should apply these or other 
Linux network tweaks, or if it's recommended to go with the OS defaults?


Thanks,
Donovan

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20130628/212f8e9e/attachment.htm>