To anyone using (or considering) the nginx-sflow-module for monitoring,
then please use version 0.9.9 or later. Previous versions generated
incorrect sFlow output in deployments with multiple worker-processes.
Symptoms included spurious spikes in Ganglia graphs.
This sFlow-HTTP output is the same as you get from the corresponding
agents for apache, tomcat, node.js and haproxy, as well as hardware
load-balancers from F5 and A10. However it can be extended, so if
is anything you would like to see added then please let me know, or
Developers Discussion Board @ sFlow.org.