For example, 2 instances of 1 process produce some metric using gmetric. They launched on one host:
instance 1, measurement 1: 350 giraffes per minute
instance 2, measurement 1: 450 giraffes per minute
instance 1, measurement 2: 450 giraffes per minute
instance 2, measurement 2: 550 giraffes per minute
Ganglia thinks that this is one metric considering with 4 measurements:
measurement 1: 350 giraffes per minute
measurement 2: 450 giraffes per minute
measurement 3: 450 giraffes per minute
measurement 4: 550 giraffes per minute
In ganglia-web I want to see aggregated metric:
measurement 1: 800 giraffes per minute
measurement 2: 1000 giraffes per minute
One solution for that is to use spoofing. That may be cool, but what if there are 1000 instances? In ganglia there will be 1000 virtual hosts. Also, spoof should be set for each instance. That's painful. Is there a better solution for that?
Statsd with Ganglia as backend (not default graphite) was the answer.