I'm trying to measure the number of requests that my Broker serve and I found various MBeans that report a metric for that. However I looked at them and they all seem to be report very different numbers. I'm clueless why they are so different.
For clarification: I export the `Count` attribute so that I can run calculate the per second values on my own in Prometheus. For instance: `kafka.network.RequestMetrics,name=TotalTimeMs,request=Fetch<Count>` vs. `kafka.network.RequestMetrics,name=ErrorsPerSec,request=Fetch,error=*<Count>` (MBean name might be missleading, but since there is an error=NONE label, this should report the same when summed across all errors) vs. `kafka.server.BrokerTopicMetrics,name=TotalFetchRequestsPerSec<Count>`
