A problem about flink latency metrics

[Lee tinker]

Hi all :

I have a problem about flink latency metrics. It is described by this link — “https://ci.apache.org/projects/flink/flink-docs-release-1.7/monitoring/metrics.html#io”.

As  it described above , “<source_id>.<source_subtask_index>.<operator_id>.<operator_subtask_index>.latency” means “The latency distributions from a given source subtask to an operator subtask (in milliseconds).” But  when i get  flink metrics by using graphite,  i found  the metrics value of different  operator are the same!  What make me more confused is  that when  i  compare two  different  flink job , the  values  are all the same too!  I want to know how this value is calculated and why the collected values are the same.  

Can someone  help me to explain that ?  Appreciate it  

Thanks 

Hanmiao Li