mufiye commented on issue #10341: URL: https://github.com/apache/skywalking/issues/10341#issuecomment-1440051786
> I just read https://opentelemetry.io/docs/reference/specification/metrics/data-model/#exponentialhistogram, it seems it is just the typical Prometheus Histogram setup in practice. > > Back to you question > > > Should I make this exponentialHistogram type be supported or use summary type for timer metric? > > We should transfer this to our histogram, I think. You need to get the bucket transfer correctly from `exponentialHistogram` to `histogram`. I will try to do it later. And there are some other essential points that need to be discussed. 1. There are some labels in airflow metric names such as <job_name>, <task_id>, <dag_id>, <operator_name>, <pool_name> which represent the components in airflow. Which level should I classify these components to? In airflow concepts, `dag` contains lots of `tasks` to be run, `pool` is where tasks run in, `operator_name` is one kind of `task` and `job` I think is a larger concept than a task because it also includes the scheduler job. I think we classify all these components as endpoints? 2. As before, we transform the "delta counter metric" to "gauge metric", we can represent some monotonic delta metric as the current metric in one specific period of time. For example, `ti_successes` metric describes "Overall task instances successes", so we can think the gauge metric as "the successful task instances in this period". But for some nonmonotonic delta counter metric, such as "dag_processing.processes" which means number of currently running DAG parsing processes, it can be negative in one period in my test. How should we handle this kind of metric? Maybe just show them originally for showing the trend of the value. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
