Inaccurate results in Prometheus due to counter resets for longer sampling intervals

[aishwaryaraimule21]

Currently, the approach is to use the maximum value of each interval to calculate the rate of change.
In case of a Counter reset inside the bucket due to a process restart,
the count delta = max of current interval - max of previous interval
This is a problem because we are ignoring any requests received after counter reset (if that number of count < count before reset).
This approximation works fine for shorter sampling intervals, for larger sampling intervals this results in an higher error.

[welcome]

Thanks for opening this issue. A team member should give feedback soon. In the meantime, feel free to check out the contributing guidelines.