Github user eperott commented on the pull request:
https://github.com/apache/cassandra/commit/2e90259669bddf04b6c1dba38b604aa6a33dcd47#commitcomment-30417260
In
src/java/org/apache/cassandra/metrics/DecayingEstimatedHistogramReservoir.java:
In
src/java/org/apache/cassandra/metrics/DecayingEstimatedHistogramReservoir.java
on line 419:
If I remember, reason for simply returning size of bucket array was that
most (if not all) implementations in the metrics library was actually returning
a value which represented the size of the container holding samples rather than
the number of recorded values. So was just trying to stay consistent with that.
A better option would perhaps have been to return the number of samples being
considered in the snapshot, but that would have been non-trivial (I think) when
taking the forward decay into account. Changing behavior here would possibly
have side effects within the metrics library, but I will have to check a little
more to tell.
I'm not sure I understand you comment wrt count? Are you saying all
snapshots get a count() of zero?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
