Github user anmolnar commented on the issue:
https://github.com/apache/zookeeper/pull/415
@phunt That's a perfectly valid point Pat, I've been asked about it several
times before. Therefore I added some more comments to the parent Jira of
monitoring, please take a look at the description:
https://issues.apache.org/jira/browse/ZOOKEEPER-2933
I've also populated the description of subtasks too to help understanding
them.
Regarding to statistics tracking in Zookeeper: I wonder how much does it
make sense to implementing something complicated within the server. One
approach (which this PR tries to follow) is to expose basic values which can be
sampled by a more sophisticated monitoring tool. Average, sliding window, etc.
can be implemented in there.
On the other hand we could add some more clever logic to ZK as well such as
https://google.github.io/guava/releases/21.0/api/docs/com/google/common/math/Quantiles.html
or
https://commons.apache.org/proper/commons-math/javadocs/api-3.0/org/apache/commons/math3/stat/descriptive/rank/Percentile.html
Basically calculating percentiles (50-90-99-Median-Max-etc.) with the
appropriate library and expose the values with JMX. This will probably require
us to gather samples with a sliding window within Zookeeper.
---
