[
https://issues.apache.org/jira/browse/HBASE-24367?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17112521#comment-17112521
]
Andrew Kyle Purtell commented on HBASE-24367:
---------------------------------------------
What I plan to do:
- Log the elapsed time in milliseconds, which most operators are very familiar
with.
- Remove the log lines for the average time, which is not interesting, and can
be derived from logs using any of the common solutions for log scraping that
operators will use. I agree with Nick that a histogram might be better, but it
should be exported as a real metric, not dumped in logs, in that case.
> ScheduledChore log elapsed timespan in a human-friendly format
> --------------------------------------------------------------
>
> Key: HBASE-24367
> URL: https://issues.apache.org/jira/browse/HBASE-24367
> Project: HBase
> Issue Type: Task
> Components: master, regionserver
> Affects Versions: 2.3.0
> Reporter: Nick Dimiduk
> Assignee: Andrew Kyle Purtell
> Priority: Minor
> Fix For: 3.0.0-alpha-1, 2.3.0, 2.4.0, 2.2.5
>
>
> I noticed this in a log line,
> {noformat}
> 2020-04-23 18:31:14,183 INFO org.apache.hadoop.hbase.ScheduledChore:
> host-a.example.com,16000,1587577999888-ClusterStatusChore average execution
> time: 68488258 ns.
> {noformat}
> I'm not sure if there's a case when elapsed time in nanoseconds is meaningful
> for these background chores, but we could do a little work before printing
> the number and time unit to truncate precision down to something a little
> more intuitive for operators. This number purports to be an average, so a
> high level of precision isn't necessarily meaningful.
> Separately, or while we're here, if we think an operator really cares about
> the performance of this chore, we should print a histogram of elapsed times,
> rather than an opaque average.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
