Andrew Wong has posted comments on this change. ( http://gerrit.cloudera.org:8080/10693 )
Change subject: parse_metrics_log: update to the new format ...................................................................... Patch Set 2: > Patch Set 2: > > (4 comments) > > I guess the histogram-parsing code was "off" because now we only include > buckets that have changed/are non-zero? > > Did you test it against some actual logs to make sure it has the expected > results? > > Also, wondering if we should bite the bullet soon and port this to C++ so it > runs faster and ships with kudu. Re: the histogram-parsing code, I'm confused about why it was calculating the percentiles using the delta between the previous and current snapshots. I would've expected it to be the aggregate of the two instead. E.g. in the case the total count stayed the same between snapshots, but the distribution changed, we would return UNKNOWN_PERCENTILES, but it seems reasonable still compute percentiles since we still have some distribution of the metric at that time, no? -- To view, visit http://gerrit.cloudera.org:8080/10693 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: If11b7ecc93a3f64db3b7c1994f47308b3ec44029 Gerrit-Change-Number: 10693 Gerrit-PatchSet: 2 Gerrit-Owner: Andrew Wong <[email protected]> Gerrit-Reviewer: Andrew Wong <[email protected]> Gerrit-Reviewer: David Ribeiro Alves <[email protected]> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Todd Lipcon <[email protected]> Gerrit-Comment-Date: Wed, 13 Jun 2018 00:44:10 +0000 Gerrit-HasComments: No
