----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/61126/#review184381 -----------------------------------------------------------
3rdparty/libprocess/include/process/metrics/timer.hpp Lines 45-46 (patched) <https://reviews.apache.org/r/61126/#comment260535> If we need a "sum", can we just add a field "sum"? Can we keep "value" to mean what it used to mean for backwards compatibility? It doesn't hurt right? Also, why do we (which must be Prometheus' rationale as well) need a sum? What useful thing can we do about it? This is related to the abstract concept of "counting semantics" that we need to document in a design. We need a whiteboard discussion offline but perhaps we should write it down for other reviewers. - Jiang Yan Xu On July 25, 2017, 5:03 p.m., James Peach wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/61126/ > ----------------------------------------------------------- > > (Updated July 25, 2017, 5:03 p.m.) > > > Review request for mesos, Benjamin Mahler, Kevin Klues, and Jiang Yan Xu. > > > Bugs: MESOS-6918 > https://issues.apache.org/jira/browse/MESOS-6918 > > > Repository: mesos > > > Description > ------- > > The value of a Timer was the elapsed time of the last timed > interval which is not a value that can be easily graphed > or interpreted. Replaced this with a counter which tracks > the cumulative elapsed time of all the measurements on this > Timer. This allows a monitoring application to plot the elapsed > time as a rate. > > > Diffs > ----- > > 3rdparty/libprocess/include/process/metrics/timer.hpp > 0a9c0227c457c6c81a59f65f901a5464ee00983d > 3rdparty/libprocess/src/tests/metrics_tests.cpp > 161ca0dc7aea526d450d71a80839d8cc075aaa31 > > > Diff: https://reviews.apache.org/r/61126/diff/3/ > > > Testing > ------- > > make check (Fedora 26). Note that this requires Mesos test changes in r/61127. > > > Thanks, > > James Peach > >
