> On Feb. 3, 2016, 10:19 p.m., Maxim Khutornenko wrote: > > src/main/java/org/apache/aurora/scheduler/storage/db/DbStorage.java, line > > 138 > > <https://reviews.apache.org/r/43150/diff/1/?file=1231802#file1231802line138> > > > > Recommend using statsProvider.untracked() here. > > Zameer Manji wrote: > What's the reason for that? > > Maxim Khutornenko wrote: > I just see no reason not to :) Unless you want to observe them in > /graphview... > > Bill Farner wrote: > Unless your external stats system polls pretty frequently, O(seconds), > you might actually miss useful activity in these gauges. > > Maxim Khutornenko wrote: > This is the part I am not clear, does the time series matter for the > stats collection? Does not it only see the immediate value at the time of > collection? > > John Sirois wrote: > I think that's Bill's point. Since Guages are instantaneous, if the > external sampling is, say, every 60s, and the interesting pool starvation > happens 10s but then corrects 10s later at the 20s mark since last > collection, you'll not see it ... unless you have tracking enabled and > collection at - default - 1s. Even then, you'll only see it via /graphview > > ... rusty recollections warning ... > > Bill Farner wrote: > John's recollection is correct.
Aha, then we are on the same page. I guess the ideal would be collecting max counters instead? That way we would not miss anything and catch spikes with rate()? - Maxim ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/43150/#review117687 ----------------------------------------------------------- On Feb. 3, 2016, 10:23 p.m., Zameer Manji wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/43150/ > ----------------------------------------------------------- > > (Updated Feb. 3, 2016, 10:23 p.m.) > > > Review request for Aurora, John Sirois and Maxim Khutornenko. > > > Bugs: AURORA-1607 > https://issues.apache.org/jira/browse/AURORA-1607 > > > Repository: aurora > > > Description > ------- > > To better understand the MyBatis connection pool this patch exposes the pool > state via stats. > > > Diffs > ----- > > src/main/java/org/apache/aurora/scheduler/storage/db/DbStorage.java > 7674b8af6d18b39153ecdf02015f0970e035e874 > src/test/java/org/apache/aurora/scheduler/storage/db/DbStorageTest.java > 3ab95c615b01f3201e3b87089119abf01d71dbb7 > > Diff: https://reviews.apache.org/r/43150/diff/ > > > Testing > ------- > > ./gradlew build -Pq > Checked the metrics in vagrant > > > Thanks, > > Zameer Manji > >