> On March 18, 2014, 5:29 p.m., kturner wrote: > > src/server/src/main/java/org/apache/accumulo/server/test/randomwalk/concurrent/CheckBalance.java, > > line 100 > > <https://reviews.apache.org/r/19352/diff/2/?file=526277#file526277line100> > > > > could use o.a.a.core.util.Stat > > Bill Havanki wrote: > I just looked at Stat. I don't think it calculates standard deviation > correctly, actually. > > Bill Havanki wrote: > For future reference: See ACCUMULO-2494 for discussion. The calculation > in Stat is OK but not great. I'll leave this one as is.
Could use the commons math function you referenced in the 2494 code review instead of Stat. - kturner ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/19352/#review37568 ----------------------------------------------------------- On March 18, 2014, 6:43 p.m., Bill Havanki wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/19352/ > ----------------------------------------------------------- > > (Updated March 18, 2014, 6:43 p.m.) > > > Review request for accumulo and Mike Drob. > > > Bugs: ACCUMULO-2488 > https://issues.apache.org/jira/browse/ACCUMULO-2488 > > > Repository: accumulo > > > Description > ------- > > The Concurrent randomwalk test used to consider servers unbalanced if any > server's tablet count differed from the cluster average by more than a fifth > of the average or by one, whichever was larger. This would cause failures > under typical balancings from the default balancer. > > This commit changes the criterion for an unbalanced server to be double the > standard deviation from the cluster average. > > > Diffs > ----- > > > src/server/src/main/java/org/apache/accumulo/server/test/randomwalk/concurrent/CheckBalance.java > d00e2b4 > > Diff: https://reviews.apache.org/r/19352/diff/ > > > Testing > ------- > > Ran Concurrent randomwalk three times on 7-node cluster. > > > Thanks, > > Bill Havanki > >
