> On July 27, 2013, 2:28 a.m., Kishore Gopalakrishna wrote: > > Have you tested for the case where N>P but number of livenodes<P and vice > > versa
Yes, I tried it for a few basic cases. It shouldn't perform worse in the case you brought up because the same number of partitions will have their preferred node live, and as more nodes come up, there should be roughly the same amount of initial churn and less eventual churn. - Kanak ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/12984/#review24005 ----------------------------------------------------------- On July 26, 2013, 11:23 p.m., Kanak Biscuitwala wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/12984/ > ----------------------------------------------------------- > > (Updated July 26, 2013, 11:23 p.m.) > > > Review request for helix, Zhen Zhang, Kishore Gopalakrishna, and Shi Lu. > > > Bugs: HELIX-148 > > > Repository: helix-git > > > Description > ------- > > Fix for HELIX-148: when there are fewer partitions than nodes, then use a > different placement strategy to ensure every node has some preferred > partitions > > > Diffs > ----- > > > helix-core/src/main/java/org/apache/helix/controller/strategy/AutoRebalanceStrategy.java > 740d17d > > Diff: https://reviews.apache.org/r/12984/diff/ > > > Testing > ------- > > Ran my end-to-end test to check for regressions > > > Thanks, > > Kanak Biscuitwala > >
