NealSun96 commented on issue #772: Unexpected randomness in the WAGED rebalancer partition assignment URL: https://github.com/apache/helix/issues/772#issuecomment-598350263 Update to the issue: after experiments and discussions, it is concluded that the divergence is not caused by "unexpected randomness" in the algorithm. When the IdealState is changed (which is necessary to scale a resource up or down), the Baseline is recalculated; due to the design of the algorithm, the new Baseline could be very different from the old Baseline depending on the order of resources. Therefore, the divergence is caused by the change of Baseline, while the Best Possible is respecting the old Baseline. As a result, to combat divergence, the answer lies within the algorithm design; there is no clean-cut solution that "eliminates divergence". We cannot eliminate the difference between old Baseline and new Baseline because eliminating it is minimizing movement while sacrificing replica evenness - we need to achieve a balance between the two factors. I will continue to do research on this topic. In terms of the scope of this issue, the tie-breaking code change will still go through to make scoring deterministic.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
