jtuglu1 commented on PR #19422: URL: https://github.com/apache/druid/pull/19422#issuecomment-4580871835
> Have you tried this out in a real cluster? I wonder if we should have a property to restore the old behavior in case something undesirable happens. > > I also struggle with the naming. The term "normalized" makes me think that costs will be scaled somehow into some bounded range. The adjusted cost function in the PR is doing something different from normalization, it's blowing up the cost as servers fill up. > > I wonder if this could all be cleaner by having the cost adjustments be implemented as options to the `cost` balancer rather than entirely distinct balancer strategies. What are your thoughts on this? > > Btw, the docs would need to be updated with this formula change, since the "multiplies the cost of placing a segment on a server by the server's disk usage ratio (diskUsed / maxSize)" is no longer what happens. @gianm old behavior was already broken (docs recommended it not be used in production). I'm attempting to make this strategy actually usable. > I wonder if this could all be cleaner by having the cost adjustments be implemented as options to the cost balancer rather than entirely distinct balancer strategies. What are your thoughts on this? I think this is fine, but I'd rather fix it in isolation for backwards compatibility first. We can then see if we can unify both `cost` and `diskNormalized`. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
