HoustonPutman commented on PR #1650: URL: https://github.com/apache/solr/pull/1650#issuecomment-1584636842
Awesome, I'm so glad this weighted approach is a good path to go down. Should be even easier to add new PlacementPlugins, and we get balanceReplicas for "free". I can't even imagine trying to implement the balance API using the previous affinityPlugin logic... > I was thinking a cool feature in the future we could have some way of simulating a change (via API maybe, like a dry-run) and get an explain back of why a node was chosen. A dry run would be great. (And very much mimicks something I have done for a Solr replica management system in the past, it was very useful). Explanations would also be great, but I think even harder given the weighted nature of these moves. "Why" is hard to elaborate on when we just have a single number for a weight. But maybe a weighted node could give explanations for the weight... Would be an interesting addition for sure. > The AffinityPlacement could use some more comments Absolutely, will do much more commenting, now that I know I'm not going down a rabbit hole. > Have you looked at the "scalability" test after the changes? When working on the spread domain changes I noticed some of the implementations I attempted were causing a bit impact in performance there. Yes, and I love that its there to help keep us in line performance-wise. I ran the test 20 times via IntelliJ, and got the following numbers: - 20.5 seconds - With this patch - 31.75 seconds - Without the patch (main) So it looks like a 50% improvement in speed, which I'm honestly very surprised by. I thought it would have been slower... -- 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: issues-unsubscr...@solr.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@solr.apache.org For additional commands, e-mail: issues-h...@solr.apache.org
