gianm opened a new pull request, #12679: URL: https://github.com/apache/druid/pull/12679
This reverts commit 8fbf92e047f792ff1c69bf67d14784ac55eee88f (PR #12600). While that patch fixed a correctness bug, it also introduced a performance regression: the number of segments returned from the metadata store to Druid servers increases substantially. We observed noticeable slowdowns on Overlord APIs as a result of this. The main problem is that 4-digit, 5-digit, 6-digit, etc, years are all jumbled up when sorted lexicographically. We can't write a range filter (such that metadata store indexes can be used) without also potentially returning a bunch of extra records that need to be filtered out on the Overlord. I can't see a good way to solve the original correctness bug while maintaining performance, other than introducing bigint-typed versions of `start` and `end`. So I think we'll need to do that, unless someone else has a better idea. -- 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]
