Gabriel39 commented on PR #69768: URL: https://github.com/apache/doris/pull/69768#issuecomment-6649267841
Addressed the three points from https://github.com/apache/doris/pull/69768#issuecomment-6598121258 in 42fa718ef6c, with the same changes synchronized to #69779 in c2c8025f091. - Compare actual service times within eight request-size buckets. Each complete remote call contributes its own size and latency; cache writeback remains outside that timer. - Own the controller in each BlockFileCache, isolating healthy and congested cache paths while sharing observations across queries on the same path. - Replace synthetic minimum feedback with real requests through the unselected source. Across all buckets of one cache path, allow at most one probe in flight, at least one second between probe admissions, and at least 64 eligible full-hit decisions between probes. Every success, error, short-read and cancellation exit releases the permit. Full-range fallback, HDFS exclusion and per-reader optional-remote failure isolation are preserved. Tests cover mixed sizes, individual remote calls, separate cache paths, recovery with no previous fast local sample, remote-history bootstrap, rate/concurrency limits and exceptional cleanup. Local ASan validation passed 104 BE tests on branch-4.2 and 96 distinct BE tests on master; four FE session/wire tests and FE Checkstyle passed on the rebased branch. clang-format 16 and changed-code clang-tidy passed. Both PRs have new buildall and review requests; full CI validation is pending. -- 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]
