sollhui opened a new pull request, #66894: URL: https://github.com/apache/doris/pull/66894
### What problem does this PR solve? Issue Number: None Related PR: #62007 #### Problem summary When row binlog is enabled, a `GROUP` request flushes both the data memtable and the row-binlog memtable. These are two independent physical flush tasks, and both increment `flush_running_count`. Write backpressure previously compared this physical task count with the same configured limit used by a normal data-only memtable. As a result, row-binlog imports effectively had half of the configured logical flush concurrency. With a limit of 2, one group memtable could consume the entire allowance and subsequent writes had to wait, making memtable writing and flushing nearly serial. #### Changes - Keep the existing `flush_running_count` backpressure mechanism for storage-compute integrated mode. - Use `memtable_flush_running_count_limit * 2` for `GROUP` requests because each logical row-binlog memtable creates two physical flush tasks. - Keep the configured limit unchanged for normal data-only requests. #### Impact Row-binlog imports can use the intended logical memtable flush concurrency and keep memtable writing overlapped with flushing. Imports without row binlog are unchanged. ### Release note None ### Check List (For Author) - Test: Not run (per request) - Behavior changed: Yes. Row-binlog `GROUP` requests use an effective physical flush-task limit of twice the configured logical limit. - Does this need documentation: No -- 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]
