morningman commented on PR #69525: URL: https://github.com/apache/doris/pull/69525#issuecomment-6365952000
Pushed b39b34ec859: - `PublishVersionDaemon.tryFinishOneTxn` now skips a transaction that a publish executor is finishing (`publishingTxnIds`). Since this PR, `setTableIdToTabletDeltaRows` takes the monitor of the transaction state, and the finish holds that monitor while it serializes the state (or, with `enable_txn_log_outside_lock` off, while it writes the journal). The daemon thread could wait there and hold up every other publish. The finish already uses the delta rows of the round that started it, so setting them again is not needed. New test: `TransactionStateTest#testPublishDaemonSkipsTransactionBeingFinished`. - The PR description has a new section on the memory, CPU and lock cost of serializing in the submitting thread, with measured serialization times for transactions of several sizes. The FE unit tests listed in the description pass (119 tests, 0 failures), and checkstyle reports no violations. -- 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]
