Alexey Serbin has posted comments on this change. ( http://gerrit.cloudera.org:8080/16515 )
Change subject: KUDU-2612: add txn memrowsets to tablet ...................................................................... Patch Set 3: (8 comments) I did a quick first look, planning to take another look tomorrow. Just some nits and one high-level question so far. http://gerrit.cloudera.org:8080/#/c/16515/3//COMMIT_MSG Commit Message: http://gerrit.cloudera.org:8080/#/c/16515/3//COMMIT_MSG@13 PS3, Line 13: be scanned by any scanner Does it mean that those would not be scanned even if using READ_YOUR_WRITES scanner mode? I guess it's OK for first implementation, but do you think it would make sense to allow READ_YOUR_WRITES scanners to access uncommitted transaction data? http://gerrit.cloudera.org:8080/#/c/16515/3//COMMIT_MSG@36 PS3, Line 36: when the transactional MRS is flushed to disk I'm curious what happens if a kudu-tserver process crashes after flushing committing MRS, but before persisting the flag. Will kudu-tserver be able to restart, bootstart and work consistently after that? http://gerrit.cloudera.org:8080/#/c/16515/3//COMMIT_MSG@39 PS3, Line 39: to to be http://gerrit.cloudera.org:8080/#/c/16515/3/src/kudu/tablet/memrowset.h File src/kudu/tablet/memrowset.h: http://gerrit.cloudera.org:8080/#/c/16515/3/src/kudu/tablet/memrowset.h@388 PS3, Line 388: std::string( nit: is this really necessary to construct this extra string? http://gerrit.cloudera.org:8080/#/c/16515/3/src/kudu/tablet/memrowset.cc File src/kudu/tablet/memrowset.cc: http://gerrit.cloudera.org:8080/#/c/16515/3/src/kudu/tablet/memrowset.cc@715 PS3, Line 715: nit: spacing? http://gerrit.cloudera.org:8080/#/c/16515/3/src/kudu/tablet/ops/write_op.h File src/kudu/tablet/ops/write_op.h: http://gerrit.cloudera.org:8080/#/c/16515/3/src/kudu/tablet/ops/write_op.h@178 PS3, Line 178: Status AcquireTxnLockCheckOpen(scoped_refptr<Txn> txn); nit: does it make sense to add a short doc comment for this method? http://gerrit.cloudera.org:8080/#/c/16515/3/src/kudu/tablet/ops/write_op.h@285 PS3, Line 285: scoped_refptr<TxnRowSets> txn_rowsets_; nit: add a short doc comment (this seems to be the only field without one in this class) http://gerrit.cloudera.org:8080/#/c/16515/3/src/kudu/tablet/txn_participant.cc File src/kudu/tablet/txn_participant.cc: http://gerrit.cloudera.org:8080/#/c/16515/3/src/kudu/tablet/txn_participant.cc@60 PS3, Line 60: txn_lock->swap(l); nit: is there any particular reason to prefer swap() to std::move() notation here? If not, maybe it's better to switch to std::move() to be in-line with the implementation on AcquireWriteLock() ? -- To view, visit http://gerrit.cloudera.org:8080/16515 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I1dea4031c0acb875993352d452dc4c233e35a502 Gerrit-Change-Number: 16515 Gerrit-PatchSet: 3 Gerrit-Owner: Andrew Wong <[email protected]> Gerrit-Reviewer: Alexey Serbin <[email protected]> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Tidy Bot (241) Gerrit-Comment-Date: Fri, 13 Nov 2020 04:26:17 +0000 Gerrit-HasComments: Yes
