Gabriella Lotz has posted comments on this change. ( http://gerrit.cloudera.org:8080/24863 )
Change subject: KUDU-3761 drop default for prefer_follower_moves ...................................................................... Patch Set 3: (3 comments) http://gerrit.cloudera.org:8080/#/c/24863/2/src/kudu/rebalance/rebalance_algo-test.cc File src/kudu/rebalance/rebalance_algo-test.cc: http://gerrit.cloudera.org:8080/#/c/24863/2/src/kudu/rebalance/rebalance_algo-test.cc@221 PS2, Line 221: /*prefer_follower_moves=*/true > Don't see any unit test with 'prefer_follower_moves=false'. Any reason for There is a false case, just not through this helper: PreferFollowerMovesByTableAndTag covers false for TwoDimensionalGreedyAlgo, LocationBalancingPrefersFollowerSource covers both values for the location algorithm, and RebalancePreferFollowerTest in rebalancer_tool-test.cc is parameterized over both. For this helper specifically I don't think a false variant would add much: ClusterConfigToClusterInfo doesn't populate ts_with_followers_by_table_and_tag (only BuildClusterInfo does), so with prefer_follower_moves=true no source ever qualifies as hosting a follower and we always end up on the leader fallback, which picks the same move as the false branch. These configs use PICK_FIRST, so there's no shuffling either and both values are deterministic and identical here. The behavioral difference only shows up once the follower map is filled in, which is what PreferFollowerMovesByTableAndTag does. http://gerrit.cloudera.org:8080/#/c/24863/2/src/kudu/rebalance/rebalance_algo.h File src/kudu/rebalance/rebalance_algo.h: http://gerrit.cloudera.org:8080/#/c/24863/2/src/kudu/rebalance/rebalance_algo.h@a198 PS2, Line 198: > question - Why not keep this, to prevent implicit conversion from braced in Good point, I have restored it. http://gerrit.cloudera.org:8080/#/c/24863/2/src/kudu/rebalance/rebalance_algo.h@198 PS2, Line 198: // 'opt' decides how ties are broken when several tables or servers are : // equally skewed: PICK_FIRST is deterministic, PICK_ > nit: Maybe add a comment above providing guidance on impact of both argumen Done -- To view, visit http://gerrit.cloudera.org:8080/24863 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I2a86f1345b3adad084813e1b61a4a38ef566d035 Gerrit-Change-Number: 24863 Gerrit-PatchSet: 3 Gerrit-Owner: Gabriella Lotz <[email protected]> Gerrit-Reviewer: Ashwani Raina <[email protected]> Gerrit-Reviewer: Gabriella Lotz <[email protected]> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Marton Greber <[email protected]> Gerrit-Comment-Date: Tue, 22 Sep 2026 12:14:04 +0000 Gerrit-HasComments: Yes
