Adar Dembo has posted comments on this change. ( http://gerrit.cloudera.org:8080/14977 )
Change subject: KUDU-3011 p2: mechanism to quiesce leadership ...................................................................... Patch Set 2: (2 comments) http://gerrit.cloudera.org:8080/#/c/14977/2/src/kudu/consensus/raft_consensus.cc File src/kudu/consensus/raft_consensus.cc: http://gerrit.cloudera.org:8080/#/c/14977/2/src/kudu/consensus/raft_consensus.cc@444 PS2, Line 444: return Status::IllegalState("leader elections are disabled"); Will it be possible to quiesce a tserver as it is bootstrapping replicas? If so, it looks like this failure will percolate up into outright replica failure, which seems undesirable. Though I guess this also applies to new replica creation: a failure here will cause the newly created replica to be marked as failed (see RaftConsensus::Start). http://gerrit.cloudera.org:8080/#/c/14977/2/src/kudu/tserver/tablet_server.h File src/kudu/tserver/tablet_server.h: http://gerrit.cloudera.org:8080/#/c/14977/2/src/kudu/tserver/tablet_server.h@107 PS2, Line 107: mutable simple_spinlock quiescing_lock_; : bool quiescing_; : Make it an std::atomic and forgo the lock? Then you could also plumb the member directly into the replicas rather than indirecting thru a functor, right? -- To view, visit http://gerrit.cloudera.org:8080/14977 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I0eb3778fe094647bc5f61f97971087d7efb8af5e Gerrit-Change-Number: 14977 Gerrit-PatchSet: 2 Gerrit-Owner: Andrew Wong <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Tidy Bot (241) Gerrit-Comment-Date: Mon, 06 Jan 2020 05:36:16 +0000 Gerrit-HasComments: Yes
