helifu has posted comments on this change. ( http://gerrit.cloudera.org:8080/12254 )
Change subject: KUDU-2665: deflake block_manager-stress-test ...................................................................... Patch Set 1: (1 comment) http://gerrit.cloudera.org:8080/#/c/12254/1/src/kudu/fs/block_manager-stress-test.cc File src/kudu/fs/block_manager-stress-test.cc: http://gerrit.cloudera.org:8080/#/c/12254/1/src/kudu/fs/block_manager-stress-test.cc@550 PS1, Line 550: // Quiesce the block manager before injecting inconsistencies so that the two : // don't interfere with one another. : this->bm_.reset(); I have just read through the code of 'block_manager-stress-test'. I agree with adar's analysis that it's now incorrect to assume that a cessation in user threads implies an end to LBM activity. Please take a look at the code of log_block_manager-test-util.cc:Line70/Line95, it is not safe while the threads in 'dd_manager_' are deleting dead container via punch hole and drop container's last ref. And i am sure Percy's issue is from Line95. Maybe we can have two ways: 1)destroy the 'dd_manager_' before injecting inconsistencies, 2)check data file existence at Line95. Sorry for leaving this issue:( -- To view, visit http://gerrit.cloudera.org:8080/12254 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I0be328f740056cd6b64c9881759225c8b961a935 Gerrit-Change-Number: 12254 Gerrit-PatchSet: 1 Gerrit-Owner: Adar Dembo <a...@cloudera.com> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Mike Percy <mpe...@apache.org> Gerrit-Reviewer: helifu <hzhel...@corp.netease.com> Gerrit-Comment-Date: Wed, 23 Jan 2019 03:11:47 +0000 Gerrit-HasComments: Yes