Todd Lipcon has posted comments on this change. ( http://gerrit.cloudera.org:8080/8352 )
Change subject: tool: new actions for adding and removing data directories ...................................................................... Patch Set 3: (1 comment) http://gerrit.cloudera.org:8080/#/c/8352/2/src/kudu/tools/kudu-tool-test.cc File src/kudu/tools/kudu-tool-test.cc: http://gerrit.cloudera.org:8080/#/c/8352/2/src/kudu/tools/kudu-tool-test.cc@2330 PS2, Line 2330: NO_FATALS(RunActionStdoutNone(Substitute( > We do data dir group resolution while loading the tablet's superblock, and Imagine you have two tablets A and B: >> TS starts - create tablet A - write block 1 on drive /data/1, flush tablet A's metadata, pointing to it - create tablet B - write block 2 on drive /data/2, flush tablet B's metadata, pointing to it >>shutdown tserver >> run tool to remove /data/2 >> start tserver - LBM decides that max_block_id is '1' - tablet A bootstraps first -- immediately when done bootstrapping, it causes an MRS flush, and writes a new block id 2 (re-used) to /data/1 - tablet B bootstraps -- it thinks it has block '2' but in fact it's pointing to the new block 2 I would have expected tablet B in this case to see "ah, I have a data dir that has been removed" and mark itself as failed. But your assertion above is having it say "can't find block" which implies that it is getting past the point of checking data dir existence, right? -- To view, visit http://gerrit.cloudera.org:8080/8352 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I6ddbdd6cc6231996e7802a622a8b4691527a0643 Gerrit-Change-Number: 8352 Gerrit-PatchSet: 3 Gerrit-Owner: Adar Dembo <a...@cloudera.com> Gerrit-Reviewer: Adar Dembo <a...@cloudera.com> Gerrit-Reviewer: Andrew Wong <aw...@cloudera.com> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Todd Lipcon <t...@apache.org> Gerrit-Comment-Date: Mon, 23 Oct 2017 23:57:02 +0000 Gerrit-HasComments: Yes