Todd Lipcon has posted comments on this change. Change subject: KUDU-1760. Add test coverage of reading pre-REINSERT after ALTER ......................................................................
Patch Set 1: (1 comment) http://gerrit.cloudera.org:8080/#/c/5488/1/src/kudu/integration-tests/alter_table-test.cc File src/kudu/integration-tests/alter_table-test.cc: PS1, Line 891: this seems odd that the : // "latest observed" timestamp is one higher than the thing we wrote > you mean from the write response, right? sorry, I think I described the problem backwards. I think the client _is_ getting the exact timestamp of its write as the "latest observed timestamp". However, a READ_AT_SNAPSHOT with the SetSnapshotRaw equal to that time is not seeing the row inserted at that time (ts1 is the time of the latest insert). So maybe when we are creating the snapshot in the scan code, we are off-by-one somewhere in determining what's committed? I see that consistency-itest works around the same "bug" by adding a + 1 to the timestamps provided to the GetRowCount() helper. Should I file a JIRA about this? -- To view, visit http://gerrit.cloudera.org:8080/5488 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: Id7d44f6e2fe4d3cae03e0024c70e9f2ae84b614c Gerrit-PatchSet: 1 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Todd Lipcon <t...@apache.org> Gerrit-Reviewer: David Ribeiro Alves <dral...@apache.org> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Mike Percy <mpe...@apache.org> Gerrit-Reviewer: Todd Lipcon <t...@apache.org> Gerrit-HasComments: Yes