Will Berkeley has posted comments on this change. ( http://gerrit.cloudera.org:8080/9722 )
Change subject: [delete_table-itest] fix flake in TestUnknownTabletsAreNotDeleted ...................................................................... Patch Set 1: (6 comments) http://gerrit.cloudera.org:8080/#/c/9722/1//COMMIT_MSG Commit Message: http://gerrit.cloudera.org:8080/#/c/9722/1//COMMIT_MSG@10 PS1, Line 10: OS X nit: s/OS X/ macOS/g http://gerrit.cloudera.org:8080/#/c/9722/1//COMMIT_MSG@10 PS1, Line 10: every second : run of the test would fail Wait, was the flake that it would fail every second run? Or you mean that's about how often we observed it fail on macOS? http://gerrit.cloudera.org:8080/#/c/9722/1//COMMIT_MSG@13 PS1, Line 13: the nit: remove http://gerrit.cloudera.org:8080/#/c/9722/1//COMMIT_MSG@15 PS1, Line 15: However, with 10ms : heartbeat interval in the tserver is able to get a response from the : restarted master before the master re-generated its IPKI records. I'm having trouble parsing this sentence. I think you mean "However, with a 10ms heartbeat interval, the tserver is able to get a response from the restarted master before the master regenerates its IPKI records." Is that right? If so, would you mind editing the sentence to , and also explaining how this causes the test to fail? Actually, on my machine, I saw it get stuck in a tight loop, repeatedly failing negotiation; presumably that's due to the 10ms heartbeat interval. http://gerrit.cloudera.org:8080/#/c/9722/1/src/kudu/integration-tests/delete_table-itest.cc File src/kudu/integration-tests/delete_table-itest.cc: http://gerrit.cloudera.org:8080/#/c/9722/1/src/kudu/integration-tests/delete_table-itest.cc@1097 PS1, Line 1097: NOTE on disabled RPC authentication and encryption: Ah ok, I understand the issue based on this comment. Could you add an abbreviated version of this explanation to the commit message? I think it would be helpful to have that in the git log. http://gerrit.cloudera.org:8080/#/c/9722/1/src/kudu/integration-tests/delete_table-itest.cc@1104 PS1, Line 1104: CA private key. If the tserver does heartbeat before the master regenerates the IPKI stuff, why does the test pass? In that case, does the tablet server count as registered? Shouldn't we refuse to register in that situation? -- To view, visit http://gerrit.cloudera.org:8080/9722 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: Ib6fd439c0ef5fb66b752f7f49175e4c2d818412e Gerrit-Change-Number: 9722 Gerrit-PatchSet: 1 Gerrit-Owner: Alexey Serbin <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Grant Henke <[email protected]> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Will Berkeley <[email protected]> Gerrit-Comment-Date: Tue, 20 Mar 2018 06:36:10 +0000 Gerrit-HasComments: Yes
