Andrew Wong has posted comments on this change. ( http://gerrit.cloudera.org:8080/17055 )
Change subject: [test] Set PROCESSORS to 4 for maintenance_mode-itest ...................................................................... Patch Set 1: Code-Review+2 > Patch Set 1: > > > > Patch Set 1: > > > > > > These are tests that use external mini-clusters with up to 6 > > tablet servers. It makes sense that they use more CPU resources. > > They are already tagged as a slow running test (requiring > > KUDU_ALLOW_SLOW_TESTS=1 to run) and most other itests also set a > > PROCESSORS value > 1. The default PROCESSORS is 1 so ctest will > > assume that a single test shard only uses 1 core. > > > > > > I think not setting PROCESSORS > 1 was likely just an oversight > > given we almost always run the tests on dist-test or more powerful > > developer machines. > > > > Thanks for the explanation! > > > > So why does test fail with 1 processor? Timeout settings are too > > aggressive? > > I guess we have many other tests which are running at least 4 processes: one > master, three tablet servers: most of those tests are resilient enough to run > on one- or two-core CPU. Indeed, maybe it's possible to make > > > > Patch Set 1: > > > > > > These are tests that use external mini-clusters with up to 6 > > tablet servers. It makes sense that they use more CPU resources. > > They are already tagged as a slow running test (requiring > > KUDU_ALLOW_SLOW_TESTS=1 to run) and most other itests also set a > > PROCESSORS value > 1. The default PROCESSORS is 1 so ctest will > > assume that a single test shard only uses 1 core. > > > > > > I think not setting PROCESSORS > 1 was likely just an oversight > > given we almost always run the tests on dist-test or more powerful > > developer machines. > > > > Thanks for the explanation! > > > > So why does test fail with 1 processor? Timeout settings are too > > aggressive? > > We have many other tests which are running at least one master and three > tablet servers, and others which run three masters and three tablet servers, > but it seems the majority of those are stable enough to pass even if run on 2 > or 1 core CPUs. Indeed, maybe it's possible to harden maintenance_mode-itest > to be less flaky in such environments? I can work on deflaking this test, but happy to merge this since it should still help. -- To view, visit http://gerrit.cloudera.org:8080/17055 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I89b479374266a64edf7fd455837c4d57a684694b Gerrit-Change-Number: 17055 Gerrit-PatchSet: 1 Gerrit-Owner: Grant Henke <[email protected]> Gerrit-Reviewer: Alexey Serbin <[email protected]> Gerrit-Reviewer: Andrew Wong <[email protected]> Gerrit-Reviewer: Bankim Bhavsar <[email protected]> Gerrit-Reviewer: Grant Henke <[email protected]> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Comment-Date: Wed, 10 Feb 2021 18:04:07 +0000 Gerrit-HasComments: No
