> On Sept. 26, 2017, 1:08 a.m., Gilbert Song wrote: > > src/tests/containerizer/cgroups_isolator_tests.cpp > > Line 445 (original), 444 (patched) > > <https://reviews.apache.org/r/62553/diff/1/?file=1834730#file1834730line445> > > > > The flaky behavior is still reproducable: > > ``` > > [ RUN ] CgroupsIsolatorTest.ROOT_CGROUPS_CFS_EnableCfs > > I0926 01:07:11.409624 20616 exec.cpp:162] Version: 1.5.0 > > I0926 01:07:11.555510 20637 exec.cpp:237] Executor registered on agent > > e36d5653-eee2-4f69-92ee-3d8aec9a3cdc-S0 > > I0926 01:07:11.563771 20636 executor.cpp:171] Received SUBSCRIBED event > > I0926 01:07:11.565313 20636 executor.cpp:175] Subscribed executor on > > vagrant-ubuntu-wily-64 > > I0926 01:07:11.566074 20636 executor.cpp:171] Received LAUNCH event > > I0926 01:07:11.567462 20636 executor.cpp:633] Starting task > > 7995dc33-b190-4295-b643-b274f278ad50 > > I0926 01:07:11.583531 20636 executor.cpp:477] Running > > '/vagrant/mesos/build/src/mesos-containerizer launch > > <POSSIBLY-SENSITIVE-DATA>' > > I0926 01:07:11.594823 20636 executor.cpp:646] Forked command at 20639 > > ../../src/tests/containerizer/cgroups_isolator_tests.cpp:444: Failure > > Expected: (0.35) >= (cpuTime), actual: 0.35 vs 0.39 > > *** Aborted at 1506388031 (unix time) try "date -d @1506388031" if you > > are using GNU date *** > > PC: @ 0x263f886 testing::UnitTest::AddTestPartResult() > > *** SIGSEGV (@0x0) received by PID 17734 (TID 0x7f106c83a800) from PID > > 0; stack trace: *** > > @ 0x7f1062d24d10 (unknown) > > @ 0x263f886 testing::UnitTest::AddTestPartResult() > > @ 0x263f409 testing::internal::AssertHelper::operator=() > > @ 0x246ebd7 > > mesos::internal::tests::CgroupsIsolatorTest_ROOT_CGROUPS_CFS_EnableCfs_Test::TestBody() > > @ 0x267d903 > > testing::internal::HandleSehExceptionsInMethodIfSupported<>() > > @ 0x266a467 > > testing::internal::HandleExceptionsInMethodIfSupported<>() > > @ 0x26495d5 testing::Test::Run() > > @ 0x264a301 testing::TestInfo::Run() > > @ 0x264aa17 testing::TestCase::Run() > > @ 0x2652418 testing::internal::UnitTestImpl::RunAllTests() > > @ 0x267ae33 > > testing::internal::HandleSehExceptionsInMethodIfSupported<>() > > @ 0x266c457 > > testing::internal::HandleExceptionsInMethodIfSupported<>() > > @ 0x26520d5 testing::UnitTest::Run() > > @ 0x1619be1 RUN_ALL_TESTS() > > @ 0x161861d main > > @ 0x7f106296aa40 (unknown) > > @ 0xaec8a9 _start > > I0926 01:07:11.910209 20630 exec.cpp:517] Agent exited ... shutting down > > I0926 01:07:11.910732 20630 executor.cpp:171] Received SHUTDOWN event > > I0926 01:07:11.911000 20630 executor.cpp:743] Shutting down > > I0926 01:07:11.911231 20630 executor.cpp:850] Sending SIGTERM to > > process tree at pid 20639 > > ``` > > > > Should we give a 200ms buffer instead?
Yes, that's fine with me. But I don't love this test. The real crux here is that this test is really dependent on the environment, i.e., what is running on that machine. Is there a better way to do this? Is there something else we can "test" to ensure that CFS has been turned on, rather than trying to actually exercise the CFS algorithm? - Benjamin ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/62553/#review186190 ----------------------------------------------------------- On Sept. 25, 2017, 10:47 p.m., Benjamin Hindman wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/62553/ > ----------------------------------------------------------- > > (Updated Sept. 25, 2017, 10:47 p.m.) > > > Review request for mesos, Gilbert Song and Jie Yu. > > > Repository: mesos > > > Description > ------- > > Fixed a flaky test. > > > Diffs > ----- > > src/tests/containerizer/cgroups_isolator_tests.cpp > 3fc93417f2d3febf2feca3ec1c8476c9edcfbf4d > > > Diff: https://reviews.apache.org/r/62553/diff/1/ > > > Testing > ------- > > make check > > > Thanks, > > Benjamin Hindman > >
