> On Dec. 7, 2016, 7:25 p.m., Joseph Wu wrote: > > I think you should mention this in your review description: > > https://reviews.apache.org/r/41665/ > > > > And it looks like we've added quite a few master failover tests since then > > :) > > By code inspection (grep for `master.*reset`, i.e. any test that restarts > > the master), we'll also need to change: > > > > fault_tolerance_tests.cpp: > > - TEST_F(FaultToleranceTest, MasterFailover) > > - TEST_F(FaultToleranceTest, ReregisterCompletedFrameworks) > > - TEST_F(FaultToleranceTest, ReregisterFrameworkExitedExecutor) > > - TEST_F(FaultToleranceTest, FrameworkReregisterEmptyExecutor) > > > > master_authorization_tests: > > - TYPED_TEST(MasterAuthorizerTest, FilterOrphanedTasks) > > > > master_tests.cpp > > - TEST_F(MasterTest, RecoveredSlaveReregisters) > > - TEST_F(MasterTest, RecoveredSlaveReregisterThenUnreachableRace) > > - TEST_F(MasterTest, OrphanTasks) > > - TEST_F(MasterTest, OrphanTasksMultipleAgents) > > - TEST_F(MasterTest, MasterFailoverLongLivedExecutor) > > - TEST_F(MasterTest, DuplicatedSlaveIdWhenSlaveReregister) > > - TEST_F(MasterTest, DISABLED_RecoverResourcesOrphanedTask) > > > > partition_tests.cpp: > > - TEST_F(PartitionTest, PartitionedSlaveReregistrationMasterFailover) > > - TEST_F(PartitionTest, DisconnectedFramework) > > > > reconciliation_tests.cpp: > > - TEST_F(ReconciliationTest, ReconcileStatusUpdateTaskState) > > > > scheduler_event_call_tests.cpp: > > - TEST_F(SchedulerDriverEventTest, SubscribedMasterFailover) > > > > scheduler_tests.cpp: > > - TEST_P(SchedulerTest, MasterFailover) > > > > slave_recovery_tests.cpp: > > - TYPED_TEST(SlaveRecoveryTest, MasterFailover)
Per our conversation, I set the flag to `"replicated_log"` only for the tests that failed when we turned it off by default. As we agreed, I will look at this a little closer today and file a bug if it looks like they shouldn't pass, or if I can't tell if they should pass. I will update this thread when I create the bug. - Alex ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/54453/#review158371 ----------------------------------------------------------- On Dec. 6, 2016, 10:34 p.m., Alex Clemmer wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/54453/ > ----------------------------------------------------------- > > (Updated Dec. 6, 2016, 10:34 p.m.) > > > Review request for mesos, Andrew Schwartzmeyer, Daniel Pravat, Joris Van > Remoortere, and Joseph Wu. > > > Bugs: MESOS-6695 and MESOS-6697 > https://issues.apache.org/jira/browse/MESOS-6695 > https://issues.apache.org/jira/browse/MESOS-6697 > > > Repository: mesos > > > Description > ------- > > Currently, nearly all instances of the Master in our tests have the > `replicated_log` flag turned on. On Windows, this will cause all tests > that involve instantiating a Master to fail immediately. > > This commit will solve this problem by setting the default master flags > to set `in_memory` instead, and then selectively turning on > `replicated_log` when it's necessary. > > NOTE: This commit is largely a resurrection of #41665. And, although it > is meant specifically to address MESOS-6697, this issue actually affects > almost all open issues in MESOS-6695. > > > Diffs > ----- > > src/tests/dynamic_weights_tests.cpp > 6f1e249e51e41aee7fdb22a2ccbfa9be71774e6d > src/tests/master_tests.cpp dfedbbdf78e8054813872e9eeebccc7504097751 > src/tests/mesos.cpp 8fd8bcb033f47e2538aa36cd373c892a882afdfd > src/tests/partition_tests.cpp 5a0d4bd2de6a5aa0e9fdf0d34cd10d16fd4e34a1 > src/tests/reconciliation_tests.cpp 71073a05a702dffd4c8e10c3c125a144ac8ca41d > > Diff: https://reviews.apache.org/r/54453/diff/ > > > Testing > ------- > > > Thanks, > > Alex Clemmer > >
