Repository: mesos Updated Branches: refs/heads/master 82979512b -> 0cc636b2d
Fixed flaky `ROOT_CGROUPS_RecoverLauncherOrphans` test. Containerizer recovery returns control to the caller before completion of destruction of orphaned containers. Previously, `wait` was called on a container right after calling `recover`, so `wait` was almost always successfull, because destruction of the orphaned container takes some time to complete. This patch replaces check for the container existence with the check that a related freezer cgroup has been destroyed. The freezer cgroup is destroyed during container destruction initiated by a containerizer recovery process. Review: https://reviews.apache.org/r/64680/ Project: http://git-wip-us.apache.org/repos/asf/mesos/repo Commit: http://git-wip-us.apache.org/repos/asf/mesos/commit/0cc636b2 Tree: http://git-wip-us.apache.org/repos/asf/mesos/tree/0cc636b2 Diff: http://git-wip-us.apache.org/repos/asf/mesos/diff/0cc636b2 Branch: refs/heads/master Commit: 0cc636b2d5ad5c934b8b7f350bc8c99b9282b5ab Parents: 8297951 Author: Andrei Budnik <abud...@mesosphere.com> Authored: Tue Dec 19 19:45:33 2017 +0100 Committer: Alexander Rukletsov <al...@apache.org> Committed: Tue Dec 19 19:45:33 2017 +0100 ---------------------------------------------------------------------- .../containerizer/nested_mesos_containerizer_tests.cpp | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/mesos/blob/0cc636b2/src/tests/containerizer/nested_mesos_containerizer_tests.cpp ---------------------------------------------------------------------- diff --git a/src/tests/containerizer/nested_mesos_containerizer_tests.cpp b/src/tests/containerizer/nested_mesos_containerizer_tests.cpp index 22bd995..92832e7 100644 --- a/src/tests/containerizer/nested_mesos_containerizer_tests.cpp +++ b/src/tests/containerizer/nested_mesos_containerizer_tests.cpp @@ -1890,15 +1890,20 @@ TEST_F(NestedMesosContainerizerTest, ROOT_CGROUPS_RecoverLauncherOrphans) buildPath(containerId, "mesos", JOIN)); ASSERT_SOME(cgroups::create(freezerHierarchy.get(), cgroup, true)); + ASSERT_SOME_TRUE(cgroups::exists(freezerHierarchy.get(), cgroup)); SlaveState state; state.id = SlaveID(); AWAIT_READY(containerizer->recover(state)); - Future<Option<ContainerTermination>> wait = containerizer->wait(containerId); - AWAIT_READY(wait); - ASSERT_SOME(wait.get()); + // We expect that containerizer recovery will detect orphan container and + // will destroy it, so we check here that the freezer cgroup is destroyed. + // + // NOTE: `wait()` can return `Some` or `None` due to a race condition between + // `recover()` and `______destroy()` for an orphan container. + AWAIT_READY(containerizer->wait(containerId)); + ASSERT_SOME_FALSE(cgroups::exists(freezerHierarchy.get(), cgroup)); Future<hashset<ContainerID>> containers = containerizer->containers(); AWAIT_READY(containers);
