Added tests for the agent gone operation.
This change adds tests for marking an active agent as gone and also
another test for ensuring that running task on an agent marked as gone
are correctly transitioned to 'TASK_GONE_BY_OPERATOR'.
Review: https://reviews.apache.org/r/62480/
Project: http://git-wip-us.apache.org/repos/asf/mesos/repo
Commit: http://git-wip-us.apache.org/repos/asf/mesos/commit/b491239f
Tree: http://git-wip-us.apache.org/repos/asf/mesos/tree/b491239f
Diff: http://git-wip-us.apache.org/repos/asf/mesos/diff/b491239f
Branch: refs/heads/master
Commit: b491239f1b177c1e5ca70ddeb302f4c388debe93
Parents: 3da4f8a
Author: Anand Mazumdar <an...@apache.org>
Authored: Fri Sep 29 11:57:59 2017 -0700
Committer: Anand Mazumdar <an...@apache.org>
Committed: Fri Sep 29 13:31:08 2017 -0700
----------------------------------------------------------------------
src/tests/api_tests.cpp | 195 ++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 194 insertions(+), 1 deletion(-)
----------------------------------------------------------------------
http://git-wip-us.apache.org/repos/asf/mesos/blob/b491239f/src/tests/api_tests.cpp
----------------------------------------------------------------------
diff --git a/src/tests/api_tests.cpp b/src/tests/api_tests.cpp
index d260a1c..9e2945c 100644
--- a/src/tests/api_tests.cpp
+++ b/src/tests/api_tests.cpp
@@ -355,7 +355,7 @@ TEST_P(MasterAPITest, GetExecutors)
ASSERT_SOME(slave);
AWAIT_READY(slaveRegisteredMessage);
- SlaveID slaveId = slaveRegisteredMessage->slave_id();
+ const SlaveID& slaveId = slaveRegisteredMessage->slave_id();
MockScheduler sched;
MesosSchedulerDriver driver(
@@ -3516,6 +3516,199 @@ TEST_P(MasterAPITest, Teardown)
}
+// This test verifies that a registered agent can be marked as gone and
+// shutdown by the master subsequently. Upon restarting the agent, it
+// should not be able to reregister with the master.
+//
+// TODO(andschwa): Enable this when MESOS-7604 is fixed.
+TEST_P_TEMP_DISABLED_ON_WINDOWS(MasterAPITest, MarkRegisteredAgentGone)
+{
+ master::Flags masterFlags = CreateMasterFlags();
+
+ Try<Owned<cluster::Master>> master = this->StartMaster(masterFlags);
+ ASSERT_SOME(master);
+
+ Owned<MasterDetector> detector = master.get()->createDetector();
+
+ Future<SlaveRegisteredMessage> slaveRegisteredMessage =
+ FUTURE_PROTOBUF(SlaveRegisteredMessage(), master.get()->pid, _);
+
+ slave::Flags slaveFlags = CreateSlaveFlags();
+
+ Try<Owned<cluster::Slave>> slave = StartSlave(detector.get(), slaveFlags);
+ ASSERT_SOME(slave);
+
+ // Ensure that the agent is registered successfully with the master
+ // before marking it as gone.
+ AWAIT_READY(slaveRegisteredMessage);
+
+ // Mark the agent as gone. This should result in the agent being shutdown.
+
+ Future<ShutdownMessage> shutdownMessage =
+ FUTURE_PROTOBUF(ShutdownMessage(), master.get()->pid, _);
+
+ ContentType contentType = GetParam();
+
+ v1::master::Call v1Call;
+ v1Call.set_type(v1::master::Call::MARK_AGENT_GONE);
+
+ v1::master::Call::MarkAgentGone* markAgentGone =
+ v1Call.mutable_mark_agent_gone();
+
+ markAgentGone->mutable_agent_id()->CopyFrom(
+ evolve(slaveRegisteredMessage->slave_id()));
+
+ Future<http::Response> response = http::post(
+ master.get()->pid,
+ "api/v1",
+ createBasicAuthHeaders(DEFAULT_CREDENTIAL),
+ serialize(contentType, v1Call),
+ stringify(contentType));
+
+ AWAIT_EXPECT_RESPONSE_STATUS_EQ(http::OK().status, response);
+ AWAIT_READY(shutdownMessage);
+
+ // The agent should not be able to re-register with
+ // the master upon restart.
+
+ slave.get()->terminate();
+ slave->reset();
+
+ Future<ShutdownMessage> shutdownMessage2 =
+ FUTURE_PROTOBUF(ShutdownMessage(), master.get()->pid, _);
+
+ slave = StartSlave(detector.get(), slaveFlags);
+
+ AWAIT_READY(shutdownMessage2);
+}
+
+
+// This test verifies that the master correctly sends 'TASK_GONE_BY_OPERATOR'
+// status updates when an agent running the tasks is marked as gone.
+//
+// TODO(andschwa): Enable this when MESOS-7604 is fixed.
+TEST_P_TEMP_DISABLED_ON_WINDOWS(MasterAPITest, TaskUpdatesUponAgentGone)
+{
+ master::Flags masterFlags = CreateMasterFlags();
+
+ Try<Owned<cluster::Master>> master = this->StartMaster(masterFlags);
+ ASSERT_SOME(master);
+
+ MockExecutor exec(DEFAULT_EXECUTOR_ID);
+ TestContainerizer containerizer(&exec);
+
+ Owned<MasterDetector> detector = master.get()->createDetector();
+
+ slave::Flags slaveFlags = CreateSlaveFlags();
+
+ Try<Owned<cluster::Slave>> slave =
+ StartSlave(detector.get(), &containerizer, slaveFlags);
+
+ ASSERT_SOME(slave);
+
+ FrameworkInfo framework = DEFAULT_FRAMEWORK_INFO;
+ framework.add_capabilities()->set_type(
+ FrameworkInfo::Capability::PARTITION_AWARE);
+
+ MockScheduler sched;
+ MesosSchedulerDriver driver(
+ &sched, framework, master.get()->pid, DEFAULT_CREDENTIAL);
+
+ Future<Nothing> registered;
+ EXPECT_CALL(sched, registered(&driver, _, _))
+ .WillOnce(FutureSatisfy(®istered));
+
+ Future<vector<Offer>> offers;
+ EXPECT_CALL(sched, resourceOffers(&driver, _))
+ .WillOnce(FutureArg<1>(&offers));
+
+ driver.start();
+
+ AWAIT_READY(registered);
+
+ AWAIT_READY(offers);
+ ASSERT_FALSE(offers->empty());
+
+ TaskInfo task = createTask(offers.get()[0], "", DEFAULT_EXECUTOR_ID);
+
+ EXPECT_CALL(exec, registered(_, _, _, _));
+
+ EXPECT_CALL(exec, launchTask(_, _))
+ .WillOnce(SendStatusUpdateFromTask(TASK_RUNNING));
+
+ Future<TaskStatus> update;
+ EXPECT_CALL(sched, statusUpdate(&driver, _))
+ .WillOnce(FutureArg<1>(&update));
+
+ driver.launchTasks(offers.get()[0].id(), {task});
+
+ AWAIT_READY(update);
+ ASSERT_EQ(TASK_RUNNING, update->state());
+
+ Future<TaskStatus> update2;
+ EXPECT_CALL(sched, statusUpdate(&driver, _))
+ .WillOnce(FutureArg<1>(&update2));
+
+ EXPECT_CALL(sched, slaveLost(&driver, _))
+ .Times(AtMost(1));
+
+ EXPECT_CALL(exec, shutdown(_))
+ .Times(AtMost(1));
+
+ // Mark the agent as gone. This should result in the master sending
+ // a 'TASK_GONE_BY_OPERATOR' update for the running task.
+
+ ContentType contentType = GetParam();
+
+ SlaveID slaveId = offers.get()[0].slave_id();
+
+ v1::master::Call v1Call;
+ v1Call.set_type(v1::master::Call::MARK_AGENT_GONE);
+
+ v1::master::Call::MarkAgentGone* markAgentGone =
+ v1Call.mutable_mark_agent_gone();
+
+ markAgentGone->mutable_agent_id()->CopyFrom(evolve(slaveId));
+
+ Future<http::Response> response = http::post(
+ master.get()->pid,
+ "api/v1",
+ createBasicAuthHeaders(DEFAULT_CREDENTIAL),
+ serialize(contentType, v1Call),
+ stringify(contentType));
+
+ AWAIT_EXPECT_RESPONSE_STATUS_EQ(http::OK().status, response);
+
+ AWAIT_READY(update2);
+
+ ASSERT_EQ(TASK_GONE_BY_OPERATOR, update2->state());
+ ASSERT_EQ(TaskStatus::REASON_SLAVE_REMOVED_BY_OPERATOR, update2->reason());
+
+ // Performing reconciliation for an unknown task on the gone agent should
+ // result in a 'TASK_GONE_BY_OPERATOR' update.
+
+ Future<TaskStatus> update3;
+ EXPECT_CALL(sched, statusUpdate(&driver, _))
+ .WillOnce(FutureArg<1>(&update3));
+
+ vector<TaskStatus> statuses;
+
+ TaskStatus status;
+ status.mutable_slave_id()->CopyFrom(slaveId);
+ status.mutable_task_id()->set_value("dummy-task");
+
+ statuses.push_back(status);
+
+ driver.reconcileTasks(statuses);
+
+ AWAIT_READY(update3);
+ ASSERT_EQ(TASK_GONE_BY_OPERATOR, update3->state());
+
+ driver.stop();
+ driver.join();
+}
+
+
class AgentAPITest
: public MesosTest,
public WithParamInterface<ContentType>