Search in sources :

Example 31 with ActorTaskManagerGateway

use of org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway in project flink by apache.

the class ExecutionVertexCancelTest method testCancelConcurrentlyToDeploying_CallsOvertaking.

@Test
public void testCancelConcurrentlyToDeploying_CallsOvertaking() {
    try {
        final JobVertexID jid = new JobVertexID();
        final TestingUtils.QueuedActionExecutionContext executionContext = TestingUtils.queuedActionExecutionContext();
        final TestingUtils.ActionQueue actions = executionContext.actionQueue();
        final ExecutionJobVertex ejv = getExecutionVertex(jid, executionContext);
        final ExecutionVertex vertex = new ExecutionVertex(ejv, 0, new IntermediateResult[0], AkkaUtils.getDefaultTimeout());
        setVertexState(vertex, ExecutionState.SCHEDULED);
        assertEquals(ExecutionState.SCHEDULED, vertex.getExecutionState());
        // task manager cancel sequence mock actor
        // first return NOT SUCCESS (task not found, cancel call overtook deploy call), then success (cancel call after deploy call)
        ActorGateway actorGateway = new CancelSequenceActorGateway(executionContext, 2);
        Instance instance = getInstance(new ActorTaskManagerGateway(actorGateway));
        SimpleSlot slot = instance.allocateSimpleSlot(new JobID());
        vertex.deployToSlot(slot);
        assertEquals(ExecutionState.DEPLOYING, vertex.getExecutionState());
        vertex.cancel();
        assertEquals(ExecutionState.CANCELING, vertex.getExecutionState());
        // first action happens (deploy)
        Runnable deployAction = actions.popNextAction();
        Runnable cancelAction = actions.popNextAction();
        // cancel call first
        cancelAction.run();
        // process onComplete callback
        actions.triggerNextAction();
        // did not find the task, not properly cancelled, stay in canceling
        assertEquals(ExecutionState.CANCELING, vertex.getExecutionState());
        // deploy action next
        deployAction.run();
        // the deploy call found itself in canceling after it returned and needs to send a cancel call
        // the call did not yet execute, so it is still in canceling
        assertEquals(ExecutionState.CANCELING, vertex.getExecutionState());
        vertex.getCurrentExecutionAttempt().cancelingComplete();
        assertEquals(ExecutionState.CANCELED, vertex.getExecutionState());
        assertTrue(slot.isReleased());
        assertNull(vertex.getFailureCause());
        assertTrue(vertex.getStateTimestamp(ExecutionState.CREATED) > 0);
        assertTrue(vertex.getStateTimestamp(ExecutionState.CANCELING) > 0);
        assertTrue(vertex.getStateTimestamp(ExecutionState.CANCELED) > 0);
    } catch (Exception e) {
        e.printStackTrace();
        fail(e.getMessage());
    }
}
Also used : Instance(org.apache.flink.runtime.instance.Instance) JobVertexID(org.apache.flink.runtime.jobgraph.JobVertexID) SimpleSlot(org.apache.flink.runtime.instance.SimpleSlot) IOException(java.io.IOException) ActorTaskManagerGateway(org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway) TestingUtils(org.apache.flink.runtime.testingUtils.TestingUtils) BaseTestingActorGateway(org.apache.flink.runtime.instance.BaseTestingActorGateway) DummyActorGateway(org.apache.flink.runtime.instance.DummyActorGateway) ActorGateway(org.apache.flink.runtime.instance.ActorGateway) JobID(org.apache.flink.api.common.JobID) Test(org.junit.Test)

Example 32 with ActorTaskManagerGateway

use of org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway in project flink by apache.

the class ExecutionVertexCancelTest method testActionsWhileCancelling.

@Test
public void testActionsWhileCancelling() {
    try {
        final JobVertexID jid = new JobVertexID();
        final ExecutionJobVertex ejv = getExecutionVertex(jid);
        // scheduling while canceling is an illegal state transition
        try {
            ExecutionVertex vertex = new ExecutionVertex(ejv, 0, new IntermediateResult[0], AkkaUtils.getDefaultTimeout());
            setVertexState(vertex, ExecutionState.CANCELING);
            Scheduler scheduler = mock(Scheduler.class);
            vertex.scheduleForExecution(scheduler, false);
        } catch (Exception e) {
            fail("should not throw an exception");
        }
        // deploying while in canceling state is illegal (should immediately go to canceled)
        try {
            ExecutionVertex vertex = new ExecutionVertex(ejv, 0, new IntermediateResult[0], AkkaUtils.getDefaultTimeout());
            setVertexState(vertex, ExecutionState.CANCELING);
            Instance instance = getInstance(new ActorTaskManagerGateway(DummyActorGateway.INSTANCE));
            SimpleSlot slot = instance.allocateSimpleSlot(new JobID());
            vertex.deployToSlot(slot);
            fail("Method should throw an exception");
        } catch (IllegalStateException e) {
        // that is what we expect
        }
        // fail while canceling
        {
            ExecutionVertex vertex = new ExecutionVertex(ejv, 0, new IntermediateResult[0], AkkaUtils.getDefaultTimeout());
            Instance instance = getInstance(new ActorTaskManagerGateway(DummyActorGateway.INSTANCE));
            SimpleSlot slot = instance.allocateSimpleSlot(new JobID());
            setVertexResource(vertex, slot);
            setVertexState(vertex, ExecutionState.CANCELING);
            Exception failureCause = new Exception("test exception");
            vertex.fail(failureCause);
            assertEquals(ExecutionState.CANCELED, vertex.getExecutionState());
            assertTrue(slot.isReleased());
        }
    } catch (Exception e) {
        e.printStackTrace();
        fail(e.getMessage());
    }
}
Also used : Instance(org.apache.flink.runtime.instance.Instance) Scheduler(org.apache.flink.runtime.jobmanager.scheduler.Scheduler) JobVertexID(org.apache.flink.runtime.jobgraph.JobVertexID) SimpleSlot(org.apache.flink.runtime.instance.SimpleSlot) IOException(java.io.IOException) JobID(org.apache.flink.api.common.JobID) ActorTaskManagerGateway(org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway) Test(org.junit.Test)

Example 33 with ActorTaskManagerGateway

use of org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway in project flink by apache.

the class ExecutionVertexDeploymentTest method testFailExternallyDuringDeploy.

@Test
public void testFailExternallyDuringDeploy() {
    try {
        final JobVertexID jid = new JobVertexID();
        final TestingUtils.QueuedActionExecutionContext ec = TestingUtils.queuedActionExecutionContext();
        final TestingUtils.ActionQueue queue = ec.actionQueue();
        final ExecutionJobVertex ejv = getExecutionVertex(jid, ec);
        final ExecutionVertex vertex = new ExecutionVertex(ejv, 0, new IntermediateResult[0], AkkaUtils.getDefaultTimeout());
        final Instance instance = getInstance(new ActorTaskManagerGateway(new SimpleActorGateway(TestingUtils.directExecutionContext())));
        final SimpleSlot slot = instance.allocateSimpleSlot(ejv.getJobId());
        assertEquals(ExecutionState.CREATED, vertex.getExecutionState());
        vertex.deployToSlot(slot);
        assertEquals(ExecutionState.DEPLOYING, vertex.getExecutionState());
        Exception testError = new Exception("test error");
        vertex.fail(testError);
        assertEquals(ExecutionState.FAILED, vertex.getExecutionState());
        assertEquals(testError, vertex.getFailureCause());
        queue.triggerNextAction();
        queue.triggerNextAction();
        assertTrue(vertex.getStateTimestamp(ExecutionState.CREATED) > 0);
        assertTrue(vertex.getStateTimestamp(ExecutionState.DEPLOYING) > 0);
        assertTrue(vertex.getStateTimestamp(ExecutionState.FAILED) > 0);
    } catch (Exception e) {
        e.printStackTrace();
        fail(e.getMessage());
    }
}
Also used : TestingUtils(org.apache.flink.runtime.testingUtils.TestingUtils) Instance(org.apache.flink.runtime.instance.Instance) ExecutionGraphTestUtils.getInstance(org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.getInstance) JobVertexID(org.apache.flink.runtime.jobgraph.JobVertexID) SimpleActorGateway(org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.SimpleActorGateway) SimpleSlot(org.apache.flink.runtime.instance.SimpleSlot) ExecutionGraphTestUtils.getExecutionVertex(org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.getExecutionVertex) ActorTaskManagerGateway(org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway) Test(org.junit.Test)

Example 34 with ActorTaskManagerGateway

use of org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway in project flink by apache.

the class ExecutionVertexDeploymentTest method testDeployFailedSynchronous.

@Test
public void testDeployFailedSynchronous() {
    try {
        final JobVertexID jid = new JobVertexID();
        final ExecutionJobVertex ejv = getExecutionVertex(jid, new DirectScheduledExecutorService());
        final ExecutionVertex vertex = new ExecutionVertex(ejv, 0, new IntermediateResult[0], AkkaUtils.getDefaultTimeout());
        final Instance instance = getInstance(new ActorTaskManagerGateway(new SimpleFailingActorGateway(TestingUtils.directExecutionContext())));
        final SimpleSlot slot = instance.allocateSimpleSlot(ejv.getJobId());
        assertEquals(ExecutionState.CREATED, vertex.getExecutionState());
        vertex.deployToSlot(slot);
        assertEquals(ExecutionState.FAILED, vertex.getExecutionState());
        assertNotNull(vertex.getFailureCause());
        assertTrue(vertex.getFailureCause().getMessage().contains(ERROR_MESSAGE));
        assertTrue(vertex.getStateTimestamp(ExecutionState.CREATED) > 0);
        assertTrue(vertex.getStateTimestamp(ExecutionState.DEPLOYING) > 0);
        assertTrue(vertex.getStateTimestamp(ExecutionState.FAILED) > 0);
    } catch (Exception e) {
        e.printStackTrace();
        fail(e.getMessage());
    }
}
Also used : Instance(org.apache.flink.runtime.instance.Instance) ExecutionGraphTestUtils.getInstance(org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.getInstance) DirectScheduledExecutorService(org.apache.flink.runtime.testutils.DirectScheduledExecutorService) JobVertexID(org.apache.flink.runtime.jobgraph.JobVertexID) SimpleFailingActorGateway(org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.SimpleFailingActorGateway) SimpleSlot(org.apache.flink.runtime.instance.SimpleSlot) ExecutionGraphTestUtils.getExecutionVertex(org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.getExecutionVertex) ActorTaskManagerGateway(org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway) Test(org.junit.Test)

Example 35 with ActorTaskManagerGateway

use of org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway in project flink by apache.

the class ExecutionVertexDeploymentTest method testDeployCall.

@Test
public void testDeployCall() {
    try {
        final JobVertexID jid = new JobVertexID();
        final ExecutionJobVertex ejv = getExecutionVertex(jid);
        // mock taskmanager to simply accept the call
        Instance instance = getInstance(new ActorTaskManagerGateway(new SimpleActorGateway(TestingUtils.directExecutionContext())));
        final SimpleSlot slot = instance.allocateSimpleSlot(ejv.getJobId());
        final ExecutionVertex vertex = new ExecutionVertex(ejv, 0, new IntermediateResult[0], AkkaUtils.getDefaultTimeout());
        assertEquals(ExecutionState.CREATED, vertex.getExecutionState());
        vertex.deployToSlot(slot);
        assertEquals(ExecutionState.DEPLOYING, vertex.getExecutionState());
        // no repeated scheduling
        try {
            vertex.deployToSlot(slot);
            fail("Scheduled from wrong state");
        } catch (IllegalStateException e) {
        // as expected
        }
        assertNull(vertex.getFailureCause());
        assertTrue(vertex.getStateTimestamp(ExecutionState.CREATED) > 0);
        assertTrue(vertex.getStateTimestamp(ExecutionState.DEPLOYING) > 0);
    } catch (Exception e) {
        e.printStackTrace();
        fail(e.getMessage());
    }
}
Also used : Instance(org.apache.flink.runtime.instance.Instance) ExecutionGraphTestUtils.getInstance(org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.getInstance) JobVertexID(org.apache.flink.runtime.jobgraph.JobVertexID) SimpleActorGateway(org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.SimpleActorGateway) SimpleSlot(org.apache.flink.runtime.instance.SimpleSlot) ExecutionGraphTestUtils.getExecutionVertex(org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.getExecutionVertex) ActorTaskManagerGateway(org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway) Test(org.junit.Test)

Aggregations

ActorTaskManagerGateway (org.apache.flink.runtime.jobmanager.slots.ActorTaskManagerGateway)40 Test (org.junit.Test)36 Instance (org.apache.flink.runtime.instance.Instance)29 SimpleSlot (org.apache.flink.runtime.instance.SimpleSlot)22 JobVertexID (org.apache.flink.runtime.jobgraph.JobVertexID)22 JobID (org.apache.flink.api.common.JobID)20 Scheduler (org.apache.flink.runtime.jobmanager.scheduler.Scheduler)16 IOException (java.io.IOException)14 ExecutionGraphTestUtils.getInstance (org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.getInstance)12 SimpleActorGateway (org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.SimpleActorGateway)11 ExecutionGraphTestUtils.getExecutionVertex (org.apache.flink.runtime.executiongraph.ExecutionGraphTestUtils.getExecutionVertex)11 JobVertex (org.apache.flink.runtime.jobgraph.JobVertex)11 InetAddress (java.net.InetAddress)9 ResourceID (org.apache.flink.runtime.clusterframework.types.ResourceID)9 TaskManagerLocation (org.apache.flink.runtime.taskmanager.TaskManagerLocation)9 ExecutionConfig (org.apache.flink.api.common.ExecutionConfig)8 ActorGateway (org.apache.flink.runtime.instance.ActorGateway)8 BaseTestingActorGateway (org.apache.flink.runtime.instance.BaseTestingActorGateway)8 DirectScheduledExecutorService (org.apache.flink.runtime.testutils.DirectScheduledExecutorService)8 DummyActorGateway (org.apache.flink.runtime.instance.DummyActorGateway)7