Examples with DO_NOTHING software.amazon.awssdk.services.greengrassv2.model.DeploymentFailureHandlingPolicy.DO_NOTHING used on opensource projects

Example 1 with DO_NOTHING

use of software.amazon.awssdk.services.greengrassv2.model.DeploymentFailureHandlingPolicy.DO_NOTHING in project aws-greengrass-nucleus by aws-greengrass.

the class BaseE2ETestCase method draftAndCreateDeployment.

@SuppressWarnings("PMD.LinguisticNaming")
protected CreateDeploymentResponse draftAndCreateDeployment(CreateDeploymentRequest createDeploymentRequest) {
    // update package name with random suffix to avoid conflict in cloud
    Map<String, ComponentDeploymentSpecification> updatedPkgMetadata = new HashMap<>();
    createDeploymentRequest.components().forEach((key, val) -> updatedPkgMetadata.put(getTestComponentNameInCloud(key), val));
    createDeploymentRequest = createDeploymentRequest.toBuilder().components(updatedPkgMetadata).build();
    // set default value
    if (createDeploymentRequest.targetArn() == null) {
        createDeploymentRequest = createDeploymentRequest.toBuilder().targetArn(thingGroupArn).build();
    }
    if (createDeploymentRequest.deploymentPolicies() == null) {
        createDeploymentRequest = createDeploymentRequest.toBuilder().deploymentPolicies(DeploymentPolicies.builder().configurationValidationPolicy(DeploymentConfigurationValidationPolicy.builder().timeoutInSeconds(120).build()).componentUpdatePolicy(DeploymentComponentUpdatePolicy.builder().action(NOTIFY_COMPONENTS).timeoutInSeconds(120).build()).failureHandlingPolicy(DO_NOTHING).build()).build();
    }
    logger.atInfo().kv("CreateDeploymentRequest", createDeploymentRequest).log();
    CreateDeploymentResponse createDeploymentResult = greengrassClient.createDeployment(createDeploymentRequest);
    logger.atInfo().kv("CreateDeploymentResult", createDeploymentResult).log();
    // Keep track of deployments to clean up
    createdDeployments.add(CancelDeploymentRequest.builder().deploymentId(createDeploymentResult.deploymentId()).build());
    return createDeploymentResult;
}

Example 2 with DO_NOTHING

use of software.amazon.awssdk.services.greengrassv2.model.DeploymentFailureHandlingPolicy.DO_NOTHING in project aws-greengrass-nucleus by aws-greengrass.

the class DeploymentE2ETest method GIVEN_deployment_fails_due_to_service_broken_WHEN_deploy_fix_THEN_service_run_and_job_is_successful.

@Timeout(value = 10, unit = TimeUnit.MINUTES)
@Test
void GIVEN_deployment_fails_due_to_service_broken_WHEN_deploy_fix_THEN_service_run_and_job_is_successful(ExtensionContext context) throws Exception {
    ignoreExceptionUltimateCauseWithMessage(context, "Service " + getTestComponentNameInCloud("CustomerApp") + " in broken state after deployment");
    // Create first Job Doc with a faulty service (CustomerApp-0.9.0)
    CreateDeploymentRequest createDeploymentRequest = CreateDeploymentRequest.builder().deploymentPolicies(DeploymentPolicies.builder().configurationValidationPolicy(DeploymentConfigurationValidationPolicy.builder().timeoutInSeconds(120).build()).componentUpdatePolicy(DeploymentComponentUpdatePolicy.builder().action(SKIP_NOTIFY_COMPONENTS).timeoutInSeconds(120).build()).failureHandlingPolicy(DO_NOTHING).build()).components(Utils.immutableMap("CustomerApp", ComponentDeploymentSpecification.builder().componentVersion("0.9.0").build())).build();
    CreateDeploymentResponse createDeploymentResult = draftAndCreateDeployment(createDeploymentRequest);
    // Wait for deployment job to fail after three retries of starting CustomerApp
    IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, createDeploymentResult.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(7), s -> s.equals(JobExecutionStatus.FAILED));
    // CustomerApp should be in BROKEN state
    assertEquals(State.BROKEN, getCloudDeployedComponent("CustomerApp").getState());
    // Create another job with a fix to the faulty service (CustomerApp-0.9.1).
    CreateDeploymentRequest createDeploymentRequest2 = CreateDeploymentRequest.builder().components(Utils.immutableMap("CustomerApp", ComponentDeploymentSpecification.builder().componentVersion("0.9.1").build())).build();
    CreateDeploymentResponse createDeploymentResult2 = draftAndCreateDeployment(createDeploymentRequest2);
    IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, createDeploymentResult2.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(5), s -> s.equals(JobExecutionStatus.SUCCEEDED));
    // Ensure that main is FINISHED and CustomerApp is RUNNING.
    assertThat(kernel.getMain()::getState, eventuallyEval(is(State.FINISHED)));
    assertEquals(State.RUNNING, getCloudDeployedComponent("CustomerApp").getState());
}

Example 3 with DO_NOTHING

use of software.amazon.awssdk.services.greengrassv2.model.DeploymentFailureHandlingPolicy.DO_NOTHING in project aws-greengrass-nucleus by aws-greengrass.

the class DeploymentE2ETest method GIVEN_deployment_received_WHEN_skip_update_policy_check_THEN_update_policy_check_skipped.

@Timeout(value = 10, unit = TimeUnit.MINUTES)
@Test
void GIVEN_deployment_received_WHEN_skip_update_policy_check_THEN_update_policy_check_skipped() throws Exception {
    // GIVEN
    // First Deployment to have a service running in Kernel which has a update policy that always returns
    // false, i.e. keeps waiting forever
    CreateDeploymentRequest createDeploymentRequest1 = CreateDeploymentRequest.builder().components(Utils.immutableMap("NonDisruptableService", ComponentDeploymentSpecification.builder().componentVersion("1.0.0").build())).build();
    CreateDeploymentResponse result1 = draftAndCreateDeployment(createDeploymentRequest1);
    IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, result1.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(3), s -> s.equals(JobExecutionStatus.SUCCEEDED));
    CountDownLatch updatePolicyCheckSkipped = new CountDownLatch(1);
    Consumer<GreengrassLogMessage> logListener = m -> {
        if (m.getMessage().contains("Deployment is configured to skip update policy check")) {
            updatePolicyCheckSkipped.countDown();
        }
    };
    Slf4jLogAdapter.addGlobalListener(logListener);
    // WHEN
    // Second deployment to update the service with SKIP_NOTIFY_COMPONENTS
    CreateDeploymentRequest createDeploymentRequest2 = CreateDeploymentRequest.builder().deploymentPolicies(DeploymentPolicies.builder().configurationValidationPolicy(DeploymentConfigurationValidationPolicy.builder().timeoutInSeconds(120).build()).failureHandlingPolicy(DO_NOTHING).componentUpdatePolicy(DeploymentComponentUpdatePolicy.builder().action(SKIP_NOTIFY_COMPONENTS).timeoutInSeconds(120).build()).build()).components(Utils.immutableMap("NonDisruptableService", ComponentDeploymentSpecification.builder().componentVersion("1.0.1").build())).build();
    CreateDeploymentResponse result2 = draftAndCreateDeployment(createDeploymentRequest2);
    IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, result2.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(3), s -> s.equals(JobExecutionStatus.IN_PROGRESS));
    // THEN
    assertTrue(updatePolicyCheckSkipped.await(60, TimeUnit.SECONDS));
    Slf4jLogAdapter.removeGlobalListener(logListener);
    IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, result2.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(3), s -> s.equals(JobExecutionStatus.SUCCEEDED));
    // Ensure that main is finished, which is its terminal state, so this means that all updates ought to be done
    assertThat(kernel.getMain()::getState, eventuallyEval(is(State.FINISHED)));
    assertThat(getCloudDeployedComponent("NonDisruptableService")::getState, eventuallyEval(is(State.RUNNING)));
    assertEquals("1.0.1", getCloudDeployedComponent("NonDisruptableService").getConfig().find("version").getOnce());
}

Example 4 with DO_NOTHING

use of software.amazon.awssdk.services.greengrassv2.model.DeploymentFailureHandlingPolicy.DO_NOTHING in project aws-greengrass-nucleus by aws-greengrass.

the class DeploymentE2ETest method GIVEN_deployment_in_progress_with_more_jobs_queued_in_cloud_WHEN_cancel_event_received_and_kernel_is_waiting_for_safe_time_THEN_deployment_should_be_canceled.

@Timeout(value = 10, unit = TimeUnit.MINUTES)
@Test
void GIVEN_deployment_in_progress_with_more_jobs_queued_in_cloud_WHEN_cancel_event_received_and_kernel_is_waiting_for_safe_time_THEN_deployment_should_be_canceled() throws Exception {
    // First Deployment to have a service running in Kernel which has a safety check that always returns
    // false, i.e. keeps waiting forever
    CreateDeploymentRequest createDeploymentRequest1 = CreateDeploymentRequest.builder().components(Utils.immutableMap("NonDisruptableService", ComponentDeploymentSpecification.builder().componentVersion("1.0.0").build())).build();
    CreateDeploymentResponse result1 = draftAndCreateDeployment(createDeploymentRequest1);
    IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, result1.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(3), s -> s.equals(JobExecutionStatus.SUCCEEDED));
    Consumer<GreengrassLogMessage> logListener = null;
    try (EventStreamRPCConnection connection = IPCTestUtils.getEventStreamRpcConnection(kernel, "NonDisruptableService" + testComponentSuffix)) {
        GreengrassCoreIPCClient ipcClient = new GreengrassCoreIPCClient(connection);
        ipcClient.subscribeToComponentUpdates(new SubscribeToComponentUpdatesRequest(), Optional.of(new StreamResponseHandler<ComponentUpdatePolicyEvents>() {

            @Override
            public void onStreamEvent(ComponentUpdatePolicyEvents streamEvent) {
                if (streamEvent.getPreUpdateEvent() != null) {
                    logger.atInfo().log("Got pre component update event");
                    DeferComponentUpdateRequest deferComponentUpdateRequest = new DeferComponentUpdateRequest();
                    deferComponentUpdateRequest.setRecheckAfterMs(TimeUnit.SECONDS.toMillis(60));
                    deferComponentUpdateRequest.setMessage("NonDisruptableService");
                    logger.atInfo().log("Sending defer request");
                    // Cannot wait inside a callback
                    ipcClient.deferComponentUpdate(deferComponentUpdateRequest, Optional.empty());
                }
            }

            @Override
            public boolean onStreamError(Throwable error) {
                logger.atError().setCause(error).log("Caught stream error while subscribing for component update");
                return false;
            }

            @Override
            public void onStreamClosed() {
            }
        }));
        CountDownLatch updateRegistered = new CountDownLatch(1);
        CountDownLatch deploymentCancelled = new CountDownLatch(1);
        logListener = m -> {
            if ("register-service-update-action".equals(m.getEventType())) {
                updateRegistered.countDown();
            }
            if (m.getMessage() != null && m.getMessage().contains("Deployment was cancelled")) {
                deploymentCancelled.countDown();
            }
        };
        Slf4jLogAdapter.addGlobalListener(logListener);
        // Second deployment to update the service which is currently running an important task so deployment should
        // keep waiting for a safe time to update
        CreateDeploymentRequest createDeploymentRequest2 = CreateDeploymentRequest.builder().deploymentPolicies(DeploymentPolicies.builder().configurationValidationPolicy(DeploymentConfigurationValidationPolicy.builder().timeoutInSeconds(120).build()).failureHandlingPolicy(DO_NOTHING).componentUpdatePolicy(DeploymentComponentUpdatePolicy.builder().action(NOTIFY_COMPONENTS).timeoutInSeconds(120).build()).build()).components(Utils.immutableMap("NonDisruptableService", ComponentDeploymentSpecification.builder().componentVersion("1.0.1").build())).build();
        CreateDeploymentResponse result2 = draftAndCreateDeployment(createDeploymentRequest2);
        IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, result2.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(3), s -> s.equals(JobExecutionStatus.IN_PROGRESS));
        // Create one more deployment so that it's queued in cloud
        CreateDeploymentRequest createDeploymentRequest3 = CreateDeploymentRequest.builder().deploymentPolicies(DeploymentPolicies.builder().configurationValidationPolicy(DeploymentConfigurationValidationPolicy.builder().timeoutInSeconds(120).build()).failureHandlingPolicy(DO_NOTHING).componentUpdatePolicy(DeploymentComponentUpdatePolicy.builder().action(NOTIFY_COMPONENTS).timeoutInSeconds(120).build()).build()).components(Utils.immutableMap("NonDisruptableService", ComponentDeploymentSpecification.builder().componentVersion("1.0.1").build())).build();
        CreateDeploymentResponse result3 = draftAndCreateDeployment(createDeploymentRequest3);
        // Wait for the second deployment to start waiting for safe time to update and
        // then cancel it's corresponding job from cloud
        assertTrue(updateRegistered.await(60, TimeUnit.SECONDS));
        UpdateSystemPolicyService updateSystemPolicyService = kernel.getContext().get(UpdateSystemPolicyService.class);
        assertThat("The UpdateSystemService should have one pending action.", updateSystemPolicyService.getPendingActions(), IsCollectionWithSize.hasSize(1));
        // Get the value of the pending Action
        String pendingAction = updateSystemPolicyService.getPendingActions().iterator().next();
        // GG_NEEDS_REVIEW: TODO : Call Fleet configuration service's cancel API when ready instead of calling IoT Jobs API
        IotJobsUtils.cancelJob(iotClient, result2.iotJobId());
        // Wait for indication that cancellation has gone through
        assertTrue(deploymentCancelled.await(240, TimeUnit.SECONDS));
        // the third deployment may have reached device.
        Set<String> pendingActions = updateSystemPolicyService.getPendingActions();
        if (pendingActions.size() == 1) {
            String newPendingAction = pendingActions.iterator().next();
            assertNotEquals(pendingAction, newPendingAction, "The UpdateSystemService's one pending action should be be replaced.");
        } else if (pendingActions.size() > 1) {
            fail("Deployment not cancelled, pending actions: " + updateSystemPolicyService.getPendingActions());
        }
        // Now that we've verified that the job got cancelled, let's verify that the next job was picked up
        // and put into IN_PROGRESS state
        IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, result3.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(3), s -> s.equals(JobExecutionStatus.IN_PROGRESS));
        // Ensure that main is finished, which is its terminal state, so this means that all updates ought to be done
        assertThat(kernel.getMain()::getState, eventuallyEval(is(State.FINISHED)));
        assertThat(getCloudDeployedComponent("NonDisruptableService")::getState, eventuallyEval(is(State.RUNNING)));
        assertEquals("1.0.0", getCloudDeployedComponent("NonDisruptableService").getConfig().find("version").getOnce());
    } finally {
        if (logListener != null) {
            Slf4jLogAdapter.removeGlobalListener(logListener);
        }
    }
}

Example 5 with DO_NOTHING

use of software.amazon.awssdk.services.greengrassv2.model.DeploymentFailureHandlingPolicy.DO_NOTHING in project aws-greengrass-nucleus by aws-greengrass.

the class DeploymentE2ETest method GIVEN_some_running_services_WHEN_cancel_event_received_and_kernel_is_waiting_for_disruptable_time_THEN_deployment_should_be_canceled.

@Timeout(value = 10, unit = TimeUnit.MINUTES)
@Test
void GIVEN_some_running_services_WHEN_cancel_event_received_and_kernel_is_waiting_for_disruptable_time_THEN_deployment_should_be_canceled() throws Exception {
    // First Deployment to have a service running in Kernel which has a update policy check that always returns
    // false, i.e. keeps waiting forever
    CreateDeploymentRequest createDeploymentRequest1 = CreateDeploymentRequest.builder().components(Utils.immutableMap("NonDisruptableService", ComponentDeploymentSpecification.builder().componentVersion("1.0.0").build())).build();
    CreateDeploymentResponse createDeploymentResult1 = draftAndCreateDeployment(createDeploymentRequest1);
    IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, createDeploymentResult1.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(3), s -> s.equals(JobExecutionStatus.SUCCEEDED));
    CountDownLatch postUpdateEventReceived = new CountDownLatch(1);
    Consumer<GreengrassLogMessage> logListener = null;
    try (EventStreamRPCConnection connection = IPCTestUtils.getEventStreamRpcConnection(kernel, "NonDisruptableService" + testComponentSuffix)) {
        GreengrassCoreIPCClient ipcClient = new GreengrassCoreIPCClient(connection);
        ipcClient.subscribeToComponentUpdates(new SubscribeToComponentUpdatesRequest(), Optional.of(new StreamResponseHandler<ComponentUpdatePolicyEvents>() {

            @Override
            public void onStreamEvent(ComponentUpdatePolicyEvents streamEvent) {
                if (streamEvent.getPreUpdateEvent() != null) {
                    DeferComponentUpdateRequest deferComponentUpdateRequest = new DeferComponentUpdateRequest();
                    deferComponentUpdateRequest.setRecheckAfterMs(TimeUnit.SECONDS.toMillis(60));
                    deferComponentUpdateRequest.setMessage("NonDisruptableService");
                    // Cannot wait for response inside a callback
                    ipcClient.deferComponentUpdate(deferComponentUpdateRequest, Optional.empty());
                }
                if (streamEvent.getPostUpdateEvent() != null) {
                    postUpdateEventReceived.countDown();
                }
            }

            @Override
            public boolean onStreamError(Throwable error) {
                logger.atError().setCause(error).log("Caught stream error while subscribing for component update");
                return false;
            }

            @Override
            public void onStreamClosed() {
            }
        }));
        // Second deployment to update the service which is currently running an important task so deployment should
        // wait for a disruptable time to update
        CreateDeploymentRequest createDeploymentRequest2 = CreateDeploymentRequest.builder().deploymentPolicies(DeploymentPolicies.builder().failureHandlingPolicy(DO_NOTHING).configurationValidationPolicy(DeploymentConfigurationValidationPolicy.builder().timeoutInSeconds(120).build()).componentUpdatePolicy(DeploymentComponentUpdatePolicy.builder().action(NOTIFY_COMPONENTS).timeoutInSeconds(120).build()).build()).components(Utils.immutableMap("NonDisruptableService", ComponentDeploymentSpecification.builder().componentVersion("1.0.1").build())).build();
        CreateDeploymentResponse createDeploymentResult2 = draftAndCreateDeployment(createDeploymentRequest2);
        CountDownLatch updateRegistered = new CountDownLatch(1);
        CountDownLatch deploymentCancelled = new CountDownLatch(1);
        logListener = m -> {
            if ("register-service-update-action".equals(m.getEventType())) {
                updateRegistered.countDown();
            }
            if (m.getMessage() != null && m.getMessage().contains("Deployment was cancelled")) {
                deploymentCancelled.countDown();
            }
        };
        Slf4jLogAdapter.addGlobalListener(logListener);
        IotJobsUtils.waitForJobExecutionStatusToSatisfy(iotClient, createDeploymentResult2.iotJobId(), thingInfo.getThingName(), Duration.ofMinutes(3), s -> s.equals(JobExecutionStatus.IN_PROGRESS));
        // Wait for the second deployment to start waiting for safe time to update and
        // then cancel it's corresponding job from cloud
        assertTrue(updateRegistered.await(60, TimeUnit.SECONDS));
        assertThat("The UpdateSystemService should have one pending action.", kernel.getContext().get(UpdateSystemPolicyService.class).getPendingActions(), IsCollectionWithSize.hasSize(1));
        // GG_NEEDS_REVIEW: TODO : Call Fleet configuration service's cancel API when ready instead of calling IoT Jobs API
        IotJobsUtils.cancelJob(iotClient, createDeploymentResult2.iotJobId());
        // Wait for indication that cancellation has gone through
        assertTrue(deploymentCancelled.await(60, TimeUnit.SECONDS));
        assertThat("The UpdateSystemService's one pending action should be be removed.", kernel.getContext().get(UpdateSystemPolicyService.class).getPendingActions(), IsCollectionWithSize.hasSize(0));
        // Component should be told to resume its work since the change it has been waiting for is cancelled
        assertTrue(postUpdateEventReceived.await(60, TimeUnit.SECONDS));
        // Ensure that main is finished, which is its terminal state, so this means that all updates ought to be done
        assertThat(kernel.getMain()::getState, eventuallyEval(is(State.FINISHED)));
        assertThat(getCloudDeployedComponent("NonDisruptableService")::getState, eventuallyEval(is(State.RUNNING)));
        assertEquals("1.0.0", getCloudDeployedComponent("NonDisruptableService").getConfig().find("version").getOnce());
    } finally {
        if (logListener != null) {
            Slf4jLogAdapter.removeGlobalListener(logListener);
        }
    }
}