Examples with ClusterDecomissionService com.sequenceiq.cloudbreak.cluster.api.ClusterDecomissionService used on opensource projects

Example 1 with ClusterDecomissionService

use of com.sequenceiq.cloudbreak.cluster.api.ClusterDecomissionService in project cloudbreak by hortonworks.

the class DecommissionHandler method accept.

@Override
public void accept(Event<DecommissionRequest> event) {
    DecommissionRequest request = event.getData();
    DecommissionResult result;
    Set<String> hostNames = Collections.emptySet();
    boolean forced = request.getDetails() != null && request.getDetails().isForced();
    try {
        Stack stack = stackService.getByIdWithListsInTransaction(request.getResourceId());
        hostNames = getHostNamesForPrivateIds(request, stack);
        ClusterDecomissionService clusterDecomissionService = getClusterDecomissionService(stack);
        Map<String, InstanceMetaData> hostsToRemove = new HashMap<>();
        Set<String> hostGroupNames = request.getHostGroupNames();
        for (String hostGroup : hostGroupNames) {
            hostsToRemove.putAll(getRemovableHosts(clusterDecomissionService, stack, hostGroup, hostNames));
        }
        updateInstancesToDeleteRequested(hostsToRemove.values());
        if (!hostsToRemove.isEmpty()) {
            executePreTerminationRecipes(stack, hostsToRemove.keySet());
        }
        Optional<String> runtimeVersion = runtimeVersionService.getRuntimeVersion(stack.getCluster().getId());
        if (entitlementService.bulkHostsRemovalFromCMSupported(Crn.fromString(stack.getResourceCrn()).getAccountId()) && CMRepositoryVersionUtil.isCmBulkHostsRemovalAllowed(runtimeVersion)) {
            result = bulkHostsRemoval(request, hostNames, forced, stack, clusterDecomissionService, hostsToRemove);
        } else {
            result = singleHostsRemoval(request, hostNames, forced, stack, clusterDecomissionService, hostsToRemove);
        }
    } catch (Exception e) {
        LOGGER.info("Exception occurred during decommission.", e);
        if (isTolerableError(e) && forced && !request.getDetails().isRepair()) {
            eventService.fireCloudbreakEvent(request.getResourceId(), UPDATE_IN_PROGRESS.name(), CLUSTER_DECOMMISSION_FAILED_FORCE_DELETE_CONTINUE, Collections.singletonList(e.getMessage()));
            result = new DecommissionResult(request, hostNames);
        } else {
            result = new DecommissionResult(e.getMessage(), e, request, hostNames, UNKNOWN_ERROR_PHASE);
        }
    }
    eventBus.notify(result.selector(), new Event<>(event.getHeaders(), result));
}

Example 2 with ClusterDecomissionService

use of com.sequenceiq.cloudbreak.cluster.api.ClusterDecomissionService in project cloudbreak by hortonworks.

the class StopStartDownscaleDecommissionViaCMHandler method doAccept.

@Override
protected Selectable doAccept(HandlerEvent<StopStartDownscaleDecommissionViaCMRequest> event) {
    StopStartDownscaleDecommissionViaCMRequest request = event.getData();
    LOGGER.info("StopStartDownscaleDecommissionViaCMHandler for: {}, {}", event.getData().getResourceId(), event.getData());
    try {
        Stack stack = stackService.getByIdWithLists(request.getResourceId());
        Cluster cluster = stack.getCluster();
        ClusterDecomissionService clusterDecomissionService = clusterApiConnectors.getConnector(stack).clusterDecomissionService();
        Set<String> hostNames = getHostNamesForPrivateIds(request.getInstanceIdsToDecommission(), stack);
        LOGGER.debug("Attempting to decommission hosts. count={}, hostnames={}", hostNames.size(), hostNames);
        HostGroup hostGroup = hostGroupService.getByClusterIdAndName(cluster.getId(), request.getHostGroupName()).orElseThrow(NotFoundException.notFound("hostgroup", request.getHostGroupName()));
        Map<String, InstanceMetaData> hostsToRemove = clusterDecomissionService.collectHostsToRemove(hostGroup, hostNames);
        List<String> missingHostsInCm = Collections.emptyList();
        if (hostNames.size() != hostsToRemove.size()) {
            missingHostsInCm = hostNames.stream().filter(h -> !hostsToRemove.containsKey(h)).collect(Collectors.toList());
            LOGGER.info("Found fewer instances in CM to decommission, as compared to initial ask. foundCount={}, initialCount={}, missingHostsInCm={}", hostsToRemove.size(), hostNames.size(), missingHostsInCm);
        }
        // TODO CB-14929: Potentially put the nodes into maintenance mode before decommissioning?
        // TODO CB-15132: Eventually, try parsing the results of the CM decommission, and see if a partial decommission went through in the
        // timebound specified.
        Set<String> decommissionedHostNames = Collections.emptySet();
        if (hostsToRemove.size() > 0) {
            decommissionedHostNames = clusterDecomissionService.decommissionClusterNodesStopStart(hostsToRemove, POLL_FOR_10_MINUTES);
            updateInstanceStatuses(hostsToRemove, decommissionedHostNames, InstanceStatus.DECOMMISSIONED, "decommission requested for instances");
        }
        // This doesn't handle failures. It handles scenarios where CM list APIs don't have the necessary hosts available.
        List<String> allMissingHostnames = null;
        if (missingHostsInCm.size() > 0) {
            allMissingHostnames = new LinkedList<>(missingHostsInCm);
        }
        if (hostsToRemove.size() != decommissionedHostNames.size()) {
            Set<String> finalDecommissionedHostnames = decommissionedHostNames;
            List<String> additionalMissingDecommissionHostnames = hostsToRemove.keySet().stream().filter(h -> !finalDecommissionedHostnames.contains(h)).collect(Collectors.toList());
            LOGGER.info("Decommissioned fewer instances than requested. decommissionedCount={}, expectedCount={}, initialCount={}, notDecommissioned=[{}]", decommissionedHostNames.size(), hostsToRemove.size(), hostNames.size(), additionalMissingDecommissionHostnames);
            if (allMissingHostnames == null) {
                allMissingHostnames = new LinkedList<>();
            }
            allMissingHostnames.addAll(additionalMissingDecommissionHostnames);
        }
        LOGGER.info("hostsDecommissioned: count={}, hostNames={}", decommissionedHostNames.size(), decommissionedHostNames);
        if (decommissionedHostNames.size() > 0) {
            LOGGER.debug("Attempting to put decommissioned hosts into maintenance mode. count={}", decommissionedHostNames.size());
            flowMessageService.fireEventAndLog(stack.getId(), UPDATE_IN_PROGRESS.name(), CLUSTER_SCALING_STOPSTART_DOWNSCALE_ENTERINGCMMAINTMODE, String.valueOf(decommissionedHostNames.size()));
            clusterDecomissionService.enterMaintenanceMode(decommissionedHostNames);
            flowMessageService.fireEventAndLog(stack.getId(), UPDATE_IN_PROGRESS.name(), CLUSTER_SCALING_STOPSTART_DOWNSCALE_ENTEREDCMMAINTMODE, String.valueOf(decommissionedHostNames.size()));
            LOGGER.debug("Successfully put decommissioned hosts into maintenance mode. count={}", decommissionedHostNames.size());
        } else {
            LOGGER.debug("No nodes decommissioned, hence no nodes being put into maintenance mode");
        }
        return new StopStartDownscaleDecommissionViaCMResult(request, decommissionedHostNames, allMissingHostnames);
    } catch (Exception e) {
        // TODO CB-15132: This can be improved based on where and when the Exception occurred to potentially rollback certain aspects.
        // ClusterClientInitException is one which is explicitly thrown.
        String message = "Failed while attempting to decommission nodes via CM";
        LOGGER.error(message, e);
        return new StopStartDownscaleDecommissionViaCMResult(message, e, request);
    }
}

Example 3 with ClusterDecomissionService

use of com.sequenceiq.cloudbreak.cluster.api.ClusterDecomissionService in project cloudbreak by hortonworks.

the class StopStartDownscaleDecommissionViaCMHandlerTest method testErrorFromCmHostCollection.

@Test
void testErrorFromCmHostCollection() {
    int instancesToDecommissionCount = 5;
    int expcetedInstanceToCollectCount = 5;
    int expectedInstancesDecommissionedCount = 5;
    List<InstanceMetaData> instancesToDecommission = getInstancesToDecommission(instancesToDecommissionCount);
    HostGroup hostGroup = createHostGroup(instancesToDecommission);
    Map<String, InstanceMetaData> collected = instancesToDecommission.stream().limit(expcetedInstanceToCollectCount).collect(Collectors.toMap(i -> i.getDiscoveryFQDN(), i -> i));
    List<InstanceMetaData> decommissionedMetadataList = collected.values().stream().limit(expectedInstancesDecommissionedCount).collect(Collectors.toList());
    Set<String> fqdnsDecommissioned = decommissionedMetadataList.stream().map(InstanceMetaData::getDiscoveryFQDN).collect(Collectors.toUnmodifiableSet());
    Set<Long> instanceIdsToDecommission = instancesToDecommission.stream().map(InstanceMetaData::getPrivateId).collect(Collectors.toUnmodifiableSet());
    Set<String> hostnamesToDecommission = instancesToDecommission.stream().map(InstanceMetaData::getDiscoveryFQDN).collect(Collectors.toUnmodifiableSet());
    setupAdditionalMocks(hostGroup, instancesToDecommission, collected, fqdnsDecommissioned);
    when(clusterDecomissionService.collectHostsToRemove(eq(hostGroup), eq(hostnamesToDecommission))).thenThrow(new RuntimeException("collectHostsToDecommissionError"));
    StopStartDownscaleDecommissionViaCMRequest request = new StopStartDownscaleDecommissionViaCMRequest(1L, INSTANCE_GROUP_NAME, instanceIdsToDecommission);
    HandlerEvent handlerEvent = new HandlerEvent(Event.wrap(request));
    Selectable selectable = underTest.doAccept(handlerEvent);
    verify(clusterDecomissionService).collectHostsToRemove(eq(hostGroup), eq(hostnamesToDecommission));
    assertThat(selectable).isInstanceOf(StopStartDownscaleDecommissionViaCMResult.class);
    StopStartDownscaleDecommissionViaCMResult result = (StopStartDownscaleDecommissionViaCMResult) selectable;
    assertThat(result.getNotDecommissionedHostFqdns()).hasSize(0);
    assertThat(result.getDecommissionedHostFqdns()).hasSize(0);
    assertThat(result.getErrorDetails().getMessage()).isEqualTo("collectHostsToDecommissionError");
    assertThat(result.getStatus()).isEqualTo(EventStatus.FAILED);
    assertThat(result.selector()).isEqualTo("STOPSTARTDOWNSCALEDECOMMISSIONVIACMRESULT_ERROR");
    verifyNoMoreInteractions(instanceMetaDataService);
    verifyNoMoreInteractions(flowMessageService);
    verifyNoMoreInteractions(clusterDecomissionService);
}

Example 4 with ClusterDecomissionService

use of com.sequenceiq.cloudbreak.cluster.api.ClusterDecomissionService in project cloudbreak by hortonworks.

the class StopStartDownscaleDecommissionViaCMHandlerTest method testNoNodesAvailableInCm.

@Test
void testNoNodesAvailableInCm() {
    int instancesToDecommissionCount = 5;
    int expcetedInstanceToCollectCount = 0;
    int expectedInstancesDecommissionedCount = 0;
    List<InstanceMetaData> instancesToDecommission = getInstancesToDecommission(instancesToDecommissionCount);
    HostGroup hostGroup = createHostGroup(instancesToDecommission);
    Map<String, InstanceMetaData> collected = instancesToDecommission.stream().limit(expcetedInstanceToCollectCount).collect(Collectors.toMap(i -> i.getDiscoveryFQDN(), i -> i));
    List<InstanceMetaData> decommissionedMetadataList = collected.values().stream().limit(expectedInstancesDecommissionedCount).collect(Collectors.toList());
    Set<String> fqdnsDecommissioned = decommissionedMetadataList.stream().map(InstanceMetaData::getDiscoveryFQDN).collect(Collectors.toUnmodifiableSet());
    setupAdditionalMocks(hostGroup, instancesToDecommission, collected, fqdnsDecommissioned);
    Set<Long> instanceIdsToDecommission = instancesToDecommission.stream().map(InstanceMetaData::getPrivateId).collect(Collectors.toUnmodifiableSet());
    Set<String> hostnamesToDecommission = instancesToDecommission.stream().map(InstanceMetaData::getDiscoveryFQDN).collect(Collectors.toUnmodifiableSet());
    StopStartDownscaleDecommissionViaCMRequest request = new StopStartDownscaleDecommissionViaCMRequest(1L, INSTANCE_GROUP_NAME, instanceIdsToDecommission);
    HandlerEvent handlerEvent = new HandlerEvent(Event.wrap(request));
    Selectable selectable = underTest.doAccept(handlerEvent);
    assertThat(selectable).isInstanceOf(StopStartDownscaleDecommissionViaCMResult.class);
    StopStartDownscaleDecommissionViaCMResult result = (StopStartDownscaleDecommissionViaCMResult) selectable;
    assertThat(result.getDecommissionedHostFqdns()).hasSize(expectedInstancesDecommissionedCount);
    assertThat(result.getNotDecommissionedHostFqdns()).hasSize(instancesToDecommissionCount - expectedInstancesDecommissionedCount);
    verifyNoMoreInteractions(instanceMetaDataService);
    verify(clusterDecomissionService).collectHostsToRemove(eq(hostGroup), eq(hostnamesToDecommission));
    verifyNoMoreInteractions(flowMessageService);
    verifyNoMoreInteractions(clusterDecomissionService);
}

Example 5 with ClusterDecomissionService

use of com.sequenceiq.cloudbreak.cluster.api.ClusterDecomissionService in project cloudbreak by hortonworks.

the class StopStartDownscaleDecommissionViaCMHandlerTest method testNoNodesFromCMDecommission.

@Test
void testNoNodesFromCMDecommission() {
    int instancesToDecommissionCount = 5;
    int expcetedInstanceToCollectCount = 4;
    int expectedInstancesDecommissionedCount = 0;
    List<InstanceMetaData> instancesToDecommission = getInstancesToDecommission(instancesToDecommissionCount);
    HostGroup hostGroup = createHostGroup(instancesToDecommission);
    Map<String, InstanceMetaData> collected = instancesToDecommission.stream().limit(expcetedInstanceToCollectCount).collect(Collectors.toMap(i -> i.getDiscoveryFQDN(), i -> i));
    List<InstanceMetaData> decommissionedMetadataList = collected.values().stream().limit(expectedInstancesDecommissionedCount).collect(Collectors.toList());
    Set<String> fqdnsDecommissioned = decommissionedMetadataList.stream().map(InstanceMetaData::getDiscoveryFQDN).collect(Collectors.toUnmodifiableSet());
    setupAdditionalMocks(hostGroup, instancesToDecommission, collected, fqdnsDecommissioned);
    Set<Long> instanceIdsToDecommission = instancesToDecommission.stream().map(InstanceMetaData::getPrivateId).collect(Collectors.toUnmodifiableSet());
    Set<String> hostnamesToDecommission = instancesToDecommission.stream().map(InstanceMetaData::getDiscoveryFQDN).collect(Collectors.toUnmodifiableSet());
    StopStartDownscaleDecommissionViaCMRequest request = new StopStartDownscaleDecommissionViaCMRequest(1L, INSTANCE_GROUP_NAME, instanceIdsToDecommission);
    HandlerEvent handlerEvent = new HandlerEvent(Event.wrap(request));
    Selectable selectable = underTest.doAccept(handlerEvent);
    assertThat(selectable).isInstanceOf(StopStartDownscaleDecommissionViaCMResult.class);
    StopStartDownscaleDecommissionViaCMResult result = (StopStartDownscaleDecommissionViaCMResult) selectable;
    assertThat(result.getDecommissionedHostFqdns()).hasSize(expectedInstancesDecommissionedCount);
    assertThat(result.getNotDecommissionedHostFqdns()).hasSize(instancesToDecommissionCount - expectedInstancesDecommissionedCount);
    verifyNoMoreInteractions(instanceMetaDataService);
    verify(clusterDecomissionService).collectHostsToRemove(eq(hostGroup), eq(hostnamesToDecommission));
    verify(clusterDecomissionService).decommissionClusterNodesStopStart(eq(collected), anyLong());
    verifyNoMoreInteractions(flowMessageService);
    verifyNoMoreInteractions(clusterDecomissionService);
}