blob: cb22e81f3669b97ef1d69b40e2eb505f94159e4b [file] [log] [blame]
// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing,
// software distributed under the License is distributed on an
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, either express or implied. See the License for the
// specific language governing permissions and limitations
// under the License.
package com.cloud.deploy;
import static com.cloud.utils.NumbersUtil.toHumanReadableSize;
import java.util.ArrayList;
import java.util.Collections;
import java.util.Comparator;
import java.util.HashMap;
import java.util.HashSet;
import java.util.List;
import java.util.Map;
import java.util.Objects;
import java.util.Optional;
import java.util.Set;
import java.util.Timer;
import java.util.TreeSet;
import java.util.stream.Collectors;
import javax.inject.Inject;
import javax.naming.ConfigurationException;
import org.apache.cloudstack.affinity.AffinityGroupDomainMapVO;
import org.apache.cloudstack.affinity.AffinityGroupProcessor;
import org.apache.cloudstack.affinity.AffinityGroupService;
import org.apache.cloudstack.affinity.AffinityGroupVMMapVO;
import org.apache.cloudstack.affinity.AffinityGroupVO;
import org.apache.cloudstack.affinity.dao.AffinityGroupDao;
import org.apache.cloudstack.affinity.dao.AffinityGroupDomainMapDao;
import org.apache.cloudstack.affinity.dao.AffinityGroupVMMapDao;
import org.apache.cloudstack.engine.cloud.entity.api.db.VMReservationVO;
import org.apache.cloudstack.engine.cloud.entity.api.db.dao.VMReservationDao;
import org.apache.cloudstack.engine.subsystem.api.storage.DataStore;
import org.apache.cloudstack.engine.subsystem.api.storage.DataStoreManager;
import org.apache.cloudstack.engine.subsystem.api.storage.StoragePoolAllocator;
import org.apache.cloudstack.framework.config.ConfigKey;
import org.apache.cloudstack.framework.config.Configurable;
import org.apache.cloudstack.framework.config.dao.ConfigurationDao;
import org.apache.cloudstack.framework.messagebus.MessageBus;
import org.apache.cloudstack.framework.messagebus.MessageSubscriber;
import org.apache.cloudstack.managed.context.ManagedContextTimerTask;
import org.apache.cloudstack.storage.datastore.db.PrimaryDataStoreDao;
import org.apache.cloudstack.storage.datastore.db.StoragePoolVO;
import org.apache.cloudstack.utils.identity.ManagementServerNode;
import org.apache.commons.collections.CollectionUtils;
import org.apache.commons.lang3.StringUtils;
import org.apache.log4j.Logger;
import com.cloud.agent.AgentManager;
import com.cloud.agent.Listener;
import com.cloud.agent.api.AgentControlAnswer;
import com.cloud.agent.api.AgentControlCommand;
import com.cloud.agent.api.Answer;
import com.cloud.agent.api.Command;
import com.cloud.agent.api.StartupCommand;
import com.cloud.agent.api.StartupRoutingCommand;
import com.cloud.agent.manager.allocator.HostAllocator;
import com.cloud.capacity.CapacityManager;
import com.cloud.capacity.dao.CapacityDao;
import com.cloud.configuration.Config;
import com.cloud.configuration.ConfigurationManagerImpl;
import com.cloud.dc.ClusterDetailsDao;
import com.cloud.dc.ClusterDetailsVO;
import com.cloud.dc.ClusterVO;
import com.cloud.dc.DataCenter;
import com.cloud.dc.DataCenterVO;
import com.cloud.dc.DedicatedResourceVO;
import com.cloud.dc.Pod;
import com.cloud.dc.dao.ClusterDao;
import com.cloud.dc.dao.DataCenterDao;
import com.cloud.dc.dao.DedicatedResourceDao;
import com.cloud.dc.dao.HostPodDao;
import com.cloud.deploy.DeploymentPlanner.ExcludeList;
import com.cloud.deploy.DeploymentPlanner.PlannerResourceUsage;
import com.cloud.deploy.dao.PlannerHostReservationDao;
import com.cloud.exception.AffinityConflictException;
import com.cloud.exception.ConnectionException;
import com.cloud.exception.InsufficientServerCapacityException;
import com.cloud.exception.StorageUnavailableException;
import com.cloud.gpu.GPU;
import com.cloud.host.DetailVO;
import com.cloud.host.Host;
import com.cloud.host.HostVO;
import com.cloud.host.Status;
import com.cloud.host.dao.HostDao;
import com.cloud.host.dao.HostDetailsDao;
import com.cloud.hypervisor.Hypervisor.HypervisorType;
import com.cloud.offering.ServiceOffering;
import com.cloud.org.Cluster;
import com.cloud.org.Grouping;
import com.cloud.resource.ResourceManager;
import com.cloud.service.ServiceOfferingDetailsVO;
import com.cloud.service.dao.ServiceOfferingDetailsDao;
import com.cloud.storage.DiskOfferingVO;
import com.cloud.storage.GuestOSVO;
import com.cloud.storage.ScopeType;
import com.cloud.storage.StorageManager;
import com.cloud.storage.StoragePool;
import com.cloud.storage.StoragePoolHostVO;
import com.cloud.storage.VMTemplateVO;
import com.cloud.storage.Volume;
import com.cloud.storage.VolumeVO;
import com.cloud.storage.dao.DiskOfferingDao;
import com.cloud.storage.dao.GuestOSCategoryDao;
import com.cloud.storage.dao.GuestOSDao;
import com.cloud.storage.dao.StoragePoolHostDao;
import com.cloud.storage.dao.VMTemplateDao;
import com.cloud.storage.dao.VolumeDao;
import com.cloud.user.AccountManager;
import com.cloud.user.AccountVO;
import com.cloud.user.dao.AccountDao;
import com.cloud.utils.DateUtil;
import com.cloud.utils.NumbersUtil;
import com.cloud.utils.Pair;
import com.cloud.utils.component.Manager;
import com.cloud.utils.component.ManagerBase;
import com.cloud.utils.db.DB;
import com.cloud.utils.db.Filter;
import com.cloud.utils.db.SearchCriteria;
import com.cloud.utils.db.Transaction;
import com.cloud.utils.db.TransactionCallback;
import com.cloud.utils.db.TransactionStatus;
import com.cloud.utils.exception.CloudRuntimeException;
import com.cloud.utils.fsm.StateListener;
import com.cloud.utils.fsm.StateMachine2;
import com.cloud.vm.DiskProfile;
import com.cloud.vm.VMInstanceVO;
import com.cloud.vm.VirtualMachine;
import com.cloud.vm.VirtualMachine.Event;
import com.cloud.vm.VirtualMachine.State;
import com.cloud.vm.VirtualMachineProfile;
import com.cloud.vm.dao.UserVmDao;
import com.cloud.vm.dao.VMInstanceDao;
public class DeploymentPlanningManagerImpl extends ManagerBase implements DeploymentPlanningManager, Manager, Listener,
StateListener<State, VirtualMachine.Event, VirtualMachine>, Configurable {
private static final Logger s_logger = Logger.getLogger(DeploymentPlanningManagerImpl.class);
@Inject
AgentManager _agentMgr;
@Inject
private AccountDao accountDao;
@Inject
protected UserVmDao _vmDao;
@Inject
protected VMInstanceDao _vmInstanceDao;
@Inject
protected AffinityGroupDao _affinityGroupDao;
@Inject
protected AffinityGroupVMMapDao _affinityGroupVMMapDao;
@Inject
protected AffinityGroupDomainMapDao _affinityGroupDomainMapDao;
@Inject
AffinityGroupService _affinityGroupService;
@Inject
DataCenterDao _dcDao;
@Inject
PlannerHostReservationDao _plannerHostReserveDao;
private int _vmCapacityReleaseInterval;
@Inject
MessageBus _messageBus;
private Timer _timer = null;
private long _hostReservationReleasePeriod = 60L * 60L * 1000L; // one hour by default
@Inject
protected VMReservationDao _reservationDao;
@Inject
HostDetailsDao _hostDetailsDao;
@Inject
private VMTemplateDao templateDao;
private static final long ADMIN_ACCOUNT_ROLE_ID = 1l;
private static final long INITIAL_RESERVATION_RELEASE_CHECKER_DELAY = 30L * 1000L; // thirty seconds expressed in milliseconds
protected long _nodeId = -1;
protected List<StoragePoolAllocator> _storagePoolAllocators;
public List<StoragePoolAllocator> getStoragePoolAllocators() {
return _storagePoolAllocators;
}
public void setStoragePoolAllocators(List<StoragePoolAllocator> storagePoolAllocators) {
_storagePoolAllocators = storagePoolAllocators;
}
protected List<HostAllocator> _hostAllocators;
public List<HostAllocator> getHostAllocators() {
return _hostAllocators;
}
public void setHostAllocators(List<HostAllocator> hostAllocators) {
_hostAllocators = hostAllocators;
}
@Inject
protected HostDao _hostDao;
@Inject
protected HostPodDao _podDao;
@Inject
protected ClusterDao _clusterDao;
@Inject
protected DedicatedResourceDao _dedicatedDao;
@Inject
protected GuestOSDao _guestOSDao = null;
@Inject
protected GuestOSCategoryDao _guestOSCategoryDao = null;
@Inject
protected DiskOfferingDao _diskOfferingDao;
@Inject
protected StoragePoolHostDao _poolHostDao;
@Inject
protected VolumeDao _volsDao;
@Inject
protected CapacityManager _capacityMgr;
@Inject
protected ConfigurationDao _configDao;
@Inject
protected PrimaryDataStoreDao _storagePoolDao;
@Inject
protected CapacityDao _capacityDao;
@Inject
protected AccountManager _accountMgr;
@Inject
protected StorageManager _storageMgr;
@Inject
DataStoreManager dataStoreMgr;
@Inject
protected ClusterDetailsDao _clusterDetailsDao;
@Inject
protected ResourceManager _resourceMgr;
@Inject
protected ServiceOfferingDetailsDao _serviceOfferingDetailsDao;
protected List<DeploymentPlanner> _planners;
public List<DeploymentPlanner> getPlanners() {
return _planners;
}
public void setPlanners(List<DeploymentPlanner> planners) {
_planners = planners;
}
protected List<AffinityGroupProcessor> _affinityProcessors;
public List<AffinityGroupProcessor> getAffinityGroupProcessors() {
return _affinityProcessors;
}
public void setAffinityGroupProcessors(List<AffinityGroupProcessor> affinityProcessors) {
_affinityProcessors = affinityProcessors;
}
protected void avoidOtherClustersForDeploymentIfMigrationDisabled(VirtualMachine vm, Host lastHost, ExcludeList avoids) {
if (lastHost == null || lastHost.getClusterId() == null ||
ConfigurationManagerImpl.MIGRATE_VM_ACROSS_CLUSTERS.valueIn(vm.getDataCenterId())) {
return;
}
List<VolumeVO> volumes = _volsDao.findUsableVolumesForInstance(vm.getId());
if (CollectionUtils.isEmpty(volumes)) {
return;
}
boolean storageMigrationNeededDuringClusterMigration = false;
for (Volume volume : volumes) {
StoragePoolVO pool = _storagePoolDao.findById(volume.getPoolId());
if (List.of(ScopeType.HOST, ScopeType.CLUSTER).contains(pool.getScope())) {
storageMigrationNeededDuringClusterMigration = true;
break;
}
}
if (!storageMigrationNeededDuringClusterMigration) {
return;
}
final Long lastHostClusterId = lastHost.getClusterId();
s_logger.warn(String.format("VM last host ID: %d belongs to zone ID: %s for which config - %s is false and storage migration would be needed for inter-cluster migration, therefore, adding all other clusters except ID: %d from this zone to avoid list",
lastHost.getId(), vm.getDataCenterId(), ConfigurationManagerImpl.MIGRATE_VM_ACROSS_CLUSTERS.key(), lastHostClusterId));
List<Long> clusterIds = _clusterDao.listAllClusters(lastHost.getDataCenterId());
Set<Long> existingAvoidedClusters = avoids.getClustersToAvoid();
clusterIds = clusterIds.stream().filter(x -> !Objects.equals(x, lastHostClusterId) && (existingAvoidedClusters == null || !existingAvoidedClusters.contains(x))).collect(Collectors.toList());
avoids.addClusterList(clusterIds);
}
@Override
public DeployDestination planDeployment(VirtualMachineProfile vmProfile, DeploymentPlan plan, ExcludeList avoids, DeploymentPlanner planner)
throws InsufficientServerCapacityException, AffinityConflictException {
ServiceOffering offering = vmProfile.getServiceOffering();
int cpu_requested = offering.getCpu() * offering.getSpeed();
long ram_requested = offering.getRamSize() * 1024L * 1024L;
VirtualMachine vm = vmProfile.getVirtualMachine();
DataCenter dc = _dcDao.findById(vm.getDataCenterId());
boolean volumesRequireEncryption = anyVolumeRequiresEncryption(_volsDao.findByInstance(vm.getId()));
if (vm.getType() == VirtualMachine.Type.User || vm.getType() == VirtualMachine.Type.DomainRouter) {
checkForNonDedicatedResources(vmProfile, dc, avoids);
}
if (s_logger.isDebugEnabled()) {
s_logger.debug("DeploymentPlanner allocation algorithm: " + planner);
s_logger.debug("Trying to allocate a host and storage pools from dc:" + plan.getDataCenterId() + ", pod:" + plan.getPodId() + ",cluster:" +
plan.getClusterId() + ", requested cpu: " + cpu_requested + ", requested ram: " + toHumanReadableSize(ram_requested));
s_logger.debug("Is ROOT volume READY (pool already allocated)?: " + (plan.getPoolId() != null ? "Yes" : "No"));
}
avoidDisabledResources(vmProfile, dc, avoids);
String haVmTag = (String)vmProfile.getParameter(VirtualMachineProfile.Param.HaTag);
String uefiFlag = (String)vmProfile.getParameter(VirtualMachineProfile.Param.UefiFlag);
if (plan.getHostId() != null && haVmTag == null) {
Long hostIdSpecified = plan.getHostId();
if (s_logger.isDebugEnabled()) {
s_logger.debug("DeploymentPlan has host_id specified, choosing this host: " + hostIdSpecified);
}
HostVO host = _hostDao.findById(hostIdSpecified);
if (host != null && StringUtils.isNotBlank(uefiFlag) && "yes".equalsIgnoreCase(uefiFlag)) {
DetailVO uefiHostDetail = _hostDetailsDao.findDetail(host.getId(), Host.HOST_UEFI_ENABLE);
if (uefiHostDetail == null || "false".equalsIgnoreCase(uefiHostDetail.getValue())) {
s_logger.debug("Cannot deploy to specified host as host does n't support uefi vm deployment, returning.");
return null;
}
}
if (host == null) {
s_logger.debug("The specified host cannot be found");
} else if (avoids.shouldAvoid(host)) {
s_logger.debug("The specified host is in avoid set");
} else {
if (s_logger.isDebugEnabled()) {
s_logger.debug(
"Looking for suitable pools for this host under zone: " + host.getDataCenterId() + ", pod: " + host.getPodId() + ", cluster: " + host.getClusterId());
}
Pod pod = _podDao.findById(host.getPodId());
Cluster cluster = _clusterDao.findById(host.getClusterId());
boolean displayStorage = getDisplayStorageFromVmProfile(vmProfile);
if (vm.getHypervisorType() == HypervisorType.BareMetal) {
DeployDestination dest = new DeployDestination(dc, pod, cluster, host, new HashMap<Volume, StoragePool>(), displayStorage);
s_logger.debug("Returning Deployment Destination: " + dest);
return dest;
}
// search for storage under the zone, pod, cluster of the host.
DataCenterDeployment lastPlan =
new DataCenterDeployment(host.getDataCenterId(), host.getPodId(), host.getClusterId(), hostIdSpecified, plan.getPoolId(), null,
plan.getReservationContext());
Pair<Map<Volume, List<StoragePool>>, List<Volume>> result = findSuitablePoolsForVolumes(vmProfile, lastPlan, avoids, HostAllocator.RETURN_UPTO_ALL);
Map<Volume, List<StoragePool>> suitableVolumeStoragePools = result.first();
List<Volume> readyAndReusedVolumes = result.second();
_hostDao.loadDetails(host);
if (volumesRequireEncryption && !Boolean.parseBoolean(host.getDetail(Host.HOST_VOLUME_ENCRYPTION))) {
s_logger.warn(String.format("VM's volumes require encryption support, and provided host %s can't handle it", host));
return null;
} else {
s_logger.debug(String.format("Volume encryption requirements are met by provided host %s", host));
}
// choose the potential pool for this VM for this host
if (!suitableVolumeStoragePools.isEmpty()) {
List<Host> suitableHosts = new ArrayList<Host>();
suitableHosts.add(host);
Pair<Host, Map<Volume, StoragePool>> potentialResources = findPotentialDeploymentResources(
suitableHosts, suitableVolumeStoragePools, avoids,
getPlannerUsage(planner, vmProfile, plan, avoids), readyAndReusedVolumes, plan.getPreferredHosts(), vm);
if (potentialResources != null) {
pod = _podDao.findById(host.getPodId());
cluster = _clusterDao.findById(host.getClusterId());
Map<Volume, StoragePool> storageVolMap = potentialResources.second();
// remove the reused vol<->pool from destination, since
// we don't have to prepare this volume.
for (Volume vol : readyAndReusedVolumes) {
storageVolMap.remove(vol);
}
DeployDestination dest = new DeployDestination(dc, pod, cluster, host, storageVolMap, displayStorage);
s_logger.debug("Returning Deployment Destination: " + dest);
return dest;
}
}
}
s_logger.debug("Cannot deploy to specified host, returning.");
return null;
}
// call affinitygroup chain
long vmGroupCount = _affinityGroupVMMapDao.countAffinityGroupsForVm(vm.getId());
if (vmGroupCount > 0) {
for (AffinityGroupProcessor processor : _affinityProcessors) {
processor.process(vmProfile, plan, avoids);
}
}
if (s_logger.isDebugEnabled()) {
s_logger.debug("Deploy avoids pods: " + avoids.getPodsToAvoid() + ", clusters: " + avoids.getClustersToAvoid() + ", hosts: " + avoids.getHostsToAvoid());
s_logger.debug("Deploy hosts with priorities " + plan.getHostPriorities() + " , hosts have NORMAL priority by default");
}
// call planners
// DataCenter dc = _dcDao.findById(vm.getDataCenterId());
// check if datacenter is in avoid set
if (avoids.shouldAvoid(dc)) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("DataCenter id = '" + dc.getId() + "' provided is in avoid set, DeploymentPlanner cannot allocate the VM, returning.");
}
return null;
}
if (planner == null) {
String plannerName = offering.getDeploymentPlanner();
if (plannerName == null) {
if (vm.getHypervisorType() == HypervisorType.BareMetal) {
plannerName = "BareMetalPlanner";
} else {
plannerName = _configDao.getValue(Config.VmDeploymentPlanner.key());
}
}
planner = getDeploymentPlannerByName(plannerName);
}
Host lastHost = null;
String considerLastHostStr = (String)vmProfile.getParameter(VirtualMachineProfile.Param.ConsiderLastHost);
boolean considerLastHost = vm.getLastHostId() != null && haVmTag == null &&
(considerLastHostStr == null || Boolean.TRUE.toString().equalsIgnoreCase(considerLastHostStr));
if (considerLastHost) {
s_logger.debug("This VM has last host_id specified, trying to choose the same host: " + vm.getLastHostId());
HostVO host = _hostDao.findById(vm.getLastHostId());
lastHost = host;
_hostDao.loadHostTags(host);
_hostDao.loadDetails(host);
ServiceOfferingDetailsVO offeringDetails = null;
if (host == null) {
s_logger.debug("The last host of this VM cannot be found");
} else if (avoids.shouldAvoid(host)) {
s_logger.debug("The last host of this VM is in avoid set");
} else if (plan.getClusterId() != null && host.getClusterId() != null
&& !plan.getClusterId().equals(host.getClusterId())) {
s_logger.debug("The last host of this VM cannot be picked as the plan specifies different clusterId: "
+ plan.getClusterId());
} else if (_capacityMgr.checkIfHostReachMaxGuestLimit(host)) {
s_logger.debug("The last Host, hostId: " + host.getId() +
" already has max Running VMs(count includes system VMs), skipping this and trying other available hosts");
} else if ((offeringDetails = _serviceOfferingDetailsDao.findDetail(offering.getId(), GPU.Keys.vgpuType.toString())) != null) {
ServiceOfferingDetailsVO groupName = _serviceOfferingDetailsDao.findDetail(offering.getId(), GPU.Keys.pciDevice.toString());
if(!_resourceMgr.isGPUDeviceAvailable(host.getId(), groupName.getValue(), offeringDetails.getValue())){
s_logger.debug("The last host of this VM does not have required GPU devices available");
}
} else if (volumesRequireEncryption && !Boolean.parseBoolean(host.getDetail(Host.HOST_VOLUME_ENCRYPTION))) {
s_logger.warn(String.format("The last host of this VM %s does not support volume encryption, which is required by this VM.", host));
} else {
if (host.getStatus() == Status.Up) {
if (checkVmProfileAndHost(vmProfile, host)) {
long cluster_id = host.getClusterId();
ClusterDetailsVO cluster_detail_cpu = _clusterDetailsDao.findDetail(cluster_id,
"cpuOvercommitRatio");
ClusterDetailsVO cluster_detail_ram = _clusterDetailsDao.findDetail(cluster_id,
"memoryOvercommitRatio");
Float cpuOvercommitRatio = Float.parseFloat(cluster_detail_cpu.getValue());
Float memoryOvercommitRatio = Float.parseFloat(cluster_detail_ram.getValue());
boolean hostHasCpuCapability, hostHasCapacity = false;
hostHasCpuCapability = _capacityMgr.checkIfHostHasCpuCapability(host.getId(), offering.getCpu(), offering.getSpeed());
if (hostHasCpuCapability) {
// first check from reserved capacity
hostHasCapacity = _capacityMgr.checkIfHostHasCapacity(host.getId(), cpu_requested, ram_requested, true, cpuOvercommitRatio, memoryOvercommitRatio, true);
// if not reserved, check the free capacity
if (!hostHasCapacity)
hostHasCapacity = _capacityMgr.checkIfHostHasCapacity(host.getId(), cpu_requested, ram_requested, false, cpuOvercommitRatio, memoryOvercommitRatio, true);
}
boolean displayStorage = getDisplayStorageFromVmProfile(vmProfile);
if (hostHasCapacity
&& hostHasCpuCapability) {
s_logger.debug("The last host of this VM is UP and has enough capacity");
s_logger.debug("Now checking for suitable pools under zone: " + host.getDataCenterId()
+ ", pod: " + host.getPodId() + ", cluster: " + host.getClusterId());
Pod pod = _podDao.findById(host.getPodId());
Cluster cluster = _clusterDao.findById(host.getClusterId());
if (vm.getHypervisorType() == HypervisorType.BareMetal) {
DeployDestination dest = new DeployDestination(dc, pod, cluster, host, new HashMap<Volume, StoragePool>(), displayStorage);
s_logger.debug("Returning Deployment Destination: " + dest);
return dest;
}
// search for storage under the zone, pod, cluster
// of
// the last host.
DataCenterDeployment lastPlan = new DataCenterDeployment(host.getDataCenterId(),
host.getPodId(), host.getClusterId(), host.getId(), plan.getPoolId(), null);
Pair<Map<Volume, List<StoragePool>>, List<Volume>> result = findSuitablePoolsForVolumes(
vmProfile, lastPlan, avoids, HostAllocator.RETURN_UPTO_ALL);
Map<Volume, List<StoragePool>> suitableVolumeStoragePools = result.first();
List<Volume> readyAndReusedVolumes = result.second();
// choose the potential pool for this VM for this
// host
if (!suitableVolumeStoragePools.isEmpty()) {
List<Host> suitableHosts = new ArrayList<Host>();
suitableHosts.add(host);
Pair<Host, Map<Volume, StoragePool>> potentialResources = findPotentialDeploymentResources(
suitableHosts, suitableVolumeStoragePools, avoids,
getPlannerUsage(planner, vmProfile, plan, avoids), readyAndReusedVolumes, plan.getPreferredHosts(), vm);
if (potentialResources != null) {
Map<Volume, StoragePool> storageVolMap = potentialResources.second();
// remove the reused vol<->pool from
// destination, since we don't have to
// prepare
// this volume.
for (Volume vol : readyAndReusedVolumes) {
storageVolMap.remove(vol);
}
DeployDestination dest = new DeployDestination(dc, pod, cluster, host,
storageVolMap, displayStorage);
s_logger.debug("Returning Deployment Destination: " + dest);
return dest;
}
}
} else {
s_logger.debug("The last host of this VM does not have enough capacity");
}
}
} else {
s_logger.debug("The last host of this VM is not UP or is not enabled, host status is: " + host.getStatus().name() + ", host resource state is: " +
host.getResourceState());
}
}
s_logger.debug("Cannot choose the last host to deploy this VM ");
}
avoidOtherClustersForDeploymentIfMigrationDisabled(vm, lastHost, avoids);
DeployDestination dest = null;
List<Long> clusterList = null;
if (planner != null && planner.canHandle(vmProfile, plan, avoids)) {
while (true) {
if (planner instanceof DeploymentClusterPlanner) {
ExcludeList plannerAvoidInput =
new ExcludeList(avoids.getDataCentersToAvoid(), avoids.getPodsToAvoid(), avoids.getClustersToAvoid(), avoids.getHostsToAvoid(),
avoids.getPoolsToAvoid());
clusterList = ((DeploymentClusterPlanner)planner).orderClusters(vmProfile, plan, avoids);
if (clusterList != null && !clusterList.isEmpty()) {
// planner refactoring. call allocators to list hosts
ExcludeList plannerAvoidOutput =
new ExcludeList(avoids.getDataCentersToAvoid(), avoids.getPodsToAvoid(), avoids.getClustersToAvoid(), avoids.getHostsToAvoid(),
avoids.getPoolsToAvoid());
resetAvoidSet(plannerAvoidOutput, plannerAvoidInput);
dest = checkClustersforDestination(clusterList, vmProfile, plan, avoids, dc, getPlannerUsage(planner, vmProfile, plan, avoids), plannerAvoidOutput);
if (dest != null) {
return dest;
}
// reset the avoid input to the planners
resetAvoidSet(avoids, plannerAvoidOutput);
} else {
return null;
}
} else {
dest = planner.plan(vmProfile, plan, avoids);
if (dest != null) {
long hostId = dest.getHost().getId();
avoids.addHost(dest.getHost().getId());
if (volumesRequireEncryption && !Boolean.parseBoolean(_hostDetailsDao.findDetail(hostId, Host.HOST_VOLUME_ENCRYPTION).getValue())) {
s_logger.warn(String.format("VM's volumes require encryption support, and the planner-provided host %s can't handle it", dest.getHost()));
continue;
} else {
s_logger.debug(String.format("VM's volume encryption requirements are met by host %s", dest.getHost()));
}
if (checkIfHostFitsPlannerUsage(hostId, DeploymentPlanner.PlannerResourceUsage.Shared)) {
// found destination
return dest;
} else {
// find another host - seems some concurrent
// deployment picked it up for dedicated access
continue;
}
} else {
return null;
}
}
}
}
return dest;
}
protected boolean anyVolumeRequiresEncryption(List<? extends Volume> volumes) {
for (Volume volume : volumes) {
if (volume.getPassphraseId() != null) {
return true;
}
}
return false;
}
private boolean isDeployAsIs(VirtualMachine vm) {
long templateId = vm.getTemplateId();
VMTemplateVO template = templateDao.findById(templateId);
return template != null && template.isDeployAsIs();
}
/**
* Display storage in the logs by default if the template is not deploy-as-is.
*/
private boolean getDisplayStorageFromVmProfile(VirtualMachineProfile vmProfile) {
return vmProfile == null || vmProfile.getTemplate() == null || !vmProfile.getTemplate().isDeployAsIs();
}
/**
* Adds disabled resources (Data centers, Pods, Clusters, and hosts) to exclude list (avoid) in case of disabled state.
*/
public void avoidDisabledResources(VirtualMachineProfile vmProfile, DataCenter dc, ExcludeList avoids) {
if (vmProfile.getType().isUsedBySystem() && isRouterDeployableInDisabledResources()) {
return;
}
VMInstanceVO vm = _vmInstanceDao.findById(vmProfile.getId());
AccountVO owner = accountDao.findById(vm.getAccountId());
boolean isOwnerRoleIdAdmin = false;
if (owner != null && owner.getRoleId() != null && owner.getRoleId() == ADMIN_ACCOUNT_ROLE_ID) {
isOwnerRoleIdAdmin = true;
}
if (isOwnerRoleIdAdmin && isAdminVmDeployableInDisabledResources()) {
return;
}
avoidDisabledDataCenters(dc, avoids);
avoidDisabledPods(dc, avoids);
avoidDisabledClusters(dc, avoids);
avoidDisabledHosts(dc, avoids);
}
/**
* Returns the value of the ConfigKey 'allow.router.on.disabled.resources'.
* @note this method allows mocking and testing with the respective ConfigKey parameter.
*/
protected boolean isRouterDeployableInDisabledResources() {
return allowRouterOnDisabledResource.value();
}
/**
* Returns the value of the ConfigKey 'allow.admin.vm.on.disabled.resources'.
* @note this method allows mocking and testing with the respective ConfigKey parameter.
*/
protected boolean isAdminVmDeployableInDisabledResources() {
return allowAdminVmOnDisabledResource.value();
}
/**
* Adds disabled Hosts to the ExcludeList in order to avoid them at the deployment planner.
*/
protected void avoidDisabledHosts(DataCenter dc, ExcludeList avoids) {
List<HostVO> disabledHosts = _hostDao.listDisabledByDataCenterId(dc.getId());
for (HostVO host : disabledHosts) {
avoids.addHost(host.getId());
}
}
/**
* Adds disabled Clusters to the ExcludeList in order to avoid them at the deployment planner.
*/
protected void avoidDisabledClusters(DataCenter dc, ExcludeList avoids) {
List<Long> pods = _podDao.listAllPods(dc.getId());
for (Long podId : pods) {
List<Long> disabledClusters = _clusterDao.listDisabledClusters(dc.getId(), podId);
avoids.addClusterList(disabledClusters);
}
}
/**
* Adds disabled Pods to the ExcludeList in order to avoid them at the deployment planner.
*/
protected void avoidDisabledPods(DataCenter dc, ExcludeList avoids) {
List<Long> disabledPods = _podDao.listDisabledPods(dc.getId());
avoids.addPodList(disabledPods);
}
/**
* Adds disabled Data Centers (Zones) to the ExcludeList in order to avoid them at the deployment planner.
*/
protected void avoidDisabledDataCenters(DataCenter dc, ExcludeList avoids) {
if (dc.getAllocationState() == Grouping.AllocationState.Disabled) {
avoids.addDataCenter(dc.getId());
}
}
@Override
public DeploymentPlanner getDeploymentPlannerByName(String plannerName) {
if (plannerName != null) {
for (DeploymentPlanner plannerInList : _planners) {
if (plannerName.equalsIgnoreCase(plannerInList.getName())) {
return plannerInList;
}
}
}
return null;
}
protected boolean checkVmProfileAndHost(final VirtualMachineProfile vmProfile, final HostVO host) {
ServiceOffering offering = vmProfile.getServiceOffering();
if (offering.getHostTag() != null) {
_hostDao.loadHostTags(host);
if (!host.checkHostServiceOfferingTags(offering)) {
s_logger.debug("Service Offering host tag does not match the last host of this VM");
return false;
}
}
long guestOSId = vmProfile.getTemplate().getGuestOSId();
GuestOSVO guestOS = _guestOSDao.findById(guestOSId);
if (guestOS != null) {
long guestOSCategoryId = guestOS.getCategoryId();
DetailVO hostDetail = _hostDetailsDao.findDetail(host.getId(), "guest.os.category.id");
if (hostDetail != null) {
String guestOSCategoryIdString = hostDetail.getValue();
if (String.valueOf(guestOSCategoryId) != guestOSCategoryIdString) {
s_logger.debug("The last host has different guest.os.category.id than guest os category of VM, skipping");
return false;
}
}
}
return true;
}
@Override
public void checkForNonDedicatedResources(VirtualMachineProfile vmProfile, DataCenter dc, ExcludeList avoids) {
boolean isExplicit = false;
VirtualMachine vm = vmProfile.getVirtualMachine();
// check if zone is dedicated. if yes check if vm owner has access to it.
DedicatedResourceVO dedicatedZone = _dedicatedDao.findByZoneId(dc.getId());
if (dedicatedZone != null && !_accountMgr.isRootAdmin(vmProfile.getOwner().getId())) {
long accountDomainId = vmProfile.getOwner().getDomainId();
long accountId = vmProfile.getOwner().getAccountId();
// If a zone is dedicated to an account then all hosts in this zone
// will be explicitly dedicated to
// that account. So there won't be any shared hosts in the zone, the
// only way to deploy vms from that
// account will be to use explicit dedication affinity group.
if (dedicatedZone.getAccountId() != null) {
if (dedicatedZone.getAccountId().equals(accountId)) {
return;
} else {
throw new CloudRuntimeException("Failed to deploy VM, Zone " + dc.getName() + " not available for the user account " + vmProfile.getOwner());
}
}
// if zone is dedicated to a domain. Check owner's access to the
// domain level dedication group
if (!_affinityGroupService.isAffinityGroupAvailableInDomain(dedicatedZone.getAffinityGroupId(), accountDomainId)) {
throw new CloudRuntimeException("Failed to deploy VM, Zone " + dc.getName() + " not available for the user domain " + vmProfile.getOwner());
}
}
// check affinity group of type Explicit dedication exists. If No put
// dedicated pod/cluster/host in avoid list
List<AffinityGroupVMMapVO> vmGroupMappings = _affinityGroupVMMapDao.findByVmIdType(vm.getId(), "ExplicitDedication");
if (vmGroupMappings != null && !vmGroupMappings.isEmpty()) {
isExplicit = true;
}
List<Long> allPodsInDc = _podDao.listAllPods(dc.getId());
List<Long> allDedicatedPods = _dedicatedDao.listAllPods();
allPodsInDc.retainAll(allDedicatedPods);
List<Long> allClustersInDc = _clusterDao.listAllClusters(dc.getId());
List<Long> allDedicatedClusters = _dedicatedDao.listAllClusters();
allClustersInDc.retainAll(allDedicatedClusters);
List<Long> allHostsInDc = _hostDao.listAllHosts(dc.getId());
List<Long> allDedicatedHosts = _dedicatedDao.listAllHosts();
allHostsInDc.retainAll(allDedicatedHosts);
//Only when the type is instance VM and not explicitly dedicated.
if (vm.getType() == VirtualMachine.Type.User && !isExplicit) {
//add explicitly dedicated resources in avoidList
if (s_logger.isDebugEnabled()) {
s_logger.debug("Adding pods to avoid lists for non-explicit VM deployment: " + allPodsInDc);
}
avoids.addPodList(allPodsInDc);
if (s_logger.isDebugEnabled()) {
s_logger.debug("Adding clusters to avoid lists for non-explicit VM deployment: " + allClustersInDc);
}
avoids.addClusterList(allClustersInDc);
if (s_logger.isDebugEnabled()) {
s_logger.debug("Adding hosts to avoid lists for non-explicit VM deployment: " + allHostsInDc);
}
avoids.addHostList(allHostsInDc);
}
//Handle the Virtual Router Case
//No need to check the isExplicit. As both the cases are handled.
if (vm.getType() == VirtualMachine.Type.DomainRouter) {
long vmAccountId = vm.getAccountId();
long vmDomainId = vm.getDomainId();
//Lists all explicitly dedicated resources from vm account ID or domain ID.
List<Long> allPodsFromDedicatedID = new ArrayList<Long>();
List<Long> allClustersFromDedicatedID = new ArrayList<Long>();
List<Long> allHostsFromDedicatedID = new ArrayList<Long>();
//Whether the dedicated resources belong to Domain or not. If not, it may belongs to Account or no dedication.
List<AffinityGroupDomainMapVO> domainGroupMappings = _affinityGroupDomainMapDao.listByDomain(vmDomainId);
//For temporary storage and indexing.
List<DedicatedResourceVO> tempStorage;
if (domainGroupMappings == null || domainGroupMappings.isEmpty()) {
//The dedicated resource belongs to VM Account ID.
tempStorage = _dedicatedDao.searchDedicatedPods(null, vmDomainId, vmAccountId, null, new Filter(DedicatedResourceVO.class, "id", true, 0L, 1L)).first();
for(DedicatedResourceVO vo : tempStorage) {
allPodsFromDedicatedID.add(vo.getPodId());
}
tempStorage.clear();
tempStorage = _dedicatedDao.searchDedicatedClusters(null, vmDomainId, vmAccountId, null, new Filter(DedicatedResourceVO.class, "id", true, 0L, 1L)).first();
for(DedicatedResourceVO vo : tempStorage) {
allClustersFromDedicatedID.add(vo.getClusterId());
}
tempStorage.clear();
tempStorage = _dedicatedDao.searchDedicatedHosts(null, vmDomainId, vmAccountId, null, new Filter(DedicatedResourceVO.class, "id", true, 0L, 1L)).first();
for(DedicatedResourceVO vo : tempStorage) {
allHostsFromDedicatedID.add(vo.getHostId());
}
//Remove the dedicated ones from main list
allPodsInDc.removeAll(allPodsFromDedicatedID);
allClustersInDc.removeAll(allClustersFromDedicatedID);
allHostsInDc.removeAll(allHostsFromDedicatedID);
}
else {
//The dedicated resource belongs to VM Domain ID or No dedication.
tempStorage = _dedicatedDao.searchDedicatedPods(null, vmDomainId, null, null, new Filter(DedicatedResourceVO.class, "id", true, 0L, 1L)).first();
for(DedicatedResourceVO vo : tempStorage) {
allPodsFromDedicatedID.add(vo.getPodId());
}
tempStorage.clear();
tempStorage = _dedicatedDao.searchDedicatedClusters(null, vmDomainId, null, null, new Filter(DedicatedResourceVO.class, "id", true, 0L, 1L)).first();
for(DedicatedResourceVO vo : tempStorage) {
allClustersFromDedicatedID.add(vo.getClusterId());
}
tempStorage.clear();
tempStorage = _dedicatedDao.searchDedicatedHosts(null, vmDomainId, null, null, new Filter(DedicatedResourceVO.class, "id", true, 0L, 1L)).first();
for(DedicatedResourceVO vo : tempStorage) {
allHostsFromDedicatedID.add(vo.getHostId());
}
//Remove the dedicated ones from main list
allPodsInDc.removeAll(allPodsFromDedicatedID);
allClustersInDc.removeAll(allClustersFromDedicatedID);
allHostsInDc.removeAll(allHostsFromDedicatedID);
}
//Add in avoid list or no addition if no dedication
if (s_logger.isDebugEnabled()) {
s_logger.debug("Adding pods to avoid lists: " + allPodsInDc);
}
avoids.addPodList(allPodsInDc);
if (s_logger.isDebugEnabled()) {
s_logger.debug("Adding clusters to avoid lists: " + allClustersInDc);
}
avoids.addClusterList(allClustersInDc);
if (s_logger.isDebugEnabled()) {
s_logger.debug("Adding hosts to avoid lists: " + allHostsInDc);
}
avoids.addHostList(allHostsInDc);
}
}
private void resetAvoidSet(ExcludeList avoidSet, ExcludeList removeSet) {
if (avoidSet.getDataCentersToAvoid() != null && removeSet.getDataCentersToAvoid() != null) {
avoidSet.getDataCentersToAvoid().removeAll(removeSet.getDataCentersToAvoid());
}
if (avoidSet.getPodsToAvoid() != null && removeSet.getPodsToAvoid() != null) {
avoidSet.getPodsToAvoid().removeAll(removeSet.getPodsToAvoid());
}
if (avoidSet.getClustersToAvoid() != null && removeSet.getClustersToAvoid() != null) {
avoidSet.getClustersToAvoid().removeAll(removeSet.getClustersToAvoid());
}
if (avoidSet.getHostsToAvoid() != null && removeSet.getHostsToAvoid() != null) {
avoidSet.getHostsToAvoid().removeAll(removeSet.getHostsToAvoid());
}
if (avoidSet.getPoolsToAvoid() != null && removeSet.getPoolsToAvoid() != null) {
avoidSet.getPoolsToAvoid().removeAll(removeSet.getPoolsToAvoid());
}
}
private PlannerResourceUsage getPlannerUsage(DeploymentPlanner planner, VirtualMachineProfile vmProfile, DeploymentPlan plan, ExcludeList avoids)
throws InsufficientServerCapacityException {
if (planner != null && planner instanceof DeploymentClusterPlanner) {
return ((DeploymentClusterPlanner)planner).getResourceUsage(vmProfile, plan, avoids);
} else {
return DeploymentPlanner.PlannerResourceUsage.Shared;
}
}
@DB
protected boolean checkIfHostFitsPlannerUsage(final long hostId, final PlannerResourceUsage resourceUsageRequired) {
// TODO Auto-generated method stub
// check if this host has been picked up by some other planner
// exclusively
// if planner can work with shared host, check if this host has
// been marked as 'shared'
// else if planner needs dedicated host,
PlannerHostReservationVO reservationEntry = _plannerHostReserveDao.findByHostId(hostId);
if (reservationEntry != null) {
final long id = reservationEntry.getId();
PlannerResourceUsage hostResourceType = reservationEntry.getResourceUsage();
if (hostResourceType != null) {
if (hostResourceType == resourceUsageRequired) {
return true;
} else {
s_logger.debug("Cannot use this host for usage: " + resourceUsageRequired + ", since this host has been reserved for planner usage : " +
hostResourceType);
return false;
}
} else {
final PlannerResourceUsage hostResourceTypeFinal = hostResourceType;
// reserve the host for required resourceType
// let us lock the reservation entry before updating.
return Transaction.execute(new TransactionCallback<Boolean>() {
@Override
public Boolean doInTransaction(TransactionStatus status) {
final PlannerHostReservationVO lockedEntry = _plannerHostReserveDao.lockRow(id, true);
if (lockedEntry == null) {
s_logger.error("Unable to lock the host entry for reservation, host: " + hostId);
return false;
}
// check before updating
if (lockedEntry.getResourceUsage() == null) {
lockedEntry.setResourceUsage(resourceUsageRequired);
_plannerHostReserveDao.persist(lockedEntry);
return true;
} else {
// someone updated it earlier. check if we can still use it
if (lockedEntry.getResourceUsage() == resourceUsageRequired) {
return true;
} else {
s_logger.debug("Cannot use this host for usage: " + resourceUsageRequired + ", since this host has been reserved for planner usage : " +
hostResourceTypeFinal);
return false;
}
}
}
});
}
}
return false;
}
@DB
public boolean checkHostReservationRelease(final Long hostId) {
if (hostId != null) {
PlannerHostReservationVO reservationEntry = _plannerHostReserveDao.findByHostId(hostId);
if (reservationEntry != null && reservationEntry.getResourceUsage() != null) {
// check if any VMs are starting or running on this host
List<VMInstanceVO> vms = _vmInstanceDao.listUpByHostId(hostId);
if (vms.size() > 0) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Cannot release reservation, Found " + vms.size() + " VMs Running on host " + hostId);
}
return false;
}
List<VMInstanceVO> vmsByLastHostId = _vmInstanceDao.listByLastHostId(hostId);
if (vmsByLastHostId.size() > 0) {
// check if any VMs are within skip.counting.hours, if yes
// we
// cannot release the host
for (VMInstanceVO stoppedVM : vmsByLastHostId) {
long secondsSinceLastUpdate = (DateUtil.currentGMTTime().getTime() - stoppedVM.getUpdateTime().getTime()) / 1000;
if (secondsSinceLastUpdate < _vmCapacityReleaseInterval) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Cannot release reservation, Found VM: " + stoppedVM + " Stopped but reserved on host " + hostId);
}
return false;
}
}
}
// check if any VMs are stopping on or migrating to this host
List<VMInstanceVO> vmsStoppingMigratingByHostId = _vmInstanceDao.findByHostInStates(hostId, State.Stopping, State.Migrating, State.Starting);
if (vmsStoppingMigratingByHostId.size() > 0) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Cannot release reservation, Found " + vmsStoppingMigratingByHostId.size() + " VMs stopping/migrating/starting on host " + hostId);
}
return false;
}
// check if any VMs are in starting state with no hostId set yet
// -
// just ignore host release to avoid race condition
List<VMInstanceVO> vmsStartingNoHost = _vmInstanceDao.listStartingWithNoHostId();
if (vmsStartingNoHost.size() > 0) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Cannot release reservation, Found " + vms.size() + " VMs starting as of now and no hostId yet stored");
}
return false;
}
if (s_logger.isDebugEnabled()) {
s_logger.debug("Host has no VMs associated, releasing the planner reservation for host " + hostId);
}
final long id = reservationEntry.getId();
return Transaction.execute(new TransactionCallback<Boolean>() {
@Override
public Boolean doInTransaction(TransactionStatus status) {
final PlannerHostReservationVO lockedEntry = _plannerHostReserveDao.lockRow(id, true);
if (lockedEntry == null) {
s_logger.error("Unable to lock the host entry for reservation, host: " + hostId);
return false;
}
// check before updating
if (lockedEntry.getResourceUsage() != null) {
lockedEntry.setResourceUsage(null);
_plannerHostReserveDao.persist(lockedEntry);
return true;
}
return false;
}
});
}
}
return false;
}
class HostReservationReleaseChecker extends ManagedContextTimerTask {
@Override
protected void runInContext() {
try {
s_logger.debug("Checking if any host reservation can be released ... ");
checkHostReservations();
s_logger.debug("Done running HostReservationReleaseChecker ... ");
} catch (Throwable t) {
s_logger.error("Exception in HostReservationReleaseChecker", t);
}
}
}
private void checkHostReservations() {
List<PlannerHostReservationVO> reservedHosts = _plannerHostReserveDao.listAllReservedHosts();
for (PlannerHostReservationVO hostReservation : reservedHosts) {
HostVO host = _hostDao.findById(hostReservation.getHostId());
if (host != null && host.getManagementServerId() != null && host.getManagementServerId() == _nodeId) {
checkHostReservationRelease(hostReservation.getHostId());
}
}
}
@Override
public boolean processAnswers(long agentId, long seq, Answer[] answers) {
// TODO Auto-generated method stub
return false;
}
@Override
public boolean processCommands(long agentId, long seq, Command[] commands) {
// TODO Auto-generated method stub
return false;
}
@Override
public AgentControlAnswer processControlCommand(long agentId, AgentControlCommand cmd) {
// TODO Auto-generated method stub
return null;
}
@Override
public void processHostAdded(long hostId) {
}
@Override
public void processConnect(Host host, StartupCommand cmd, boolean forRebalance) throws ConnectionException {
if (!(cmd instanceof StartupRoutingCommand)) {
return;
}
PlannerHostReservationVO reservationEntry = _plannerHostReserveDao.findByHostId(host.getId());
if (reservationEntry == null) {
// record the host in this table
PlannerHostReservationVO newHost = new PlannerHostReservationVO(host.getId(), host.getDataCenterId(), host.getPodId(), host.getClusterId());
_plannerHostReserveDao.persist(newHost);
}
}
@Override
public boolean processDisconnect(long agentId, Status state) {
// TODO Auto-generated method stub
return false;
}
@Override
public void processHostAboutToBeRemoved(long hostId) {
}
@Override
public void processHostRemoved(long hostId, long clusterId) {
}
@Override
public boolean isRecurring() {
// TODO Auto-generated method stub
return false;
}
@Override
public int getTimeout() {
// TODO Auto-generated method stub
return 0;
}
@Override
public boolean processTimeout(long agentId, long seq) {
// TODO Auto-generated method stub
return false;
}
@Override
public boolean configure(final String name, final Map<String, Object> params) throws ConfigurationException {
_agentMgr.registerForHostEvents(this, true, false, true);
VirtualMachine.State.getStateMachine().registerListener(this);
_messageBus.subscribe("VM_ReservedCapacity_Free", new MessageSubscriber() {
@Override
public void onPublishMessage(String senderAddress, String subject, Object obj) {
VMInstanceVO vm = ((VMInstanceVO)obj);
s_logger.debug("MessageBus message: host reserved capacity released for VM: " + vm.getLastHostId() +
", checking if host reservation can be released for host:" + vm.getLastHostId());
Long hostId = vm.getLastHostId();
checkHostReservationRelease(hostId);
}
});
_vmCapacityReleaseInterval = NumbersUtil.parseInt(_configDao.getValue(Config.CapacitySkipcountingHours.key()), 3600);
String hostReservationReleasePeriod = _configDao.getValue(Config.HostReservationReleasePeriod.key());
if (hostReservationReleasePeriod != null) {
_hostReservationReleasePeriod = Long.parseLong(hostReservationReleasePeriod);
if (_hostReservationReleasePeriod <= 0)
_hostReservationReleasePeriod = Long.parseLong(Config.HostReservationReleasePeriod.getDefaultValue());
}
_timer = new Timer("HostReservationReleaseChecker");
_nodeId = ManagementServerNode.getManagementServerId();
return super.configure(name, params);
}
@Override
public boolean start() {
_timer.schedule(new HostReservationReleaseChecker(), INITIAL_RESERVATION_RELEASE_CHECKER_DELAY, _hostReservationReleasePeriod);
cleanupVMReservations();
return true;
}
@Override
public boolean stop() {
_timer.cancel();
return true;
}
@Override
public void cleanupVMReservations() {
List<VMReservationVO> reservations = _reservationDao.listAll();
for (VMReservationVO reserv : reservations) {
VMInstanceVO vm = _vmInstanceDao.findById(reserv.getVmId());
if (vm != null) {
if (vm.getState() == State.Starting || (vm.getState() == State.Stopped && vm.getLastHostId() == null)) {
continue;
} else {
// delete reservation
_reservationDao.remove(reserv.getId());
}
} else {
// delete reservation
_reservationDao.remove(reserv.getId());
}
}
}
// /refactoring planner methods
private DeployDestination checkClustersforDestination(List<Long> clusterList, VirtualMachineProfile vmProfile, DeploymentPlan plan, ExcludeList avoid, DataCenter dc,
DeploymentPlanner.PlannerResourceUsage resourceUsageRequired, ExcludeList plannerAvoidOutput) {
if (s_logger.isTraceEnabled()) {
s_logger.trace("ClusterId List to consider: " + clusterList);
}
for (Long clusterId : clusterList) {
ClusterVO clusterVO = _clusterDao.findById(clusterId);
if (clusterVO.getHypervisorType() != vmProfile.getHypervisorType()) {
s_logger.debug("Cluster: " + clusterId + " has HyperVisorType that does not match the VM, skipping this cluster");
avoid.addCluster(clusterVO.getId());
continue;
}
s_logger.debug("Checking resources in Cluster: " + clusterId + " under Pod: " + clusterVO.getPodId());
// search for resources(hosts and storage) under this zone, pod,
// cluster.
DataCenterDeployment potentialPlan =
new DataCenterDeployment(plan.getDataCenterId(), clusterVO.getPodId(), clusterVO.getId(), null, plan.getPoolId(), null, plan.getReservationContext());
potentialPlan.setHostPriorities(plan.getHostPriorities());
Pod pod = _podDao.findById(clusterVO.getPodId());
if (CollectionUtils.isNotEmpty(avoid.getPodsToAvoid()) && avoid.getPodsToAvoid().contains(pod.getId())) {
s_logger.debug("The cluster is in a disabled pod : " + pod.getId());
} else {
// find suitable hosts under this cluster, need as many hosts as we
// get.
List<Host> suitableHosts = findSuitableHosts(vmProfile, potentialPlan, avoid, HostAllocator.RETURN_UPTO_ALL);
// if found suitable hosts in this cluster, find suitable storage
// pools for each volume of the VM
if (suitableHosts != null && !suitableHosts.isEmpty()) {
if (vmProfile.getHypervisorType() == HypervisorType.BareMetal) {
DeployDestination dest = new DeployDestination(dc, pod, clusterVO, suitableHosts.get(0));
return dest;
}
Pair<Map<Volume, List<StoragePool>>, List<Volume>> result = findSuitablePoolsForVolumes(vmProfile, potentialPlan, avoid, StoragePoolAllocator.RETURN_UPTO_ALL);
Map<Volume, List<StoragePool>> suitableVolumeStoragePools = result.first();
List<Volume> readyAndReusedVolumes = result.second();
// choose the potential host and pool for the VM
if (!suitableVolumeStoragePools.isEmpty()) {
Pair<Host, Map<Volume, StoragePool>> potentialResources = findPotentialDeploymentResources(suitableHosts, suitableVolumeStoragePools, avoid,
resourceUsageRequired, readyAndReusedVolumes, plan.getPreferredHosts(), vmProfile.getVirtualMachine());
if (potentialResources != null) {
Host host = _hostDao.findById(potentialResources.first().getId());
Map<Volume, StoragePool> storageVolMap = potentialResources.second();
// remove the reused vol<->pool from destination, since
// we don't have to prepare this volume.
for (Volume vol : readyAndReusedVolumes) {
storageVolMap.remove(vol);
}
boolean displayStorage = getDisplayStorageFromVmProfile(vmProfile);
DeployDestination dest = new DeployDestination(dc, pod, clusterVO, host, storageVolMap, displayStorage);
s_logger.debug("Returning Deployment Destination: " + dest);
return dest;
}
} else {
s_logger.debug("No suitable storagePools found under this Cluster: " + clusterId);
}
} else {
s_logger.debug("No suitable hosts found under this Cluster: " + clusterId);
}
}
if (canAvoidCluster(clusterVO, avoid, plannerAvoidOutput, vmProfile)) {
avoid.addCluster(clusterVO.getId());
}
}
s_logger.debug("Could not find suitable Deployment Destination for this VM under any clusters, returning. ");
return null;
}
private boolean canAvoidCluster(Cluster clusterVO, ExcludeList avoids, ExcludeList plannerAvoidOutput, VirtualMachineProfile vmProfile) {
ExcludeList allocatorAvoidOutput =
new ExcludeList(avoids.getDataCentersToAvoid(), avoids.getPodsToAvoid(), avoids.getClustersToAvoid(), avoids.getHostsToAvoid(), avoids.getPoolsToAvoid());
// remove any hosts/pools that the planners might have added
// to get the list of hosts/pools that Allocators flagged as 'avoid'
resetAvoidSet(allocatorAvoidOutput, plannerAvoidOutput);
// if all hosts or all pools in the cluster are in avoid set after this
// pass, then put the cluster in avoid set.
boolean avoidAllHosts = true;
boolean avoidAllPools = true;
boolean avoidAllLocalPools = true;
boolean avoidAllSharedPools = true;
List<HostVO> allhostsInCluster =
_hostDao.listAllUpAndEnabledNonHAHosts(Host.Type.Routing, clusterVO.getId(), clusterVO.getPodId(), clusterVO.getDataCenterId(), null);
for (HostVO host : allhostsInCluster) {
if (!allocatorAvoidOutput.shouldAvoid(host)) {
// there's some host in the cluster that is not yet in avoid set
avoidAllHosts = false;
break;
}
}
// all hosts in avoid set, avoid the cluster. Otherwise check the pools
if (avoidAllHosts) {
return true;
}
// Cluster can be put in avoid set in following scenarios:
// 1. If storage allocators haven't put any pools in avoid set means either no pools in cluster
// or pools not suitable for the allocators to handle or there is no
// linkage of any suitable host to any of the pools in cluster
// 2. If all 'shared' or 'local' pools are in avoid set
if (allocatorAvoidOutput.getPoolsToAvoid() != null && !allocatorAvoidOutput.getPoolsToAvoid().isEmpty()) {
Pair<Boolean, Boolean> storageRequirements = findVMStorageRequirements(vmProfile);
boolean vmRequiresSharedStorage = storageRequirements.first();
boolean vmRequiresLocalStorege = storageRequirements.second();
if (vmRequiresSharedStorage) {
// check shared pools
List<StoragePoolVO> allPoolsInCluster = _storagePoolDao.findPoolsByTags(clusterVO.getDataCenterId(), clusterVO.getPodId(), clusterVO.getId(), null, false, 0);
for (StoragePoolVO pool : allPoolsInCluster) {
if (!allocatorAvoidOutput.shouldAvoid(pool)) {
// there's some pool in the cluster that is not yet in avoid set
avoidAllSharedPools = false;
break;
}
}
}
if (vmRequiresLocalStorege) {
// check local pools
List<StoragePoolVO> allLocalPoolsInCluster =
_storagePoolDao.findLocalStoragePoolsByTags(clusterVO.getDataCenterId(), clusterVO.getPodId(), clusterVO.getId(), null, false);
for (StoragePoolVO pool : allLocalPoolsInCluster) {
if (!allocatorAvoidOutput.shouldAvoid(pool)) {
// there's some pool in the cluster that is not yet
// in avoid set
avoidAllLocalPools = false;
break;
}
}
}
if (vmRequiresSharedStorage && vmRequiresLocalStorege) {
avoidAllPools = (avoidAllLocalPools || avoidAllSharedPools) ? true : false;
} else if (vmRequiresSharedStorage) {
avoidAllPools = avoidAllSharedPools;
} else if (vmRequiresLocalStorege) {
avoidAllPools = avoidAllLocalPools;
}
}
if (avoidAllHosts || avoidAllPools) {
return true;
}
return false;
}
private Pair<Boolean, Boolean> findVMStorageRequirements(VirtualMachineProfile vmProfile) {
boolean requiresShared = false, requiresLocal = false;
List<VolumeVO> volumesTobeCreated = _volsDao.findUsableVolumesForInstance(vmProfile.getId());
// for each volume find whether shared or local pool is required
for (VolumeVO toBeCreated : volumesTobeCreated) {
DiskOfferingVO diskOffering = _diskOfferingDao.findById(toBeCreated.getDiskOfferingId());
if (diskOffering != null) {
if (diskOffering.isUseLocalStorage()) {
requiresLocal = true;
} else {
requiresShared = true;
}
}
}
return new Pair<Boolean, Boolean>(requiresShared, requiresLocal);
}
protected Pair<Host, Map<Volume, StoragePool>> findPotentialDeploymentResources(List<Host> suitableHosts, Map<Volume, List<StoragePool>> suitableVolumeStoragePools,
ExcludeList avoid, PlannerResourceUsage resourceUsageRequired, List<Volume> readyAndReusedVolumes, List<Long> preferredHosts, VirtualMachine vm) {
s_logger.debug("Trying to find a potenial host and associated storage pools from the suitable host/pool lists for this VM");
boolean hostCanAccessPool = false;
boolean haveEnoughSpace = false;
boolean hostAffinityCheck = false;
if (readyAndReusedVolumes == null) {
readyAndReusedVolumes = new ArrayList<Volume>();
}
Map<Volume, StoragePool> storage = new HashMap<Volume, StoragePool>();
TreeSet<Volume> volumesOrderBySizeDesc = new TreeSet<Volume>(new Comparator<Volume>() {
@Override
public int compare(Volume v1, Volume v2) {
if (v1.getSize() < v2.getSize())
return 1;
else
return -1;
}
});
volumesOrderBySizeDesc.addAll(suitableVolumeStoragePools.keySet());
boolean multipleVolume = volumesOrderBySizeDesc.size() > 1;
boolean deployAsIs = isDeployAsIs(vm);
for (Host potentialHost : suitableHosts) {
Map<StoragePool, List<Volume>> volumeAllocationMap = new HashMap<StoragePool, List<Volume>>();
if (deployAsIs) {
storage = new HashMap<>();
// Find the common suitable pools
s_logger.debug("Trying to allocate all the VM volumes to a single storage pool");
Set<StoragePool> suitablePools = new HashSet<>();
List<StoragePool> notAllowedPools = new ArrayList<>();
for (List<StoragePool> pools : suitableVolumeStoragePools.values()) {
if (CollectionUtils.isEmpty(suitablePools)) {
// All the suitable pools of the first volume
suitablePools.addAll(pools);
} else {
for (StoragePool pool : pools) {
if (!suitablePools.contains(pool)) {
s_logger.debug("Storage pool " + pool.getUuid() + " not allowed for this VM");
notAllowedPools.add(pool);
}
}
}
}
suitablePools.removeAll(notAllowedPools);
if (CollectionUtils.isEmpty(suitablePools)) {
s_logger.debug("Could not find a storage pool to fit all the VM volumes on this host");
continue;
}
List<Volume> allVolumes = new ArrayList<>();
allVolumes.addAll(volumesOrderBySizeDesc);
List<Pair<Volume, DiskProfile>> volumeDiskProfilePair = getVolumeDiskProfilePairs(allVolumes);
for (StoragePool storagePool : suitablePools) {
haveEnoughSpace = false;
hostCanAccessPool = false;
hostAffinityCheck = checkAffinity(potentialHost, preferredHosts);
if (hostCanAccessSPool(potentialHost, storagePool)) {
hostCanAccessPool = true;
if (potentialHost.getHypervisorType() == HypervisorType.VMware) {
try {
boolean isStoragePoolStoragepolicyComplaince = _storageMgr.isStoragePoolCompliantWithStoragePolicy(volumeDiskProfilePair, storagePool);
if (!isStoragePoolStoragepolicyComplaince) {
continue;
}
} catch (StorageUnavailableException e) {
s_logger.warn(String.format("Could not verify storage policy complaince against storage pool %s due to exception %s", storagePool.getUuid(), e.getMessage()));
continue;
}
haveEnoughSpace = true;
}
}
if (hostCanAccessPool && haveEnoughSpace && hostAffinityCheck) {
for (Volume vol : volumesOrderBySizeDesc) {
s_logger.debug("Found a suitable storage pool for all the VM volumes: " + storagePool.getUuid());
storage.put(vol, storagePool);
}
break;
}
}
} else {
for (Volume vol : volumesOrderBySizeDesc) {
haveEnoughSpace = false;
s_logger.debug("Checking if host: " + potentialHost.getId() + " can access any suitable storage pool for volume: " + vol.getVolumeType());
List<StoragePool> volumePoolList = suitableVolumeStoragePools.get(vol);
hostCanAccessPool = false;
hostAffinityCheck = checkAffinity(potentialHost, preferredHosts);
for (StoragePool potentialSPool : volumePoolList) {
if (hostCanAccessSPool(potentialHost, potentialSPool)) {
hostCanAccessPool = true;
if (multipleVolume && !readyAndReusedVolumes.contains(vol)) {
List<Volume> requestVolumes = null;
if (volumeAllocationMap.containsKey(potentialSPool))
requestVolumes = volumeAllocationMap.get(potentialSPool);
else
requestVolumes = new ArrayList<Volume>();
requestVolumes.add(vol);
List<Pair<Volume, DiskProfile>> volumeDiskProfilePair = getVolumeDiskProfilePairs(requestVolumes);
if (potentialHost.getHypervisorType() == HypervisorType.VMware) {
try {
boolean isStoragePoolStoragepolicyComplaince = _storageMgr.isStoragePoolCompliantWithStoragePolicy(volumeDiskProfilePair, potentialSPool);
if (!isStoragePoolStoragepolicyComplaince) {
continue;
}
} catch (StorageUnavailableException e) {
s_logger.warn(String.format("Could not verify storage policy complaince against storage pool %s due to exception %s", potentialSPool.getUuid(), e.getMessage()));
continue;
}
}
if (!_storageMgr.storagePoolHasEnoughIops(volumeDiskProfilePair, potentialSPool) ||
!_storageMgr.storagePoolHasEnoughSpace(volumeDiskProfilePair, potentialSPool, potentialHost.getClusterId()))
continue;
volumeAllocationMap.put(potentialSPool, requestVolumes);
}
storage.put(vol, potentialSPool);
haveEnoughSpace = true;
break;
}
}
if (!hostCanAccessPool) {
break;
}
if (!haveEnoughSpace) {
s_logger.warn("insufficient capacity to allocate all volumes");
break;
}
if (!hostAffinityCheck) {
s_logger.debug("Host affinity check failed");
break;
}
}
}
HostVO potentialHostVO = _hostDao.findById(potentialHost.getId());
_hostDao.loadDetails(potentialHostVO);
boolean hostHasEncryption = Boolean.parseBoolean(potentialHostVO.getDetail(Host.HOST_VOLUME_ENCRYPTION));
boolean hostMeetsEncryptionRequirements = !anyVolumeRequiresEncryption(new ArrayList<>(volumesOrderBySizeDesc)) || hostHasEncryption;
boolean plannerUsageFits = checkIfHostFitsPlannerUsage(potentialHost.getId(), resourceUsageRequired);
if (hostCanAccessPool && haveEnoughSpace && hostAffinityCheck && hostMeetsEncryptionRequirements && plannerUsageFits) {
s_logger.debug("Found a potential host " + "id: " + potentialHost.getId() + " name: " + potentialHost.getName() +
" and associated storage pools for this VM");
volumeAllocationMap.clear();
return new Pair<Host, Map<Volume, StoragePool>>(potentialHost, storage);
} else {
if (!hostMeetsEncryptionRequirements) {
s_logger.debug("Potential host " + potentialHost + " did not meet encryption requirements of all volumes");
}
avoid.addHost(potentialHost.getId());
}
}
s_logger.debug("Could not find a potential host that has associated storage pools from the suitable host/pool lists for this VM");
return null;
}
private List<Pair<Volume, DiskProfile>> getVolumeDiskProfilePairs(List<Volume> volumes) {
List<Pair<Volume, DiskProfile>> volumeDiskProfilePairs = new ArrayList<>();
for (Volume volume: volumes) {
DiskOfferingVO diskOffering = _diskOfferingDao.findById(volume.getDiskOfferingId());
DiskProfile diskProfile = new DiskProfile(volume, diskOffering, _volsDao.getHypervisorType(volume.getId()));
volumeDiskProfilePairs.add(new Pair<>(volume, diskProfile));
}
return volumeDiskProfilePairs;
}
/**
* True if:
* - Affinity is not enabled (preferred host is empty)
* - Affinity is enabled and potential host is on the preferred hosts list
*
* False if not
*/
@DB
public boolean checkAffinity(Host potentialHost, List<Long> preferredHosts) {
boolean hostAffinityEnabled = CollectionUtils.isNotEmpty(preferredHosts);
boolean hostAffinityMatches = hostAffinityEnabled && preferredHosts.contains(potentialHost.getId());
return !hostAffinityEnabled || hostAffinityMatches;
}
protected boolean hostCanAccessSPool(Host host, StoragePool pool) {
boolean hostCanAccessSPool = false;
StoragePoolHostVO hostPoolLinkage = _poolHostDao.findByPoolHost(pool.getId(), host.getId());
if (hostPoolLinkage != null && _storageMgr.canHostAccessStoragePool(host, pool)) {
hostCanAccessSPool = true;
}
s_logger.debug("Host: " + host.getId() + (hostCanAccessSPool ? " can" : " cannot") + " access pool: " + pool.getId());
return hostCanAccessSPool;
}
protected List<Host> findSuitableHosts(VirtualMachineProfile vmProfile, DeploymentPlan plan, ExcludeList avoid, int returnUpTo) {
List<Host> suitableHosts = new ArrayList<Host>();
for (HostAllocator allocator : _hostAllocators) {
suitableHosts = allocator.allocateTo(vmProfile, plan, Host.Type.Routing, avoid, returnUpTo);
if (suitableHosts != null && !suitableHosts.isEmpty()) {
break;
}
}
if (suitableHosts.isEmpty()) {
s_logger.debug("No suitable hosts found");
}
// re-order hosts by priority
reorderHostsByPriority(plan.getHostPriorities(), suitableHosts);
return suitableHosts;
}
@Override
public void reorderHostsByPriority(Map<Long, Integer> priorities, List<Host> hosts) {
s_logger.info("Re-ordering hosts " + hosts + " by priorities " + priorities);
hosts.removeIf(host -> DataCenterDeployment.PROHIBITED_HOST_PRIORITY.equals(getHostPriority(priorities, host.getId())));
Collections.sort(hosts, new Comparator<>() {
@Override
public int compare(Host host1, Host host2) {
int res = getHostPriority(priorities, host1.getId()).compareTo(getHostPriority(priorities, host2.getId()));
return -res;
}
}
);
s_logger.info("Hosts after re-ordering are: " + hosts);
}
private Integer getHostPriority(Map<Long, Integer> priorities, Long hostId) {
return priorities.get(hostId) != null ? priorities.get(hostId) : DeploymentPlan.DEFAULT_HOST_PRIORITY;
}
protected Pair<Map<Volume, List<StoragePool>>, List<Volume>> findSuitablePoolsForVolumes(VirtualMachineProfile vmProfile, DeploymentPlan plan, ExcludeList avoid,
int returnUpTo) {
List<VolumeVO> volumesTobeCreated = _volsDao.findUsableVolumesForInstance(vmProfile.getId());
Map<Volume, List<StoragePool>> suitableVolumeStoragePools = new HashMap<Volume, List<StoragePool>>();
List<Volume> readyAndReusedVolumes = new ArrayList<Volume>();
// There should be at least the ROOT volume of the VM in usable state
if (volumesTobeCreated.isEmpty()) {
// OfflineVmwareMigration: find out what is wrong with the id of the vm we try to start
throw new CloudRuntimeException("Unable to create deployment, no usable volumes found for the VM: " + vmProfile.getId());
}
// don't allow to start vm that doesn't have a root volume
if (_volsDao.findByInstanceAndType(vmProfile.getId(), Volume.Type.ROOT).isEmpty()) {
throw new CloudRuntimeException("Unable to prepare volumes for vm as ROOT volume is missing");
}
// for each volume find list of suitable storage pools by calling the
// allocators
Set<Long> originalAvoidPoolSet = new HashSet<>();
if (avoid.getPoolsToAvoid() != null) {
originalAvoidPoolSet.addAll(avoid.getPoolsToAvoid());
}
Set<Long> poolsToAvoidOutput = new HashSet<>(originalAvoidPoolSet);
for (VolumeVO toBeCreated : volumesTobeCreated) {
s_logger.debug("Checking suitable pools for volume (Id, Type): (" + toBeCreated.getId() + "," + toBeCreated.getVolumeType().name() + ")");
if (toBeCreated.getState() == Volume.State.Allocated && toBeCreated.getPoolId() != null) {
toBeCreated.setPoolId(null);
if (!_volsDao.update(toBeCreated.getId(), toBeCreated)) {
throw new CloudRuntimeException(String.format("Error updating volume [%s] to clear pool Id.", toBeCreated.getId()));
}
if (s_logger.isDebugEnabled()) {
String msg = String.format("Setting pool_id to NULL for volume id=%s as it is in Allocated state", toBeCreated.getId());
s_logger.debug(msg);
}
}
// If the plan specifies a poolId, it means that this VM's ROOT
// volume is ready and the pool should be reused.
// In this case, also check if rest of the volumes are ready and can
// be reused.
if (plan.getPoolId() != null || (toBeCreated.getVolumeType() == Volume.Type.DATADISK && toBeCreated.getPoolId() != null && toBeCreated.getState() == Volume.State.Ready)) {
s_logger.debug("Volume has pool already allocated, checking if pool can be reused, poolId: " + toBeCreated.getPoolId());
List<StoragePool> suitablePools = new ArrayList<StoragePool>();
StoragePool pool = null;
if (toBeCreated.getPoolId() != null) {
pool = (StoragePool)dataStoreMgr.getPrimaryDataStore(toBeCreated.getPoolId());
} else {
pool = (StoragePool)dataStoreMgr.getPrimaryDataStore(plan.getPoolId());
}
if (!pool.isInMaintenance()) {
if (!avoid.shouldAvoid(pool)) {
long exstPoolDcId = pool.getDataCenterId();
long exstPoolPodId = pool.getPodId() != null ? pool.getPodId() : -1;
long exstPoolClusterId = pool.getClusterId() != null ? pool.getClusterId() : -1;
boolean canReusePool = false;
if (plan.getDataCenterId() == exstPoolDcId && plan.getPodId() == exstPoolPodId && plan.getClusterId() == exstPoolClusterId) {
canReusePool = true;
} else if (plan.getDataCenterId() == exstPoolDcId) {
DataStore dataStore = dataStoreMgr.getPrimaryDataStore(pool.getId());
if (dataStore != null && dataStore.getScope() != null && dataStore.getScope().getScopeType() == ScopeType.ZONE) {
canReusePool = true;
}
} else {
s_logger.debug("Pool of the volume does not fit the specified plan, need to reallocate a pool for this volume");
canReusePool = false;
}
if (canReusePool) {
s_logger.debug("Planner need not allocate a pool for this volume since its READY");
suitablePools.add(pool);
suitableVolumeStoragePools.put(toBeCreated, suitablePools);
if (!(toBeCreated.getState() == Volume.State.Allocated || toBeCreated.getState() == Volume.State.Creating)) {
readyAndReusedVolumes.add(toBeCreated);
}
continue;
}
} else {
s_logger.debug("Pool of the volume is in avoid set, need to reallocate a pool for this volume");
}
} else {
s_logger.debug("Pool of the volume is in maintenance, need to reallocate a pool for this volume");
}
}
if (s_logger.isDebugEnabled()) {
s_logger.debug("We need to allocate new storagepool for this volume");
}
if (!isRootAdmin(vmProfile)) {
if (!isEnabledForAllocation(plan.getDataCenterId(), plan.getPodId(), plan.getClusterId())) {
if (s_logger.isDebugEnabled()) {
s_logger.debug("Cannot allocate new storagepool for this volume in this cluster, allocation state is disabled");
s_logger.debug("Cannot deploy to this specified plan, allocation state is disabled, returning.");
}
// Cannot find suitable storage pools under this cluster for
// this volume since allocation_state is disabled.
// - remove any suitable pools found for other volumes.
// All volumes should get suitable pools under this cluster;
// else we can't use this cluster.
suitableVolumeStoragePools.clear();
break;
}
}
s_logger.debug("Calling StoragePoolAllocators to find suitable pools");
DiskOfferingVO diskOffering = _diskOfferingDao.findById(toBeCreated.getDiskOfferingId());
DiskProfile diskProfile = new DiskProfile(toBeCreated, diskOffering, vmProfile.getHypervisorType());
boolean useLocalStorage = false;
if (vmProfile.getType() != VirtualMachine.Type.User) {
DataCenterVO zone = _dcDao.findById(plan.getDataCenterId());
assert (zone != null) : "Invalid zone in deployment plan";
Boolean useLocalStorageForSystemVM = ConfigurationManagerImpl.SystemVMUseLocalStorage.valueIn(zone.getId());
if (useLocalStorageForSystemVM != null) {
useLocalStorage = useLocalStorageForSystemVM.booleanValue();
s_logger.debug("System VMs will use " + (useLocalStorage ? "local" : "shared") + " storage for zone id=" + plan.getDataCenterId());
}
} else {
useLocalStorage = diskOffering.isUseLocalStorage();
}
diskProfile.setUseLocalStorage(useLocalStorage);
boolean foundPotentialPools = false;
for (StoragePoolAllocator allocator : _storagePoolAllocators) {
final List<StoragePool> suitablePools = allocator.allocateToPool(diskProfile, vmProfile, plan, avoid, returnUpTo);
if (suitablePools != null && !suitablePools.isEmpty()) {
checkForPreferredStoragePool(suitablePools, vmProfile.getVirtualMachine(), suitableVolumeStoragePools, toBeCreated);
foundPotentialPools = true;
break;
}
}
if (avoid.getPoolsToAvoid() != null) {
poolsToAvoidOutput.addAll(avoid.getPoolsToAvoid());
avoid.getPoolsToAvoid().retainAll(originalAvoidPoolSet);
}
if (!foundPotentialPools) {
s_logger.debug("No suitable pools found for volume: " + toBeCreated + " under cluster: " + plan.getClusterId());
// No suitable storage pools found under this cluster for this
// volume. - remove any suitable pools found for other volumes.
// All volumes should get suitable pools under this cluster;
// else we can't use this cluster.
suitableVolumeStoragePools.clear();
break;
}
}
HashSet<Long> toRemove = new HashSet<Long>();
for (List<StoragePool> lsp : suitableVolumeStoragePools.values()) {
for (StoragePool sp : lsp) {
toRemove.add(sp.getId());
}
}
poolsToAvoidOutput.removeAll(toRemove);
if (avoid.getPoolsToAvoid() != null) {
avoid.getPoolsToAvoid().addAll(poolsToAvoidOutput);
}
if (suitableVolumeStoragePools.isEmpty()) {
s_logger.debug("No suitable pools found");
}
return new Pair<Map<Volume, List<StoragePool>>, List<Volume>>(suitableVolumeStoragePools, readyAndReusedVolumes);
}
private void checkForPreferredStoragePool(List<StoragePool> suitablePools,
VirtualMachine vm,
Map<Volume, List<StoragePool>> suitableVolumeStoragePools,
VolumeVO toBeCreated) {
List<StoragePool> pools = new ArrayList<>();
Optional<StoragePool> storagePool = getPreferredStoragePool(suitablePools, vm);
storagePool.ifPresent(pools::add);
pools.addAll(suitablePools);
suitableVolumeStoragePools.put(toBeCreated, pools);
}
private Optional<StoragePool> getMatchingStoragePool(String preferredPoolId, List<StoragePool> storagePools) {
if (preferredPoolId == null) {
return Optional.empty();
}
return storagePools.stream()
.filter(pool -> pool.getUuid().equalsIgnoreCase(preferredPoolId))
.findFirst();
}
private Optional<StoragePool> getPreferredStoragePool(List<StoragePool> poolList, VirtualMachine vm) {
String accountStoragePoolUuid = StorageManager.PreferredStoragePool.valueIn(vm.getAccountId());
Optional<StoragePool> storagePool = getMatchingStoragePool(accountStoragePoolUuid, poolList);
if (storagePool.isPresent()) {
s_logger.debug("A storage pool is specified for this account, so we will use this storage pool for allocation: "
+ storagePool.get().getUuid());
} else {
String globalStoragePoolUuid = StorageManager.PreferredStoragePool.value();
storagePool = getMatchingStoragePool(globalStoragePoolUuid, poolList);
storagePool.ifPresent(pool -> s_logger.debug("A storage pool is specified in global setting, so we will use this storage pool for allocation: "
+ pool.getUuid()));
}
return storagePool;
}
private boolean isEnabledForAllocation(long zoneId, Long podId, Long clusterId) {
// Check if the zone exists in the system
DataCenterVO zone = _dcDao.findById(zoneId);
if (zone != null && Grouping.AllocationState.Disabled == zone.getAllocationState()) {
s_logger.info("Zone is currently disabled, cannot allocate to this zone: " + zoneId);
return false;
}
Pod pod = _podDao.findById(podId);
if (pod != null && Grouping.AllocationState.Disabled == pod.getAllocationState()) {
s_logger.info("Pod is currently disabled, cannot allocate to this pod: " + podId);
return false;
}
Cluster cluster = _clusterDao.findById(clusterId);
if (cluster != null && Grouping.AllocationState.Disabled == cluster.getAllocationState()) {
s_logger.info("Cluster is currently disabled, cannot allocate to this cluster: " + clusterId);
return false;
}
return true;
}
private boolean isRootAdmin(VirtualMachineProfile vmProfile) {
if (vmProfile != null) {
if (vmProfile.getOwner() != null) {
return _accountMgr.isRootAdmin(vmProfile.getOwner().getId());
} else {
return false;
}
}
return false;
}
@DB
@Override
public String finalizeReservation(final DeployDestination plannedDestination, final VirtualMachineProfile vmProfile, DeploymentPlan plan, ExcludeList avoids, final DeploymentPlanner planner)
throws InsufficientServerCapacityException, AffinityConflictException {
final VirtualMachine vm = vmProfile.getVirtualMachine();
final long vmGroupCount = _affinityGroupVMMapDao.countAffinityGroupsForVm(vm.getId());
return Transaction.execute(new TransactionCallback<String>() {
@Override
public String doInTransaction(TransactionStatus status) {
boolean saveReservation = true;
if (vmGroupCount > 0) {
List<Long> groupIds = _affinityGroupVMMapDao.listAffinityGroupIdsByVmId(vm.getId());
SearchCriteria<AffinityGroupVO> criteria = _affinityGroupDao.createSearchCriteria();
criteria.addAnd("id", SearchCriteria.Op.IN, groupIds.toArray(new Object[groupIds.size()]));
_affinityGroupDao.lockRows(criteria, null, true);
for (AffinityGroupProcessor processor : _affinityProcessors) {
if (!processor.check(vmProfile, plannedDestination)) {
saveReservation = false;
break;
}
}
}
if (saveReservation) {
VMReservationVO vmReservation =
new VMReservationVO(vm.getId(), plannedDestination.getDataCenter().getId(), plannedDestination.getPod().getId(), plannedDestination.getCluster()
.getId(), plannedDestination.getHost().getId());
if (planner != null) {
vmReservation.setDeploymentPlanner(planner.getName());
}
Map<Long, Long> volumeReservationMap = new HashMap<Long, Long>();
if (vm.getHypervisorType() != HypervisorType.BareMetal) {
for (Volume vo : plannedDestination.getStorageForDisks().keySet()) {
volumeReservationMap.put(vo.getId(), plannedDestination.getStorageForDisks().get(vo).getId());
}
vmReservation.setVolumeReservation(volumeReservationMap);
}
_reservationDao.persist(vmReservation);
return vmReservation.getUuid();
}
return null;
}
});
}
@Override
public boolean preStateTransitionEvent(State oldState, Event event, State newState, VirtualMachine vo, boolean status, Object opaque) {
return true;
}
@Override
public boolean postStateTransitionEvent(StateMachine2.Transition<State, Event> transition, VirtualMachine vo, boolean status, Object opaque) {
if (!status) {
return false;
}
State oldState = transition.getCurrentState();
State newState = transition.getToState();
if ((oldState == State.Starting) && (newState != State.Starting)) {
// cleanup all VM reservation entries
SearchCriteria<VMReservationVO> sc = _reservationDao.createSearchCriteria();
sc.addAnd("vmId", SearchCriteria.Op.EQ, vo.getId());
_reservationDao.expunge(sc);
}
return true;
}
@Override
public ConfigKey<?>[] getConfigKeys() {
return new ConfigKey<?>[] {allowRouterOnDisabledResource, allowAdminVmOnDisabledResource};
}
@Override
public String getConfigComponentName() {
return DeploymentPlanningManager.class.getSimpleName();
}
}