| /* |
| * Copyright 2014-present Open Networking Foundation |
| * |
| * Licensed 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 org.onosproject.cluster.impl; |
| |
| import com.codahale.metrics.Timer; |
| import com.codahale.metrics.Timer.Context; |
| import com.google.common.collect.Lists; |
| import com.google.common.collect.Sets; |
| import com.google.common.util.concurrent.Futures; |
| import org.apache.felix.scr.annotations.Activate; |
| import org.apache.felix.scr.annotations.Component; |
| import org.apache.felix.scr.annotations.Deactivate; |
| import org.apache.felix.scr.annotations.Modified; |
| import org.apache.felix.scr.annotations.Property; |
| import org.apache.felix.scr.annotations.Reference; |
| import org.apache.felix.scr.annotations.ReferenceCardinality; |
| import org.apache.felix.scr.annotations.Service; |
| import org.onlab.metrics.MetricsService; |
| import org.onosproject.cfg.ComponentConfigService; |
| import org.onosproject.cfg.ConfigProperty; |
| import org.onosproject.cluster.ClusterService; |
| import org.onosproject.cluster.ControllerNode; |
| import org.onosproject.cluster.NodeId; |
| import org.onosproject.cluster.RoleInfo; |
| import org.onosproject.core.MetricsHelper; |
| import org.onosproject.event.AbstractListenerManager; |
| import org.onosproject.mastership.MastershipAdminService; |
| import org.onosproject.mastership.MastershipEvent; |
| import org.onosproject.mastership.MastershipListener; |
| import org.onosproject.mastership.MastershipService; |
| import org.onosproject.mastership.MastershipStore; |
| import org.onosproject.mastership.MastershipStoreDelegate; |
| import org.onosproject.mastership.MastershipTerm; |
| import org.onosproject.mastership.MastershipTermService; |
| import org.onosproject.net.DeviceId; |
| import org.onosproject.net.MastershipRole; |
| import org.onosproject.net.region.Region; |
| import org.onosproject.net.region.RegionService; |
| import org.onosproject.upgrade.UpgradeEvent; |
| import org.onosproject.upgrade.UpgradeEventListener; |
| import org.onosproject.upgrade.UpgradeService; |
| import org.slf4j.Logger; |
| |
| import java.util.ArrayList; |
| import java.util.Collection; |
| import java.util.Collections; |
| import java.util.Comparator; |
| import java.util.HashMap; |
| import java.util.HashSet; |
| import java.util.Iterator; |
| import java.util.List; |
| import java.util.Map; |
| import java.util.Set; |
| import java.util.concurrent.CompletableFuture; |
| |
| import static com.google.common.base.Preconditions.checkNotNull; |
| import static com.google.common.collect.Lists.newArrayList; |
| import static java.util.concurrent.CompletableFuture.allOf; |
| import static org.onlab.metrics.MetricsUtil.startTimer; |
| import static org.onlab.metrics.MetricsUtil.stopTimer; |
| import static org.onosproject.net.MastershipRole.MASTER; |
| import static org.onosproject.security.AppGuard.checkPermission; |
| import static org.onosproject.security.AppPermission.Type.CLUSTER_READ; |
| import static org.onosproject.security.AppPermission.Type.CLUSTER_WRITE; |
| import static org.slf4j.LoggerFactory.getLogger; |
| |
| |
| /** |
| * Component providing the node-device mastership service. |
| */ |
| @Component(immediate = true) |
| @Service |
| public class MastershipManager |
| extends AbstractListenerManager<MastershipEvent, MastershipListener> |
| implements MastershipService, MastershipAdminService, MastershipTermService, |
| MetricsHelper { |
| |
| private static final String NODE_ID_NULL = "Node ID cannot be null"; |
| private static final String DEVICE_ID_NULL = "Device ID cannot be null"; |
| private static final String ROLE_NULL = "Mastership role cannot be null"; |
| |
| private final Logger log = getLogger(getClass()); |
| |
| private final MastershipStoreDelegate delegate = new InternalDelegate(); |
| private final UpgradeEventListener upgradeEventListener = new InternalUpgradeEventListener(); |
| |
| @Reference(cardinality = ReferenceCardinality.MANDATORY_UNARY) |
| protected MastershipStore store; |
| |
| @Reference(cardinality = ReferenceCardinality.MANDATORY_UNARY) |
| protected ClusterService clusterService; |
| |
| @Reference(cardinality = ReferenceCardinality.MANDATORY_UNARY) |
| protected MetricsService metricsService; |
| |
| @Reference(cardinality = ReferenceCardinality.MANDATORY_UNARY) |
| protected RegionService regionService; |
| |
| @Reference(cardinality = ReferenceCardinality.MANDATORY_UNARY) |
| protected ComponentConfigService cfgService; |
| |
| @Reference(cardinality = ReferenceCardinality.MANDATORY_UNARY) |
| protected UpgradeService upgradeService; |
| |
| private NodeId localNodeId; |
| private Timer requestRoleTimer; |
| |
| static final boolean DEFAULT_USE_REGION_FOR_BALANCE_ROLES = false; |
| @Property(name = "useRegionForBalanceRoles", boolValue = DEFAULT_USE_REGION_FOR_BALANCE_ROLES, |
| label = "Use Regions for balancing roles") |
| protected boolean useRegionForBalanceRoles; |
| |
| private static final boolean DEFAULT_REBALANCE_ROLES_ON_UPGRADE = true; |
| @Property(name = "rebalanceRolesOnUpgrade", |
| boolValue = DEFAULT_REBALANCE_ROLES_ON_UPGRADE, |
| label = "Automatically rebalance roles following an upgrade") |
| protected boolean rebalanceRolesOnUpgrade = DEFAULT_REBALANCE_ROLES_ON_UPGRADE; |
| |
| @Activate |
| public void activate() { |
| cfgService.registerProperties(getClass()); |
| modified(); |
| |
| requestRoleTimer = createTimer("Mastership", "requestRole", "responseTime"); |
| localNodeId = clusterService.getLocalNode().id(); |
| upgradeService.addListener(upgradeEventListener); |
| eventDispatcher.addSink(MastershipEvent.class, listenerRegistry); |
| store.setDelegate(delegate); |
| log.info("Started"); |
| } |
| |
| @Modified |
| public void modified() { |
| Set<ConfigProperty> configProperties = cfgService.getProperties(getClass().getCanonicalName()); |
| for (ConfigProperty property : configProperties) { |
| if ("useRegionForBalanceRoles".equals(property.name())) { |
| useRegionForBalanceRoles = property.asBoolean(); |
| } |
| } |
| } |
| |
| @Deactivate |
| public void deactivate() { |
| eventDispatcher.removeSink(MastershipEvent.class); |
| upgradeService.removeListener(upgradeEventListener); |
| store.unsetDelegate(delegate); |
| log.info("Stopped"); |
| cfgService.unregisterProperties(getClass(), false); |
| } |
| |
| @Override |
| public CompletableFuture<Void> setRole(NodeId nodeId, DeviceId deviceId, MastershipRole role) { |
| checkNotNull(nodeId, NODE_ID_NULL); |
| checkNotNull(deviceId, DEVICE_ID_NULL); |
| checkNotNull(role, ROLE_NULL); |
| |
| CompletableFuture<MastershipEvent> eventFuture = null; |
| |
| switch (role) { |
| case MASTER: |
| eventFuture = store.setMaster(nodeId, deviceId); |
| break; |
| case STANDBY: |
| eventFuture = store.setStandby(nodeId, deviceId); |
| break; |
| case NONE: |
| eventFuture = store.relinquishRole(nodeId, deviceId); |
| break; |
| default: |
| log.info("Unknown role; ignoring"); |
| return CompletableFuture.completedFuture(null); |
| } |
| |
| return eventFuture.thenAccept(this::post) |
| .thenApply(v -> null); |
| } |
| |
| @Override |
| public MastershipRole getLocalRole(DeviceId deviceId) { |
| checkPermission(CLUSTER_READ); |
| |
| checkNotNull(deviceId, DEVICE_ID_NULL); |
| return store.getRole(clusterService.getLocalNode().id(), deviceId); |
| } |
| |
| @Override |
| public CompletableFuture<Void> relinquishMastership(DeviceId deviceId) { |
| checkPermission(CLUSTER_WRITE); |
| return store.relinquishRole(localNodeId, deviceId) |
| .thenAccept(this::post) |
| .thenApply(v -> null); |
| } |
| |
| @Override |
| public CompletableFuture<MastershipRole> requestRoleFor(DeviceId deviceId) { |
| checkPermission(CLUSTER_WRITE); |
| |
| checkNotNull(deviceId, DEVICE_ID_NULL); |
| final Context timer = startTimer(requestRoleTimer); |
| return store.requestRole(deviceId).whenComplete((result, error) -> stopTimer(timer)); |
| |
| } |
| |
| @Override |
| public NodeId getMasterFor(DeviceId deviceId) { |
| checkPermission(CLUSTER_READ); |
| |
| checkNotNull(deviceId, DEVICE_ID_NULL); |
| return store.getMaster(deviceId); |
| } |
| |
| @Override |
| public Set<DeviceId> getDevicesOf(NodeId nodeId) { |
| checkPermission(CLUSTER_READ); |
| |
| checkNotNull(nodeId, NODE_ID_NULL); |
| return store.getDevices(nodeId); |
| } |
| |
| @Override |
| public RoleInfo getNodesFor(DeviceId deviceId) { |
| checkPermission(CLUSTER_READ); |
| |
| checkNotNull(deviceId, DEVICE_ID_NULL); |
| return store.getNodes(deviceId); |
| } |
| |
| @Override |
| public MastershipTerm getMastershipTerm(DeviceId deviceId) { |
| checkPermission(CLUSTER_READ); |
| return store.getTermFor(deviceId); |
| } |
| |
| @Override |
| public MetricsService metricsService() { |
| return metricsService; |
| } |
| |
| @Override |
| public void balanceRoles() { |
| List<ControllerNode> nodes = newArrayList(clusterService.getNodes()); |
| Map<ControllerNode, Set<DeviceId>> controllerDevices = new HashMap<>(); |
| Set<DeviceId> orphanedDevices = Sets.newHashSet(); |
| int deviceCount = 0; |
| |
| // Create buckets reflecting current ownership; do this irrespective of |
| // whether the node is active. |
| for (ControllerNode node : nodes) { |
| Set<DeviceId> devicesOf = new HashSet<>(getDevicesOf(node.id())); |
| if (clusterService.getState(node.id()).isActive()) { |
| log.info("Node {} has {} devices.", node.id(), devicesOf.size()); |
| deviceCount += devicesOf.size(); |
| controllerDevices.put(node, devicesOf); |
| } else if (!devicesOf.isEmpty()) { |
| log.warn("Inactive node {} has {} orphaned devices.", node.id(), devicesOf.size()); |
| orphanedDevices.addAll(getDevicesOf(node.id())); |
| } |
| } |
| |
| if (useRegionForBalanceRoles && balanceRolesUsingRegions(controllerDevices)) { |
| return; |
| } |
| |
| List<CompletableFuture<Void>> balanceBucketsFutures = Lists.newLinkedList(); |
| |
| // First re-balance the buckets until they are roughly even. |
| balanceControllerNodes(controllerDevices, deviceCount, balanceBucketsFutures); |
| |
| // Then attempt to distribute any orphaned devices among the buckets. |
| distributeOrphanedDevices(controllerDevices, orphanedDevices, balanceBucketsFutures); |
| |
| CompletableFuture<Void> balanceRolesFuture = |
| allOf(balanceBucketsFutures.toArray(new CompletableFuture[balanceBucketsFutures.size()])); |
| |
| Futures.getUnchecked(balanceRolesFuture); |
| } |
| |
| /** |
| * Balances the nodes specified in controllerDevices. |
| * |
| * @param controllerDevices controller nodes to devices map |
| * @param deviceCount number of devices mastered by controller nodes |
| * @param futures list of setRole futures for "moved" devices |
| */ |
| private void balanceControllerNodes(Map<ControllerNode, Set<DeviceId>> controllerDevices, |
| int deviceCount, |
| List<CompletableFuture<Void>> futures) { |
| // Now re-balance the buckets until they are roughly even. |
| int rounds = controllerDevices.keySet().size(); |
| for (int i = 0; i < rounds; i++) { |
| // Iterate over the buckets and find the smallest and the largest. |
| ControllerNode smallest = findBucket(true, controllerDevices); |
| ControllerNode largest = findBucket(false, controllerDevices); |
| futures.add(balanceBuckets(smallest, largest, controllerDevices, deviceCount)); |
| } |
| } |
| |
| /** |
| * Uses the set of orphaned devices to even out the load among the controllers. |
| * |
| * @param controllerDevices controller nodes to devices map |
| * @param orphanedDevices set of orphaned devices without an active master |
| * @param futures list of completable future to track the progress of the balancing operation |
| */ |
| private void distributeOrphanedDevices(Map<ControllerNode, Set<DeviceId>> controllerDevices, |
| Set<DeviceId> orphanedDevices, |
| List<CompletableFuture<Void>> futures) { |
| // Now re-distribute the orphaned devices into buckets until they are roughly even. |
| while (!orphanedDevices.isEmpty()) { |
| // Iterate over the buckets and find the smallest bucket. |
| ControllerNode smallest = findBucket(true, controllerDevices); |
| changeMastership(smallest, controllerDevices.get(smallest), |
| orphanedDevices, 1, futures); |
| } |
| } |
| |
| /** |
| * Finds node with the minimum/maximum devices from a list of nodes. |
| * |
| * @param min true: minimum, false: maximum |
| * @param controllerDevices controller nodes to devices map |
| * @return controller node with minimum/maximum devices |
| */ |
| |
| private ControllerNode findBucket(boolean min, |
| Map<ControllerNode, Set<DeviceId>> controllerDevices) { |
| int xSize = min ? Integer.MAX_VALUE : -1; |
| ControllerNode xNode = null; |
| for (ControllerNode node : controllerDevices.keySet()) { |
| int size = controllerDevices.get(node).size(); |
| if ((min && size < xSize) || (!min && size > xSize)) { |
| xSize = size; |
| xNode = node; |
| } |
| } |
| return xNode; |
| } |
| |
| /** |
| * Balance the node buckets by moving devices from largest to smallest node. |
| * |
| * @param smallest node that is master of the smallest number of devices |
| * @param largest node that is master of the largest number of devices |
| * @param controllerDevices controller nodes to devices map |
| * @param deviceCount number of devices mastered by controller nodes |
| * @return list of setRole futures for "moved" devices |
| */ |
| private CompletableFuture<Void> balanceBuckets(ControllerNode smallest, ControllerNode largest, |
| Map<ControllerNode, Set<DeviceId>> controllerDevices, |
| int deviceCount) { |
| Collection<DeviceId> minBucket = controllerDevices.get(smallest); |
| Collection<DeviceId> maxBucket = controllerDevices.get(largest); |
| int bucketCount = controllerDevices.keySet().size(); |
| |
| int delta = (maxBucket.size() - minBucket.size()) / 2; |
| delta = Math.min(deviceCount / bucketCount, delta); |
| |
| List<CompletableFuture<Void>> setRoleFutures = Lists.newLinkedList(); |
| |
| if (delta > 0) { |
| log.info("Attempting to move {} nodes from {} to {}...", delta, |
| largest.id(), smallest.id()); |
| changeMastership(smallest, minBucket, maxBucket, delta, setRoleFutures); |
| } |
| |
| return allOf(setRoleFutures.toArray(new CompletableFuture[setRoleFutures.size()])); |
| } |
| |
| /** |
| * Changes mastership for the specified number of devices in the given source |
| * bucket to the specified node and ads those devices to the given target |
| * bucket. Also adds the futures for tracking the role reassignment progress. |
| * |
| * @param toNode target controller node |
| * @param toBucket target bucket |
| * @param fromBucket source bucket |
| * @param count number of devices |
| * @param futures futures for tracking operation progress |
| */ |
| private void changeMastership(ControllerNode toNode, Collection<DeviceId> toBucket, |
| Collection<DeviceId> fromBucket, int count, |
| List<CompletableFuture<Void>> futures) { |
| int i = 0; |
| Iterator<DeviceId> it = fromBucket.iterator(); |
| while (it.hasNext() && i < count) { |
| DeviceId deviceId = it.next(); |
| log.info("Setting {} as the master for {}", toNode.id(), deviceId); |
| futures.add(setRole(toNode.id(), deviceId, MASTER)); |
| toBucket.add(deviceId); |
| it.remove(); |
| i++; |
| } |
| } |
| |
| /** |
| * Balances the nodes considering Region information. |
| * |
| * @param allControllerDevices controller nodes to devices map |
| * @return true: nodes balanced; false: nodes not balanced |
| */ |
| private boolean balanceRolesUsingRegions(Map<ControllerNode, Set<DeviceId>> allControllerDevices) { |
| Set<Region> regions = regionService.getRegions(); |
| if (regions.isEmpty()) { |
| return false; // no balancing was done using regions. |
| } |
| |
| // Handle nodes belonging to regions |
| Set<ControllerNode> nodesInRegions = Sets.newHashSet(); |
| for (Region region : regions) { |
| Map<ControllerNode, Set<DeviceId>> activeRegionControllers = |
| balanceRolesInRegion(region, allControllerDevices); |
| nodesInRegions.addAll(activeRegionControllers.keySet()); |
| } |
| |
| // Handle nodes not belonging to any region |
| Set<ControllerNode> nodesNotInRegions = Sets.difference(allControllerDevices.keySet(), nodesInRegions); |
| if (!nodesNotInRegions.isEmpty()) { |
| int deviceCount = 0; |
| Map<ControllerNode, Set<DeviceId>> controllerDevicesNotInRegions = new HashMap<>(); |
| for (ControllerNode controllerNode : nodesNotInRegions) { |
| controllerDevicesNotInRegions.put(controllerNode, allControllerDevices.get(controllerNode)); |
| deviceCount += allControllerDevices.get(controllerNode).size(); |
| } |
| // Now re-balance the buckets until they are roughly even. |
| List<CompletableFuture<Void>> balanceBucketsFutures = Lists.newArrayList(); |
| balanceControllerNodes(controllerDevicesNotInRegions, deviceCount, balanceBucketsFutures); |
| |
| CompletableFuture<Void> balanceRolesFuture = allOf( |
| balanceBucketsFutures.toArray(new CompletableFuture[balanceBucketsFutures.size()])); |
| |
| Futures.getUnchecked(balanceRolesFuture); |
| } |
| return true; // balancing was done using regions. |
| } |
| |
| /** |
| * Balances the nodes in specified region. |
| * |
| * @param region region in which nodes are to be balanced |
| * @param allControllerDevices controller nodes to devices map |
| * @return controller nodes that were balanced |
| */ |
| private Map<ControllerNode, Set<DeviceId>> |
| balanceRolesInRegion(Region region, |
| Map<ControllerNode, Set<DeviceId>> allControllerDevices) { |
| |
| // Retrieve all devices associated with specified region |
| Set<DeviceId> devicesInRegion = regionService.getRegionDevices(region.id()); |
| log.info("Region {} has {} devices.", region.id(), devicesInRegion.size()); |
| if (devicesInRegion.isEmpty()) { |
| return new HashMap<>(); // no devices in this region, so nothing to balance. |
| } |
| |
| List<Set<NodeId>> mastersList = region.masters(); |
| log.info("Region {} has {} sets of masters.", region.id(), mastersList.size()); |
| if (mastersList.isEmpty()) { |
| // TODO handle devices that belong to a region, which has no masters defined |
| return new HashMap<>(); // for now just leave devices alone |
| } |
| |
| // Get the region's preferred set of masters |
| Set<DeviceId> devicesInMasters = Sets.newHashSet(); |
| Map<ControllerNode, Set<DeviceId>> regionalControllerDevices = |
| getRegionsPreferredMasters(region, devicesInMasters, allControllerDevices); |
| |
| // Now re-balance the buckets until they are roughly even. |
| List<CompletableFuture<Void>> balanceBucketsFutures = Lists.newArrayList(); |
| balanceControllerNodes(regionalControllerDevices, devicesInMasters.size(), balanceBucketsFutures); |
| |
| // Handle devices that are not currently mastered by the master node set |
| Set<DeviceId> devicesNotMasteredWithControllers = Sets.difference(devicesInRegion, devicesInMasters); |
| if (!devicesNotMasteredWithControllers.isEmpty()) { |
| // active controllers in master node set are already balanced, just |
| // assign device mastership in sequence |
| List<ControllerNode> sorted = new ArrayList<>(regionalControllerDevices.keySet()); |
| Collections.sort(sorted, Comparator.comparingInt(o -> (regionalControllerDevices.get(o)).size())); |
| int deviceIndex = 0; |
| for (DeviceId deviceId : devicesNotMasteredWithControllers) { |
| ControllerNode cnode = sorted.get(deviceIndex % sorted.size()); |
| balanceBucketsFutures.add(setRole(cnode.id(), deviceId, MASTER)); |
| regionalControllerDevices.get(cnode).add(deviceId); |
| deviceIndex++; |
| } |
| } |
| |
| CompletableFuture<Void> balanceRolesFuture = |
| allOf(balanceBucketsFutures.toArray(new CompletableFuture[balanceBucketsFutures.size()])); |
| |
| Futures.getUnchecked(balanceRolesFuture); |
| |
| // Update the map before returning |
| regionalControllerDevices.forEach((controllerNode, deviceIds) -> { |
| regionalControllerDevices.put(controllerNode, new HashSet<>(getDevicesOf(controllerNode.id()))); |
| }); |
| |
| return regionalControllerDevices; |
| } |
| |
| /** |
| * Get region's preferred set of master nodes - the first master node set that has at |
| * least one active node. |
| * |
| * @param region region for which preferred set of master nodes is requested |
| * @param devicesInMasters device set to track devices in preferred set of master nodes |
| * @param allControllerDevices controller nodes to devices map |
| * @return region's preferred master nodes (and devices that use them as masters) |
| */ |
| private Map<ControllerNode, Set<DeviceId>> |
| getRegionsPreferredMasters(Region region, |
| Set<DeviceId> devicesInMasters, |
| Map<ControllerNode, Set<DeviceId>> allControllerDevices) { |
| Map<ControllerNode, Set<DeviceId>> regionalControllerDevices = new HashMap<>(); |
| int listIndex = 0; |
| for (Set<NodeId> masterSet : region.masters()) { |
| log.info("Region {} masters set {} has {} nodes.", |
| region.id(), listIndex, masterSet.size()); |
| if (masterSet.isEmpty()) { // nothing on this level |
| listIndex++; |
| continue; |
| } |
| // Create buckets reflecting current ownership. |
| for (NodeId nodeId : masterSet) { |
| if (clusterService.getState(nodeId).isActive()) { |
| ControllerNode controllerNode = clusterService.getNode(nodeId); |
| Set<DeviceId> devicesOf = new HashSet<>(allControllerDevices.get(controllerNode)); |
| regionalControllerDevices.put(controllerNode, devicesOf); |
| devicesInMasters.addAll(devicesOf); |
| log.info("Active Node {} has {} devices.", nodeId, devicesOf.size()); |
| } |
| } |
| if (!regionalControllerDevices.isEmpty()) { |
| break; // now have a set of >0 active controllers |
| } |
| listIndex++; // keep on looking |
| } |
| return regionalControllerDevices; |
| } |
| |
| public class InternalDelegate implements MastershipStoreDelegate { |
| @Override |
| public void notify(MastershipEvent event) { |
| post(event); |
| } |
| } |
| |
| private class InternalUpgradeEventListener implements UpgradeEventListener { |
| @Override |
| public void event(UpgradeEvent event) { |
| if (rebalanceRolesOnUpgrade && |
| (event.type() == UpgradeEvent.Type.COMMITTED || event.type() == UpgradeEvent.Type.RESET)) { |
| balanceRoles(); |
| } |
| } |
| } |
| |
| } |