2 * Copyright (c) 2014 Cisco Systems, Inc. and others. All rights reserved.
4 * This program and the accompanying materials are made available under the
5 * terms of the Eclipse Public License v1.0 which accompanies this distribution,
6 * and is available at http://www.eclipse.org/legal/epl-v10.html
9 package org.opendaylight.controller.cluster.datastore.shardmanager;
11 import static akka.pattern.Patterns.ask;
13 import akka.actor.ActorRef;
14 import akka.actor.Address;
15 import akka.actor.Cancellable;
16 import akka.actor.OneForOneStrategy;
17 import akka.actor.PoisonPill;
18 import akka.actor.Status;
19 import akka.actor.SupervisorStrategy;
20 import akka.actor.SupervisorStrategy.Directive;
21 import akka.cluster.ClusterEvent;
22 import akka.cluster.ClusterEvent.MemberWeaklyUp;
23 import akka.cluster.Member;
24 import akka.dispatch.Futures;
25 import akka.dispatch.OnComplete;
26 import akka.japi.Function;
27 import akka.pattern.Patterns;
28 import akka.persistence.DeleteSnapshotsFailure;
29 import akka.persistence.DeleteSnapshotsSuccess;
30 import akka.persistence.RecoveryCompleted;
31 import akka.persistence.SaveSnapshotFailure;
32 import akka.persistence.SaveSnapshotSuccess;
33 import akka.persistence.SnapshotOffer;
34 import akka.persistence.SnapshotSelectionCriteria;
35 import akka.util.Timeout;
36 import com.google.common.annotations.VisibleForTesting;
37 import com.google.common.base.Preconditions;
38 import java.util.ArrayList;
39 import java.util.Collection;
40 import java.util.Collections;
41 import java.util.HashMap;
42 import java.util.HashSet;
43 import java.util.List;
45 import java.util.Map.Entry;
47 import java.util.concurrent.CountDownLatch;
48 import java.util.concurrent.TimeUnit;
49 import java.util.concurrent.TimeoutException;
50 import java.util.function.Consumer;
51 import java.util.function.Supplier;
52 import org.opendaylight.controller.cluster.access.concepts.MemberName;
53 import org.opendaylight.controller.cluster.common.actor.AbstractUntypedPersistentActorWithMetering;
54 import org.opendaylight.controller.cluster.datastore.AbstractDataStore;
55 import org.opendaylight.controller.cluster.datastore.ClusterWrapper;
56 import org.opendaylight.controller.cluster.datastore.DatastoreContext;
57 import org.opendaylight.controller.cluster.datastore.DatastoreContext.Builder;
58 import org.opendaylight.controller.cluster.datastore.DatastoreContextFactory;
59 import org.opendaylight.controller.cluster.datastore.Shard;
60 import org.opendaylight.controller.cluster.datastore.config.Configuration;
61 import org.opendaylight.controller.cluster.datastore.config.ModuleShardConfiguration;
62 import org.opendaylight.controller.cluster.datastore.config.PrefixShardConfiguration;
63 import org.opendaylight.controller.cluster.datastore.exceptions.AlreadyExistsException;
64 import org.opendaylight.controller.cluster.datastore.exceptions.NoShardLeaderException;
65 import org.opendaylight.controller.cluster.datastore.exceptions.NotInitializedException;
66 import org.opendaylight.controller.cluster.datastore.exceptions.PrimaryNotFoundException;
67 import org.opendaylight.controller.cluster.datastore.identifiers.ShardIdentifier;
68 import org.opendaylight.controller.cluster.datastore.messages.ActorInitialized;
69 import org.opendaylight.controller.cluster.datastore.messages.AddPrefixShardReplica;
70 import org.opendaylight.controller.cluster.datastore.messages.AddShardReplica;
71 import org.opendaylight.controller.cluster.datastore.messages.ChangeShardMembersVotingStatus;
72 import org.opendaylight.controller.cluster.datastore.messages.CreateShard;
73 import org.opendaylight.controller.cluster.datastore.messages.FindLocalShard;
74 import org.opendaylight.controller.cluster.datastore.messages.FindPrimary;
75 import org.opendaylight.controller.cluster.datastore.messages.FlipShardMembersVotingStatus;
76 import org.opendaylight.controller.cluster.datastore.messages.GetShardRole;
77 import org.opendaylight.controller.cluster.datastore.messages.GetShardRoleReply;
78 import org.opendaylight.controller.cluster.datastore.messages.LocalPrimaryShardFound;
79 import org.opendaylight.controller.cluster.datastore.messages.LocalShardFound;
80 import org.opendaylight.controller.cluster.datastore.messages.LocalShardNotFound;
81 import org.opendaylight.controller.cluster.datastore.messages.RemoteFindPrimary;
82 import org.opendaylight.controller.cluster.datastore.messages.RemotePrimaryShardFound;
83 import org.opendaylight.controller.cluster.datastore.messages.RemovePrefixShardReplica;
84 import org.opendaylight.controller.cluster.datastore.messages.RemoveShardReplica;
85 import org.opendaylight.controller.cluster.datastore.messages.ShardLeaderStateChanged;
86 import org.opendaylight.controller.cluster.datastore.messages.UpdateSchemaContext;
87 import org.opendaylight.controller.cluster.datastore.persisted.DatastoreSnapshot;
88 import org.opendaylight.controller.cluster.datastore.persisted.ShardManagerSnapshot;
89 import org.opendaylight.controller.cluster.datastore.utils.ClusterUtils;
90 import org.opendaylight.controller.cluster.datastore.utils.CompositeOnComplete;
91 import org.opendaylight.controller.cluster.datastore.utils.Dispatchers;
92 import org.opendaylight.controller.cluster.datastore.utils.PrimaryShardInfoFutureCache;
93 import org.opendaylight.controller.cluster.notifications.RegisterRoleChangeListener;
94 import org.opendaylight.controller.cluster.notifications.RegisterRoleChangeListenerReply;
95 import org.opendaylight.controller.cluster.notifications.RoleChangeNotification;
96 import org.opendaylight.controller.cluster.raft.base.messages.FollowerInitialSyncUpStatus;
97 import org.opendaylight.controller.cluster.raft.base.messages.SwitchBehavior;
98 import org.opendaylight.controller.cluster.raft.client.messages.GetOnDemandRaftState;
99 import org.opendaylight.controller.cluster.raft.client.messages.GetSnapshot;
100 import org.opendaylight.controller.cluster.raft.client.messages.OnDemandRaftState;
101 import org.opendaylight.controller.cluster.raft.client.messages.Shutdown;
102 import org.opendaylight.controller.cluster.raft.messages.AddServer;
103 import org.opendaylight.controller.cluster.raft.messages.AddServerReply;
104 import org.opendaylight.controller.cluster.raft.messages.ChangeServersVotingStatus;
105 import org.opendaylight.controller.cluster.raft.messages.RemoveServer;
106 import org.opendaylight.controller.cluster.raft.messages.RemoveServerReply;
107 import org.opendaylight.controller.cluster.raft.messages.ServerChangeReply;
108 import org.opendaylight.controller.cluster.raft.messages.ServerChangeStatus;
109 import org.opendaylight.controller.cluster.raft.messages.ServerRemoved;
110 import org.opendaylight.controller.cluster.raft.policy.DisableElectionsRaftPolicy;
111 import org.opendaylight.controller.cluster.sharding.PrefixedShardConfigUpdateHandler;
112 import org.opendaylight.controller.cluster.sharding.messages.InitConfigListener;
113 import org.opendaylight.controller.cluster.sharding.messages.PrefixShardCreated;
114 import org.opendaylight.controller.cluster.sharding.messages.PrefixShardRemoved;
115 import org.opendaylight.controller.md.sal.common.api.data.LogicalDatastoreType;
116 import org.opendaylight.controller.md.sal.dom.api.DOMDataTreeChangeListener;
117 import org.opendaylight.mdsal.dom.api.DOMDataTreeIdentifier;
118 import org.opendaylight.yangtools.concepts.ListenerRegistration;
119 import org.opendaylight.yangtools.yang.data.api.YangInstanceIdentifier;
120 import org.opendaylight.yangtools.yang.model.api.SchemaContext;
121 import org.slf4j.Logger;
122 import org.slf4j.LoggerFactory;
123 import scala.concurrent.ExecutionContext;
124 import scala.concurrent.Future;
125 import scala.concurrent.duration.Duration;
126 import scala.concurrent.duration.FiniteDuration;
129 * Manages the shards for a data store. The ShardManager has the following jobs:
131 * <li> Create all the local shard replicas that belong on this cluster member
132 * <li> Find the address of the local shard
133 * <li> Find the primary replica for any given shard
134 * <li> Monitor the cluster members and store their addresses
137 class ShardManager extends AbstractUntypedPersistentActorWithMetering {
138 private static final Logger LOG = LoggerFactory.getLogger(ShardManager.class);
140 // Stores a mapping between a shard name and it's corresponding information
141 // Shard names look like inventory, topology etc and are as specified in
143 private final Map<String, ShardInformation> localShards = new HashMap<>();
145 // The type of a ShardManager reflects the type of the datastore itself
146 // A data store could be of type config/operational
147 private final String type;
149 private final ClusterWrapper cluster;
151 private final Configuration configuration;
153 private final String shardDispatcherPath;
155 private final ShardManagerInfo shardManagerMBean;
157 private DatastoreContextFactory datastoreContextFactory;
159 private final CountDownLatch waitTillReadyCountdownLatch;
161 private final PrimaryShardInfoFutureCache primaryShardInfoCache;
163 private final ShardPeerAddressResolver peerAddressResolver;
165 private SchemaContext schemaContext;
167 private DatastoreSnapshot restoreFromSnapshot;
169 private ShardManagerSnapshot currentSnapshot;
171 private final Set<String> shardReplicaOperationsInProgress = new HashSet<>();
173 private final Map<String, CompositeOnComplete<Boolean>> shardActorsStopping = new HashMap<>();
175 private final String persistenceId;
176 private final AbstractDataStore dataStore;
178 private ListenerRegistration<DOMDataTreeChangeListener> configListenerReg = null;
179 private PrefixedShardConfigUpdateHandler configUpdateHandler;
181 ShardManager(AbstractShardManagerCreator<?> builder) {
182 this.cluster = builder.getCluster();
183 this.configuration = builder.getConfiguration();
184 this.datastoreContextFactory = builder.getDatastoreContextFactory();
185 this.type = datastoreContextFactory.getBaseDatastoreContext().getDataStoreName();
186 this.shardDispatcherPath =
187 new Dispatchers(context().system().dispatchers()).getDispatcherPath(Dispatchers.DispatcherType.Shard);
188 this.waitTillReadyCountdownLatch = builder.getWaitTillReadyCountDownLatch();
189 this.primaryShardInfoCache = builder.getPrimaryShardInfoCache();
190 this.restoreFromSnapshot = builder.getRestoreFromSnapshot();
192 String possiblePersistenceId = datastoreContextFactory.getBaseDatastoreContext().getShardManagerPersistenceId();
193 persistenceId = possiblePersistenceId != null ? possiblePersistenceId : "shard-manager-" + type;
195 peerAddressResolver = new ShardPeerAddressResolver(type, cluster.getCurrentMemberName());
197 // Subscribe this actor to cluster member events
198 cluster.subscribeToMemberEvents(getSelf());
200 shardManagerMBean = new ShardManagerInfo(getSelf(), cluster.getCurrentMemberName(),
201 "shard-manager-" + this.type,
202 datastoreContextFactory.getBaseDatastoreContext().getDataStoreMXBeanType());
203 shardManagerMBean.registerMBean();
205 dataStore = builder.getDistributedDataStore();
209 public void preStart() {
210 LOG.info("Starting ShardManager {}", persistenceId);
214 public void postStop() {
215 LOG.info("Stopping ShardManager {}", persistenceId());
217 shardManagerMBean.unregisterMBean();
219 if (configListenerReg != null) {
220 configListenerReg.close();
221 configListenerReg = null;
226 public void handleCommand(Object message) throws Exception {
227 if (message instanceof FindPrimary) {
228 findPrimary((FindPrimary)message);
229 } else if (message instanceof FindLocalShard) {
230 findLocalShard((FindLocalShard) message);
231 } else if (message instanceof UpdateSchemaContext) {
232 updateSchemaContext(message);
233 } else if (message instanceof ActorInitialized) {
234 onActorInitialized(message);
235 } else if (message instanceof ClusterEvent.MemberUp) {
236 memberUp((ClusterEvent.MemberUp) message);
237 } else if (message instanceof ClusterEvent.MemberWeaklyUp) {
238 memberWeaklyUp((ClusterEvent.MemberWeaklyUp) message);
239 } else if (message instanceof ClusterEvent.MemberExited) {
240 memberExited((ClusterEvent.MemberExited) message);
241 } else if (message instanceof ClusterEvent.MemberRemoved) {
242 memberRemoved((ClusterEvent.MemberRemoved) message);
243 } else if (message instanceof ClusterEvent.UnreachableMember) {
244 memberUnreachable((ClusterEvent.UnreachableMember) message);
245 } else if (message instanceof ClusterEvent.ReachableMember) {
246 memberReachable((ClusterEvent.ReachableMember) message);
247 } else if (message instanceof DatastoreContextFactory) {
248 onDatastoreContextFactory((DatastoreContextFactory) message);
249 } else if (message instanceof RoleChangeNotification) {
250 onRoleChangeNotification((RoleChangeNotification) message);
251 } else if (message instanceof FollowerInitialSyncUpStatus) {
252 onFollowerInitialSyncStatus((FollowerInitialSyncUpStatus) message);
253 } else if (message instanceof ShardNotInitializedTimeout) {
254 onShardNotInitializedTimeout((ShardNotInitializedTimeout) message);
255 } else if (message instanceof ShardLeaderStateChanged) {
256 onLeaderStateChanged((ShardLeaderStateChanged) message);
257 } else if (message instanceof SwitchShardBehavior) {
258 onSwitchShardBehavior((SwitchShardBehavior) message);
259 } else if (message instanceof CreateShard) {
260 onCreateShard((CreateShard)message);
261 } else if (message instanceof AddShardReplica) {
262 onAddShardReplica((AddShardReplica) message);
263 } else if (message instanceof AddPrefixShardReplica) {
264 onAddPrefixShardReplica((AddPrefixShardReplica) message);
265 } else if (message instanceof PrefixShardCreated) {
266 onPrefixShardCreated((PrefixShardCreated) message);
267 } else if (message instanceof PrefixShardRemoved) {
268 onPrefixShardRemoved((PrefixShardRemoved) message);
269 } else if (message instanceof InitConfigListener) {
270 onInitConfigListener();
271 } else if (message instanceof ForwardedAddServerReply) {
272 ForwardedAddServerReply msg = (ForwardedAddServerReply)message;
273 onAddServerReply(msg.shardInfo, msg.addServerReply, getSender(), msg.leaderPath,
274 msg.removeShardOnFailure);
275 } else if (message instanceof ForwardedAddServerFailure) {
276 ForwardedAddServerFailure msg = (ForwardedAddServerFailure)message;
277 onAddServerFailure(msg.shardName, msg.failureMessage, msg.failure, getSender(), msg.removeShardOnFailure);
278 } else if (message instanceof RemoveShardReplica) {
279 onRemoveShardReplica((RemoveShardReplica) message);
280 } else if (message instanceof RemovePrefixShardReplica) {
281 onRemovePrefixShardReplica((RemovePrefixShardReplica) message);
282 } else if (message instanceof WrappedShardResponse) {
283 onWrappedShardResponse((WrappedShardResponse) message);
284 } else if (message instanceof GetSnapshot) {
286 } else if (message instanceof ServerRemoved) {
287 onShardReplicaRemoved((ServerRemoved) message);
288 } else if (message instanceof ChangeShardMembersVotingStatus) {
289 onChangeShardServersVotingStatus((ChangeShardMembersVotingStatus) message);
290 } else if (message instanceof FlipShardMembersVotingStatus) {
291 onFlipShardMembersVotingStatus((FlipShardMembersVotingStatus) message);
292 } else if (message instanceof SaveSnapshotSuccess) {
293 onSaveSnapshotSuccess((SaveSnapshotSuccess) message);
294 } else if (message instanceof SaveSnapshotFailure) {
295 LOG.error("{}: SaveSnapshotFailure received for saving snapshot of shards", persistenceId(),
296 ((SaveSnapshotFailure) message).cause());
297 } else if (message instanceof Shutdown) {
299 } else if (message instanceof GetLocalShardIds) {
300 onGetLocalShardIds();
301 } else if (message instanceof GetShardRole) {
302 onGetShardRole((GetShardRole) message);
303 } else if (message instanceof RunnableMessage) {
304 ((RunnableMessage)message).run();
305 } else if (message instanceof DeleteSnapshotsFailure) {
306 LOG.warn("{}: Failed to delete prior snapshots", persistenceId(),
307 ((DeleteSnapshotsFailure) message).cause());
308 } else if (message instanceof DeleteSnapshotsSuccess) {
309 LOG.debug("{}: Successfully deleted prior snapshots", persistenceId(), message);
310 } else if (message instanceof RegisterRoleChangeListenerReply) {
311 LOG.trace("{}: Received RegisterRoleChangeListenerReply", persistenceId());
312 } else if (message instanceof ClusterEvent.MemberEvent) {
313 LOG.trace("{}: Received other ClusterEvent.MemberEvent: {}", persistenceId(), message);
315 unknownMessage(message);
319 private void onGetShardRole(final GetShardRole message) {
320 LOG.debug("{}: onGetShardRole for shard: {}", persistenceId(), message.getName());
322 final String name = message.getName();
324 final ShardInformation shardInformation = localShards.get(name);
326 if (shardInformation == null) {
327 LOG.info("{}: no shard information for {} found", persistenceId(), name);
328 getSender().tell(new Status.Failure(
329 new IllegalArgumentException("Shard with name " + name + " not present.")), ActorRef.noSender());
333 getSender().tell(new GetShardRoleReply(shardInformation.getRole()), ActorRef.noSender());
336 private void onInitConfigListener() {
337 LOG.debug("{}: Initializing config listener on {}", persistenceId(), cluster.getCurrentMemberName());
339 final org.opendaylight.mdsal.common.api.LogicalDatastoreType type =
340 org.opendaylight.mdsal.common.api.LogicalDatastoreType
341 .valueOf(datastoreContextFactory.getBaseDatastoreContext().getLogicalStoreType().name());
343 if (configUpdateHandler != null) {
344 configUpdateHandler.close();
347 configUpdateHandler = new PrefixedShardConfigUpdateHandler(self(), cluster.getCurrentMemberName());
348 configUpdateHandler.initListener(dataStore, type);
351 private void onShutDown() {
352 List<Future<Boolean>> stopFutures = new ArrayList<>(localShards.size());
353 for (ShardInformation info : localShards.values()) {
354 if (info.getActor() != null) {
355 LOG.debug("{}: Issuing gracefulStop to shard {}", persistenceId(), info.getShardId());
357 FiniteDuration duration = info.getDatastoreContext().getShardRaftConfig()
358 .getElectionTimeOutInterval().$times(2);
359 stopFutures.add(Patterns.gracefulStop(info.getActor(), duration, Shutdown.INSTANCE));
363 LOG.info("Shutting down ShardManager {} - waiting on {} shards", persistenceId(), stopFutures.size());
365 ExecutionContext dispatcher = new Dispatchers(context().system().dispatchers())
366 .getDispatcher(Dispatchers.DispatcherType.Client);
367 Future<Iterable<Boolean>> combinedFutures = Futures.sequence(stopFutures, dispatcher);
369 combinedFutures.onComplete(new OnComplete<Iterable<Boolean>>() {
371 public void onComplete(Throwable failure, Iterable<Boolean> results) {
372 LOG.debug("{}: All shards shutdown - sending PoisonPill to self", persistenceId());
374 self().tell(PoisonPill.getInstance(), self());
376 if (failure != null) {
377 LOG.warn("{}: An error occurred attempting to shut down the shards", persistenceId(), failure);
380 for (Boolean result : results) {
387 LOG.warn("{}: {} shards did not shut down gracefully", persistenceId(), nfailed);
394 private void onWrappedShardResponse(WrappedShardResponse message) {
395 if (message.getResponse() instanceof RemoveServerReply) {
396 onRemoveServerReply(getSender(), message.getShardId(), (RemoveServerReply) message.getResponse(),
397 message.getLeaderPath());
401 private void onRemoveServerReply(ActorRef originalSender, ShardIdentifier shardId, RemoveServerReply replyMsg,
403 shardReplicaOperationsInProgress.remove(shardId.getShardName());
405 LOG.debug("{}: Received {} for shard {}", persistenceId(), replyMsg, shardId.getShardName());
407 if (replyMsg.getStatus() == ServerChangeStatus.OK) {
408 LOG.debug("{}: Leader shard successfully removed the replica shard {}", persistenceId(),
409 shardId.getShardName());
410 originalSender.tell(new Status.Success(null), getSelf());
412 LOG.warn("{}: Leader failed to remove shard replica {} with status {}",
413 persistenceId(), shardId, replyMsg.getStatus());
415 Exception failure = getServerChangeException(RemoveServer.class, replyMsg.getStatus(), leaderPath, shardId);
416 originalSender.tell(new Status.Failure(failure), getSelf());
420 private void removePrefixShardReplica(final RemovePrefixShardReplica contextMessage, final String shardName,
421 final String primaryPath, final ActorRef sender) {
422 if (isShardReplicaOperationInProgress(shardName, sender)) {
426 shardReplicaOperationsInProgress.add(shardName);
428 final ShardIdentifier shardId = getShardIdentifier(contextMessage.getMemberName(), shardName);
430 final DatastoreContext datastoreContext = newShardDatastoreContextBuilder(shardName).build();
432 //inform ShardLeader to remove this shard as a replica by sending an RemoveServer message
433 LOG.debug("{}: Sending RemoveServer message to peer {} for shard {}", persistenceId(),
434 primaryPath, shardId);
436 Timeout removeServerTimeout = new Timeout(datastoreContext.getShardLeaderElectionTimeout().duration());
437 Future<Object> futureObj = ask(getContext().actorSelection(primaryPath),
438 new RemoveServer(shardId.toString()), removeServerTimeout);
440 futureObj.onComplete(new OnComplete<Object>() {
442 public void onComplete(Throwable failure, Object response) {
443 if (failure != null) {
444 shardReplicaOperationsInProgress.remove(shardName);
445 String msg = String.format("RemoveServer request to leader %s for shard %s failed",
446 primaryPath, shardName);
448 LOG.debug("{}: {}", persistenceId(), msg, failure);
451 sender.tell(new Status.Failure(new RuntimeException(msg, failure)), self());
454 self().tell(new WrappedShardResponse(shardId, response, primaryPath), sender);
457 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
460 private void removeShardReplica(RemoveShardReplica contextMessage, final String shardName, final String primaryPath,
461 final ActorRef sender) {
462 if (isShardReplicaOperationInProgress(shardName, sender)) {
466 shardReplicaOperationsInProgress.add(shardName);
468 final ShardIdentifier shardId = getShardIdentifier(contextMessage.getMemberName(), shardName);
470 final DatastoreContext datastoreContext = newShardDatastoreContextBuilder(shardName).build();
472 //inform ShardLeader to remove this shard as a replica by sending an RemoveServer message
473 LOG.debug("{}: Sending RemoveServer message to peer {} for shard {}", persistenceId(),
474 primaryPath, shardId);
476 Timeout removeServerTimeout = new Timeout(datastoreContext.getShardLeaderElectionTimeout().duration());
477 Future<Object> futureObj = ask(getContext().actorSelection(primaryPath),
478 new RemoveServer(shardId.toString()), removeServerTimeout);
480 futureObj.onComplete(new OnComplete<Object>() {
482 public void onComplete(Throwable failure, Object response) {
483 if (failure != null) {
484 shardReplicaOperationsInProgress.remove(shardName);
485 String msg = String.format("RemoveServer request to leader %s for shard %s failed",
486 primaryPath, shardName);
488 LOG.debug("{}: {}", persistenceId(), msg, failure);
491 sender.tell(new Status.Failure(new RuntimeException(msg, failure)), self());
494 self().tell(new WrappedShardResponse(shardId, response, primaryPath), sender);
497 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
500 private void onShardReplicaRemoved(ServerRemoved message) {
501 removeShard(new ShardIdentifier.Builder().fromShardIdString(message.getServerId()).build());
504 @SuppressWarnings("checkstyle:IllegalCatch")
505 private void removeShard(final ShardIdentifier shardId) {
506 final String shardName = shardId.getShardName();
507 final ShardInformation shardInformation = localShards.remove(shardName);
508 if (shardInformation == null) {
509 LOG.debug("{} : Shard replica {} is not present in list", persistenceId(), shardId.toString());
513 final ActorRef shardActor = shardInformation.getActor();
514 if (shardActor != null) {
515 long timeoutInMS = Math.max(shardInformation.getDatastoreContext().getShardRaftConfig()
516 .getElectionTimeOutInterval().$times(3).toMillis(), 10000);
518 LOG.debug("{} : Sending Shutdown to Shard actor {} with {} ms timeout", persistenceId(), shardActor,
521 final Future<Boolean> stopFuture = Patterns.gracefulStop(shardActor,
522 FiniteDuration.apply(timeoutInMS, TimeUnit.MILLISECONDS), Shutdown.INSTANCE);
524 final CompositeOnComplete<Boolean> onComplete = new CompositeOnComplete<Boolean>() {
526 public void onComplete(final Throwable failure, final Boolean result) {
527 if (failure == null) {
528 LOG.debug("{} : Successfully shut down Shard actor {}", persistenceId(), shardActor);
530 LOG.warn("{}: Failed to shut down Shard actor {}", persistenceId(), shardActor, failure);
533 self().tell((RunnableMessage) () -> {
534 shardActorsStopping.remove(shardName);
535 notifyOnCompleteTasks(failure, result);
536 }, ActorRef.noSender());
540 shardActorsStopping.put(shardName, onComplete);
541 stopFuture.onComplete(onComplete, new Dispatchers(context().system().dispatchers())
542 .getDispatcher(Dispatchers.DispatcherType.Client));
545 LOG.debug("{} : Local Shard replica for shard {} has been removed", persistenceId(), shardName);
549 private void onGetSnapshot() {
550 LOG.debug("{}: onGetSnapshot", persistenceId());
552 List<String> notInitialized = null;
553 for (ShardInformation shardInfo : localShards.values()) {
554 if (!shardInfo.isShardInitialized()) {
555 if (notInitialized == null) {
556 notInitialized = new ArrayList<>();
559 notInitialized.add(shardInfo.getShardName());
563 if (notInitialized != null) {
564 getSender().tell(new Status.Failure(new IllegalStateException(String.format(
565 "%d shard(s) %s are not initialized", notInitialized.size(), notInitialized))), getSelf());
569 ActorRef replyActor = getContext().actorOf(ShardManagerGetSnapshotReplyActor.props(
570 new ArrayList<>(localShards.keySet()), type, currentSnapshot , getSender(), persistenceId(),
571 datastoreContextFactory.getBaseDatastoreContext().getShardInitializationTimeout().duration()));
573 for (ShardInformation shardInfo: localShards.values()) {
574 shardInfo.getActor().tell(GetSnapshot.INSTANCE, replyActor);
578 @SuppressWarnings("checkstyle:IllegalCatch")
579 private void onCreateShard(CreateShard createShard) {
580 LOG.debug("{}: onCreateShard: {}", persistenceId(), createShard);
584 String shardName = createShard.getModuleShardConfig().getShardName();
585 if (localShards.containsKey(shardName)) {
586 LOG.debug("{}: Shard {} already exists", persistenceId(), shardName);
587 reply = new Status.Success(String.format("Shard with name %s already exists", shardName));
589 doCreateShard(createShard);
590 reply = new Status.Success(null);
592 } catch (Exception e) {
593 LOG.error("{}: onCreateShard failed", persistenceId(), e);
594 reply = new Status.Failure(e);
597 if (getSender() != null && !getContext().system().deadLetters().equals(getSender())) {
598 getSender().tell(reply, getSelf());
602 private void onPrefixShardCreated(final PrefixShardCreated message) {
603 LOG.debug("{}: onPrefixShardCreated: {}", persistenceId(), message);
605 final PrefixShardConfiguration config = message.getConfiguration();
606 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(),
607 ClusterUtils.getCleanShardName(config.getPrefix().getRootIdentifier()));
608 final String shardName = shardId.getShardName();
610 if (isPreviousShardActorStopInProgress(shardName, message)) {
614 if (localShards.containsKey(shardName)) {
615 LOG.debug("{}: Received create for an already existing shard {}", persistenceId(), shardName);
616 final PrefixShardConfiguration existing =
617 configuration.getAllPrefixShardConfigurations().get(config.getPrefix());
619 if (existing != null && existing.equals(config)) {
620 // we don't have to do nothing here
625 doCreatePrefixShard(config, shardId, shardName);
628 private boolean isPreviousShardActorStopInProgress(final String shardName, final Object messageToDefer) {
629 final CompositeOnComplete<Boolean> stopOnComplete = shardActorsStopping.get(shardName);
630 if (stopOnComplete == null) {
634 LOG.debug("{} : Stop is in progress for shard {} - adding OnComplete callback to defer {}", persistenceId(),
635 shardName, messageToDefer);
636 final ActorRef sender = getSender();
637 stopOnComplete.addOnComplete(new OnComplete<Boolean>() {
639 public void onComplete(Throwable failure, Boolean result) {
640 LOG.debug("{} : Stop complete for shard {} - re-queing {}", persistenceId(), shardName, messageToDefer);
641 self().tell(messageToDefer, sender);
648 private void doCreatePrefixShard(PrefixShardConfiguration config, ShardIdentifier shardId, String shardName) {
649 configuration.addPrefixShardConfiguration(config);
651 final Builder builder = newShardDatastoreContextBuilder(shardName);
652 builder.logicalStoreType(LogicalDatastoreType.valueOf(config.getPrefix().getDatastoreType().name()))
653 .storeRoot(config.getPrefix().getRootIdentifier());
654 DatastoreContext shardDatastoreContext = builder.build();
656 final Map<String, String> peerAddresses = getPeerAddresses(shardName);
657 final boolean isActiveMember = true;
659 LOG.debug("{} doCreatePrefixShard: shardId: {}, memberNames: {}, peerAddresses: {}, isActiveMember: {}",
660 persistenceId(), shardId, config.getShardMemberNames(), peerAddresses, isActiveMember);
662 final ShardInformation info = new ShardInformation(shardName, shardId, peerAddresses,
663 shardDatastoreContext, Shard.builder(), peerAddressResolver);
664 info.setActiveMember(isActiveMember);
665 localShards.put(info.getShardName(), info);
667 if (schemaContext != null) {
668 info.setActor(newShardActor(schemaContext, info));
672 private void onPrefixShardRemoved(final PrefixShardRemoved message) {
673 LOG.debug("{}: onPrefixShardRemoved : {}", persistenceId(), message);
675 final DOMDataTreeIdentifier prefix = message.getPrefix();
676 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(),
677 ClusterUtils.getCleanShardName(prefix.getRootIdentifier()));
679 configuration.removePrefixShardConfiguration(prefix);
680 removeShard(shardId);
683 private void doCreateShard(final CreateShard createShard) {
684 final ModuleShardConfiguration moduleShardConfig = createShard.getModuleShardConfig();
685 final String shardName = moduleShardConfig.getShardName();
687 configuration.addModuleShardConfiguration(moduleShardConfig);
689 DatastoreContext shardDatastoreContext = createShard.getDatastoreContext();
690 if (shardDatastoreContext == null) {
691 shardDatastoreContext = newShardDatastoreContext(shardName);
693 shardDatastoreContext = DatastoreContext.newBuilderFrom(shardDatastoreContext).shardPeerAddressResolver(
694 peerAddressResolver).build();
697 ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(), shardName);
699 boolean shardWasInRecoveredSnapshot = currentSnapshot != null
700 && currentSnapshot.getShardList().contains(shardName);
702 Map<String, String> peerAddresses;
703 boolean isActiveMember;
704 if (shardWasInRecoveredSnapshot || configuration.getMembersFromShardName(shardName)
705 .contains(cluster.getCurrentMemberName())) {
706 peerAddresses = getPeerAddresses(shardName);
707 isActiveMember = true;
709 // The local member is not in the static shard member configuration and the shard did not
710 // previously exist (ie !shardWasInRecoveredSnapshot). In this case we'll create
711 // the shard with no peers and with elections disabled so it stays as follower. A
712 // subsequent AddServer request will be needed to make it an active member.
713 isActiveMember = false;
714 peerAddresses = Collections.emptyMap();
715 shardDatastoreContext = DatastoreContext.newBuilderFrom(shardDatastoreContext)
716 .customRaftPolicyImplementation(DisableElectionsRaftPolicy.class.getName()).build();
719 LOG.debug("{} doCreateShard: shardId: {}, memberNames: {}, peerAddresses: {}, isActiveMember: {}",
720 persistenceId(), shardId, moduleShardConfig.getShardMemberNames(), peerAddresses,
723 ShardInformation info = new ShardInformation(shardName, shardId, peerAddresses,
724 shardDatastoreContext, createShard.getShardBuilder(), peerAddressResolver);
725 info.setActiveMember(isActiveMember);
726 localShards.put(info.getShardName(), info);
728 if (schemaContext != null) {
729 info.setActor(newShardActor(schemaContext, info));
733 private DatastoreContext.Builder newShardDatastoreContextBuilder(String shardName) {
734 return DatastoreContext.newBuilderFrom(datastoreContextFactory.getShardDatastoreContext(shardName))
735 .shardPeerAddressResolver(peerAddressResolver);
738 private DatastoreContext newShardDatastoreContext(String shardName) {
739 return newShardDatastoreContextBuilder(shardName).build();
742 private void checkReady() {
743 if (isReadyWithLeaderId()) {
744 LOG.info("{}: All Shards are ready - data store {} is ready, available count is {}",
745 persistenceId(), type, waitTillReadyCountdownLatch.getCount());
747 waitTillReadyCountdownLatch.countDown();
751 private void onLeaderStateChanged(ShardLeaderStateChanged leaderStateChanged) {
752 LOG.info("{}: Received LeaderStateChanged message: {}", persistenceId(), leaderStateChanged);
754 ShardInformation shardInformation = findShardInformation(leaderStateChanged.getMemberId());
755 if (shardInformation != null) {
756 shardInformation.setLocalDataTree(leaderStateChanged.getLocalShardDataTree());
757 shardInformation.setLeaderVersion(leaderStateChanged.getLeaderPayloadVersion());
758 if (shardInformation.setLeaderId(leaderStateChanged.getLeaderId())) {
759 primaryShardInfoCache.remove(shardInformation.getShardName());
764 LOG.debug("No shard found with member Id {}", leaderStateChanged.getMemberId());
768 private void onShardNotInitializedTimeout(ShardNotInitializedTimeout message) {
769 ShardInformation shardInfo = message.getShardInfo();
771 LOG.debug("{}: Received ShardNotInitializedTimeout message for shard {}", persistenceId(),
772 shardInfo.getShardName());
774 shardInfo.removeOnShardInitialized(message.getOnShardInitialized());
776 if (!shardInfo.isShardInitialized()) {
777 LOG.debug("{}: Returning NotInitializedException for shard {}", persistenceId(), shardInfo.getShardName());
778 message.getSender().tell(createNotInitializedException(shardInfo.getShardId()), getSelf());
780 LOG.debug("{}: Returning NoShardLeaderException for shard {}", persistenceId(), shardInfo.getShardName());
781 message.getSender().tell(createNoShardLeaderException(shardInfo.getShardId()), getSelf());
785 private void onFollowerInitialSyncStatus(FollowerInitialSyncUpStatus status) {
786 LOG.info("{} Received follower initial sync status for {} status sync done {}", persistenceId(),
787 status.getName(), status.isInitialSyncDone());
789 ShardInformation shardInformation = findShardInformation(status.getName());
791 if (shardInformation != null) {
792 shardInformation.setFollowerSyncStatus(status.isInitialSyncDone());
794 shardManagerMBean.setSyncStatus(isInSync());
799 private void onRoleChangeNotification(RoleChangeNotification roleChanged) {
800 LOG.info("{}: Received role changed for {} from {} to {}", persistenceId(), roleChanged.getMemberId(),
801 roleChanged.getOldRole(), roleChanged.getNewRole());
803 ShardInformation shardInformation = findShardInformation(roleChanged.getMemberId());
804 if (shardInformation != null) {
805 shardInformation.setRole(roleChanged.getNewRole());
807 shardManagerMBean.setSyncStatus(isInSync());
812 private ShardInformation findShardInformation(String memberId) {
813 for (ShardInformation info : localShards.values()) {
814 if (info.getShardId().toString().equals(memberId)) {
822 private boolean isReadyWithLeaderId() {
823 boolean isReady = true;
824 for (ShardInformation info : localShards.values()) {
825 if (!info.isShardReadyWithLeaderId()) {
833 private boolean isInSync() {
834 for (ShardInformation info : localShards.values()) {
835 if (!info.isInSync()) {
842 private void onActorInitialized(Object message) {
843 final ActorRef sender = getSender();
845 if (sender == null) {
846 return; //why is a non-actor sending this message? Just ignore.
849 String actorName = sender.path().name();
850 //find shard name from actor name; actor name is stringified shardId
852 final ShardIdentifier shardId;
854 shardId = ShardIdentifier.fromShardIdString(actorName);
855 } catch (IllegalArgumentException e) {
856 LOG.debug("{}: ignoring actor {}", actorName, e);
860 markShardAsInitialized(shardId.getShardName());
863 private void markShardAsInitialized(String shardName) {
864 LOG.debug("{}: Initializing shard [{}]", persistenceId(), shardName);
866 ShardInformation shardInformation = localShards.get(shardName);
867 if (shardInformation != null) {
868 shardInformation.setActorInitialized();
870 shardInformation.getActor().tell(new RegisterRoleChangeListener(), self());
875 protected void handleRecover(Object message) throws Exception {
876 if (message instanceof RecoveryCompleted) {
877 onRecoveryCompleted();
878 } else if (message instanceof SnapshotOffer) {
879 applyShardManagerSnapshot((ShardManagerSnapshot)((SnapshotOffer) message).snapshot());
883 @SuppressWarnings("checkstyle:IllegalCatch")
884 private void onRecoveryCompleted() {
885 LOG.info("Recovery complete : {}", persistenceId());
887 if (currentSnapshot == null && restoreFromSnapshot != null
888 && restoreFromSnapshot.getShardManagerSnapshot() != null) {
889 ShardManagerSnapshot snapshot = restoreFromSnapshot.getShardManagerSnapshot();
891 LOG.debug("{}: Restoring from ShardManagerSnapshot: {}", persistenceId(), snapshot);
893 applyShardManagerSnapshot(snapshot);
899 private void sendResponse(ShardInformation shardInformation, boolean doWait,
900 boolean wantShardReady, final Supplier<Object> messageSupplier) {
901 if (!shardInformation.isShardInitialized() || wantShardReady && !shardInformation.isShardReadyWithLeaderId()) {
903 final ActorRef sender = getSender();
904 final ActorRef self = self();
906 Runnable replyRunnable = () -> sender.tell(messageSupplier.get(), self);
908 OnShardInitialized onShardInitialized = wantShardReady ? new OnShardReady(replyRunnable) :
909 new OnShardInitialized(replyRunnable);
911 shardInformation.addOnShardInitialized(onShardInitialized);
913 FiniteDuration timeout = shardInformation.getDatastoreContext()
914 .getShardInitializationTimeout().duration();
915 if (shardInformation.isShardInitialized()) {
916 // If the shard is already initialized then we'll wait enough time for the shard to
917 // elect a leader, ie 2 times the election timeout.
918 timeout = FiniteDuration.create(shardInformation.getDatastoreContext().getShardRaftConfig()
919 .getElectionTimeOutInterval().toMillis() * 2, TimeUnit.MILLISECONDS);
922 LOG.debug("{}: Scheduling {} ms timer to wait for shard {}", persistenceId(), timeout.toMillis(),
923 shardInformation.getShardName());
925 Cancellable timeoutSchedule = getContext().system().scheduler().scheduleOnce(
927 new ShardNotInitializedTimeout(shardInformation, onShardInitialized, sender),
928 getContext().dispatcher(), getSelf());
930 onShardInitialized.setTimeoutSchedule(timeoutSchedule);
932 } else if (!shardInformation.isShardInitialized()) {
933 LOG.debug("{}: Returning NotInitializedException for shard {}", persistenceId(),
934 shardInformation.getShardName());
935 getSender().tell(createNotInitializedException(shardInformation.getShardId()), getSelf());
937 LOG.debug("{}: Returning NoShardLeaderException for shard {}", persistenceId(),
938 shardInformation.getShardName());
939 getSender().tell(createNoShardLeaderException(shardInformation.getShardId()), getSelf());
945 getSender().tell(messageSupplier.get(), getSelf());
948 private static NoShardLeaderException createNoShardLeaderException(ShardIdentifier shardId) {
949 return new NoShardLeaderException(null, shardId.toString());
952 private static NotInitializedException createNotInitializedException(ShardIdentifier shardId) {
953 return new NotInitializedException(String.format(
954 "Found primary shard %s but it's not initialized yet. Please try again later", shardId));
958 static MemberName memberToName(final Member member) {
959 return MemberName.forName(member.roles().iterator().next());
962 private void memberRemoved(ClusterEvent.MemberRemoved message) {
963 MemberName memberName = memberToName(message.member());
965 LOG.info("{}: Received MemberRemoved: memberName: {}, address: {}", persistenceId(), memberName,
966 message.member().address());
968 peerAddressResolver.removePeerAddress(memberName);
970 for (ShardInformation info : localShards.values()) {
971 info.peerDown(memberName, getShardIdentifier(memberName, info.getShardName()).toString(), getSelf());
975 private void memberExited(ClusterEvent.MemberExited message) {
976 MemberName memberName = memberToName(message.member());
978 LOG.info("{}: Received MemberExited: memberName: {}, address: {}", persistenceId(), memberName,
979 message.member().address());
981 peerAddressResolver.removePeerAddress(memberName);
983 for (ShardInformation info : localShards.values()) {
984 info.peerDown(memberName, getShardIdentifier(memberName, info.getShardName()).toString(), getSelf());
988 private void memberUp(ClusterEvent.MemberUp message) {
989 MemberName memberName = memberToName(message.member());
991 LOG.info("{}: Received MemberUp: memberName: {}, address: {}", persistenceId(), memberName,
992 message.member().address());
994 memberUp(memberName, message.member().address());
997 private void memberUp(MemberName memberName, Address address) {
998 addPeerAddress(memberName, address);
1002 private void memberWeaklyUp(MemberWeaklyUp message) {
1003 MemberName memberName = memberToName(message.member());
1005 LOG.info("{}: Received MemberWeaklyUp: memberName: {}, address: {}", persistenceId(), memberName,
1006 message.member().address());
1008 memberUp(memberName, message.member().address());
1011 private void addPeerAddress(MemberName memberName, Address address) {
1012 peerAddressResolver.addPeerAddress(memberName, address);
1014 for (ShardInformation info : localShards.values()) {
1015 String shardName = info.getShardName();
1016 String peerId = getShardIdentifier(memberName, shardName).toString();
1017 info.updatePeerAddress(peerId, peerAddressResolver.getShardActorAddress(shardName, memberName), getSelf());
1019 info.peerUp(memberName, peerId, getSelf());
1023 private void memberReachable(ClusterEvent.ReachableMember message) {
1024 MemberName memberName = memberToName(message.member());
1025 LOG.info("Received ReachableMember: memberName {}, address: {}", memberName, message.member().address());
1027 addPeerAddress(memberName, message.member().address());
1029 markMemberAvailable(memberName);
1032 private void memberUnreachable(ClusterEvent.UnreachableMember message) {
1033 MemberName memberName = memberToName(message.member());
1034 LOG.info("Received UnreachableMember: memberName {}, address: {}", memberName, message.member().address());
1036 markMemberUnavailable(memberName);
1039 private void markMemberUnavailable(final MemberName memberName) {
1040 final String memberStr = memberName.getName();
1041 for (ShardInformation info : localShards.values()) {
1042 String leaderId = info.getLeaderId();
1043 // XXX: why are we using String#contains() here?
1044 if (leaderId != null && leaderId.contains(memberStr)) {
1045 LOG.debug("Marking Leader {} as unavailable.", leaderId);
1046 info.setLeaderAvailable(false);
1048 primaryShardInfoCache.remove(info.getShardName());
1051 info.peerDown(memberName, getShardIdentifier(memberName, info.getShardName()).toString(), getSelf());
1055 private void markMemberAvailable(final MemberName memberName) {
1056 final String memberStr = memberName.getName();
1057 for (ShardInformation info : localShards.values()) {
1058 String leaderId = info.getLeaderId();
1059 // XXX: why are we using String#contains() here?
1060 if (leaderId != null && leaderId.contains(memberStr)) {
1061 LOG.debug("Marking Leader {} as available.", leaderId);
1062 info.setLeaderAvailable(true);
1065 info.peerUp(memberName, getShardIdentifier(memberName, info.getShardName()).toString(), getSelf());
1069 private void onDatastoreContextFactory(DatastoreContextFactory factory) {
1070 datastoreContextFactory = factory;
1071 for (ShardInformation info : localShards.values()) {
1072 info.setDatastoreContext(newShardDatastoreContext(info.getShardName()), getSelf());
1076 private void onGetLocalShardIds() {
1077 final List<String> response = new ArrayList<>(localShards.size());
1079 for (ShardInformation info : localShards.values()) {
1080 response.add(info.getShardId().toString());
1083 getSender().tell(new Status.Success(response), getSelf());
1086 private void onSwitchShardBehavior(final SwitchShardBehavior message) {
1087 final ShardIdentifier identifier = message.getShardId();
1089 if (identifier != null) {
1090 final ShardInformation info = localShards.get(identifier.getShardName());
1092 getSender().tell(new Status.Failure(
1093 new IllegalArgumentException("Shard " + identifier + " is not local")), getSelf());
1097 switchShardBehavior(info, new SwitchBehavior(message.getNewState(), message.getTerm()));
1099 for (ShardInformation info : localShards.values()) {
1100 switchShardBehavior(info, new SwitchBehavior(message.getNewState(), message.getTerm()));
1104 getSender().tell(new Status.Success(null), getSelf());
1107 private void switchShardBehavior(final ShardInformation info, final SwitchBehavior switchBehavior) {
1108 final ActorRef actor = info.getActor();
1109 if (actor != null) {
1110 actor.tell(switchBehavior, getSelf());
1112 LOG.warn("Could not switch the behavior of shard {} to {} - shard is not yet available",
1113 info.getShardName(), switchBehavior.getNewState());
1118 * Notifies all the local shards of a change in the schema context.
1120 * @param message the message to send
1122 private void updateSchemaContext(final Object message) {
1123 schemaContext = ((UpdateSchemaContext) message).getSchemaContext();
1125 LOG.debug("Got updated SchemaContext: # of modules {}", schemaContext.getAllModuleIdentifiers().size());
1127 for (ShardInformation info : localShards.values()) {
1128 if (info.getActor() == null) {
1129 LOG.debug("Creating Shard {}", info.getShardId());
1130 info.setActor(newShardActor(schemaContext, info));
1132 info.getActor().tell(message, getSelf());
1138 protected ClusterWrapper getCluster() {
1143 protected ActorRef newShardActor(final SchemaContext shardSchemaContext, final ShardInformation info) {
1144 return getContext().actorOf(info.newProps(shardSchemaContext).withDispatcher(shardDispatcherPath),
1145 info.getShardId().toString());
1148 private void findPrimary(FindPrimary message) {
1149 LOG.debug("{}: In findPrimary: {}", persistenceId(), message);
1151 final String shardName = message.getShardName();
1152 final boolean canReturnLocalShardState = !(message instanceof RemoteFindPrimary);
1154 // First see if the there is a local replica for the shard
1155 final ShardInformation info = localShards.get(shardName);
1156 if (info != null && info.isActiveMember()) {
1157 sendResponse(info, message.isWaitUntilReady(), true, () -> {
1158 String primaryPath = info.getSerializedLeaderActor();
1159 Object found = canReturnLocalShardState && info.isLeader()
1160 ? new LocalPrimaryShardFound(primaryPath, info.getLocalShardDataTree().get()) :
1161 new RemotePrimaryShardFound(primaryPath, info.getLeaderVersion());
1163 LOG.debug("{}: Found primary for {}: {}", persistenceId(), shardName, found);
1170 final Collection<String> visitedAddresses;
1171 if (message instanceof RemoteFindPrimary) {
1172 visitedAddresses = ((RemoteFindPrimary)message).getVisitedAddresses();
1174 visitedAddresses = new ArrayList<>(1);
1177 visitedAddresses.add(peerAddressResolver.getShardManagerActorPathBuilder(cluster.getSelfAddress()).toString());
1179 for (String address: peerAddressResolver.getShardManagerPeerActorAddresses()) {
1180 if (visitedAddresses.contains(address)) {
1184 LOG.debug("{}: findPrimary for {} forwarding to remote ShardManager {}, visitedAddresses: {}",
1185 persistenceId(), shardName, address, visitedAddresses);
1187 getContext().actorSelection(address).forward(new RemoteFindPrimary(shardName,
1188 message.isWaitUntilReady(), visitedAddresses), getContext());
1192 LOG.debug("{}: No shard found for {}", persistenceId(), shardName);
1194 getSender().tell(new PrimaryNotFoundException(
1195 String.format("No primary shard found for %s.", shardName)), getSelf());
1198 private void findPrimary(final String shardName, final FindPrimaryResponseHandler handler) {
1199 Timeout findPrimaryTimeout = new Timeout(datastoreContextFactory.getBaseDatastoreContext()
1200 .getShardInitializationTimeout().duration().$times(2));
1202 Future<Object> futureObj = ask(getSelf(), new FindPrimary(shardName, true), findPrimaryTimeout);
1203 futureObj.onComplete(new OnComplete<Object>() {
1205 public void onComplete(Throwable failure, Object response) {
1206 if (failure != null) {
1207 handler.onFailure(failure);
1209 if (response instanceof RemotePrimaryShardFound) {
1210 handler.onRemotePrimaryShardFound((RemotePrimaryShardFound) response);
1211 } else if (response instanceof LocalPrimaryShardFound) {
1212 handler.onLocalPrimaryFound((LocalPrimaryShardFound) response);
1214 handler.onUnknownResponse(response);
1218 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1222 * Construct the name of the shard actor given the name of the member on
1223 * which the shard resides and the name of the shard.
1225 * @param memberName the member name
1226 * @param shardName the shard name
1229 private ShardIdentifier getShardIdentifier(MemberName memberName, String shardName) {
1230 return peerAddressResolver.getShardIdentifier(memberName, shardName);
1234 * Create shards that are local to the member on which the ShardManager runs.
1236 private void createLocalShards() {
1237 MemberName memberName = this.cluster.getCurrentMemberName();
1238 Collection<String> memberShardNames = this.configuration.getMemberShardNames(memberName);
1240 Map<String, DatastoreSnapshot.ShardSnapshot> shardSnapshots = new HashMap<>();
1241 if (restoreFromSnapshot != null) {
1242 for (DatastoreSnapshot.ShardSnapshot snapshot: restoreFromSnapshot.getShardSnapshots()) {
1243 shardSnapshots.put(snapshot.getName(), snapshot);
1247 restoreFromSnapshot = null; // null out to GC
1249 for (String shardName : memberShardNames) {
1250 ShardIdentifier shardId = getShardIdentifier(memberName, shardName);
1252 LOG.debug("{}: Creating local shard: {}", persistenceId(), shardId);
1254 Map<String, String> peerAddresses = getPeerAddresses(shardName);
1255 localShards.put(shardName, new ShardInformation(shardName, shardId, peerAddresses,
1256 newShardDatastoreContext(shardName), Shard.builder().restoreFromSnapshot(
1257 shardSnapshots.get(shardName)), peerAddressResolver));
1262 * Given the name of the shard find the addresses of all it's peers.
1264 * @param shardName the shard name
1266 private Map<String, String> getPeerAddresses(String shardName) {
1267 final Collection<MemberName> members = configuration.getMembersFromShardName(shardName);
1268 return getPeerAddresses(shardName, members);
1271 private Map<String, String> getPeerAddresses(final String shardName, final Collection<MemberName> members) {
1272 Map<String, String> peerAddresses = new HashMap<>();
1273 MemberName currentMemberName = this.cluster.getCurrentMemberName();
1275 for (MemberName memberName : members) {
1276 if (!currentMemberName.equals(memberName)) {
1277 ShardIdentifier shardId = getShardIdentifier(memberName, shardName);
1278 String address = peerAddressResolver.getShardActorAddress(shardName, memberName);
1279 peerAddresses.put(shardId.toString(), address);
1282 return peerAddresses;
1286 public SupervisorStrategy supervisorStrategy() {
1288 return new OneForOneStrategy(10, Duration.create("1 minute"),
1289 (Function<Throwable, Directive>) t -> {
1290 LOG.warn("Supervisor Strategy caught unexpected exception - resuming", t);
1291 return SupervisorStrategy.resume();
1296 public String persistenceId() {
1297 return persistenceId;
1301 ShardManagerInfoMBean getMBean() {
1302 return shardManagerMBean;
1305 private boolean isShardReplicaOperationInProgress(final String shardName, final ActorRef sender) {
1306 if (shardReplicaOperationsInProgress.contains(shardName)) {
1307 String msg = String.format("A shard replica operation for %s is already in progress", shardName);
1308 LOG.debug("{}: {}", persistenceId(), msg);
1309 sender.tell(new Status.Failure(new IllegalStateException(msg)), getSelf());
1316 private void onAddPrefixShardReplica(final AddPrefixShardReplica message) {
1317 LOG.debug("{}: onAddPrefixShardReplica: {}", persistenceId(), message);
1319 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(),
1320 ClusterUtils.getCleanShardName(message.getShardPrefix()));
1321 final String shardName = shardId.getShardName();
1323 // Create the localShard
1324 if (schemaContext == null) {
1325 String msg = String.format(
1326 "No SchemaContext is available in order to create a local shard instance for %s", shardName);
1327 LOG.debug("{}: {}", persistenceId(), msg);
1328 getSender().tell(new Status.Failure(new IllegalStateException(msg)), getSelf());
1332 findPrimary(shardName, new AutoFindPrimaryFailureResponseHandler(getSender(), shardName, persistenceId(),
1335 public void onRemotePrimaryShardFound(RemotePrimaryShardFound response) {
1336 final RunnableMessage runnable = (RunnableMessage) () -> addPrefixShard(getShardName(),
1337 message.getShardPrefix(), response, getSender());
1338 if (!isPreviousShardActorStopInProgress(getShardName(), runnable)) {
1339 getSelf().tell(runnable, getTargetActor());
1344 public void onLocalPrimaryFound(LocalPrimaryShardFound message) {
1345 sendLocalReplicaAlreadyExistsReply(getShardName(), getTargetActor());
1350 private void onAddShardReplica(final AddShardReplica shardReplicaMsg) {
1351 final String shardName = shardReplicaMsg.getShardName();
1353 LOG.debug("{}: onAddShardReplica: {}", persistenceId(), shardReplicaMsg);
1355 // verify the shard with the specified name is present in the cluster configuration
1356 if (!this.configuration.isShardConfigured(shardName)) {
1357 String msg = String.format("No module configuration exists for shard %s", shardName);
1358 LOG.debug("{}: {}", persistenceId(), msg);
1359 getSender().tell(new Status.Failure(new IllegalArgumentException(msg)), getSelf());
1363 // Create the localShard
1364 if (schemaContext == null) {
1365 String msg = String.format(
1366 "No SchemaContext is available in order to create a local shard instance for %s", shardName);
1367 LOG.debug("{}: {}", persistenceId(), msg);
1368 getSender().tell(new Status.Failure(new IllegalStateException(msg)), getSelf());
1372 findPrimary(shardName, new AutoFindPrimaryFailureResponseHandler(getSender(), shardName, persistenceId(),
1375 public void onRemotePrimaryShardFound(RemotePrimaryShardFound response) {
1376 final RunnableMessage runnable = (RunnableMessage) () ->
1377 addShard(getShardName(), response, getSender());
1378 if (!isPreviousShardActorStopInProgress(getShardName(), runnable)) {
1379 getSelf().tell(runnable, getTargetActor());
1384 public void onLocalPrimaryFound(LocalPrimaryShardFound message) {
1385 sendLocalReplicaAlreadyExistsReply(getShardName(), getTargetActor());
1390 private void sendLocalReplicaAlreadyExistsReply(String shardName, ActorRef sender) {
1391 String msg = String.format("Local shard %s already exists", shardName);
1392 LOG.debug("{}: {}", persistenceId(), msg);
1393 sender.tell(new Status.Failure(new AlreadyExistsException(msg)), getSelf());
1396 private void addPrefixShard(final String shardName, final YangInstanceIdentifier shardPrefix,
1397 final RemotePrimaryShardFound response, final ActorRef sender) {
1398 if (isShardReplicaOperationInProgress(shardName, sender)) {
1402 shardReplicaOperationsInProgress.add(shardName);
1404 final ShardInformation shardInfo;
1405 final boolean removeShardOnFailure;
1406 ShardInformation existingShardInfo = localShards.get(shardName);
1407 if (existingShardInfo == null) {
1408 removeShardOnFailure = true;
1409 ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(), shardName);
1411 final Builder builder = newShardDatastoreContextBuilder(shardName);
1412 builder.storeRoot(shardPrefix).customRaftPolicyImplementation(DisableElectionsRaftPolicy.class.getName());
1414 DatastoreContext datastoreContext = builder.build();
1416 shardInfo = new ShardInformation(shardName, shardId, getPeerAddresses(shardName), datastoreContext,
1417 Shard.builder(), peerAddressResolver);
1418 shardInfo.setActiveMember(false);
1419 localShards.put(shardName, shardInfo);
1420 shardInfo.setActor(newShardActor(schemaContext, shardInfo));
1422 removeShardOnFailure = false;
1423 shardInfo = existingShardInfo;
1426 execAddShard(shardName, shardInfo, response, removeShardOnFailure, sender);
1429 private void addShard(final String shardName, final RemotePrimaryShardFound response, final ActorRef sender) {
1430 if (isShardReplicaOperationInProgress(shardName, sender)) {
1434 shardReplicaOperationsInProgress.add(shardName);
1436 final ShardInformation shardInfo;
1437 final boolean removeShardOnFailure;
1438 ShardInformation existingShardInfo = localShards.get(shardName);
1439 if (existingShardInfo == null) {
1440 removeShardOnFailure = true;
1441 ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(), shardName);
1443 DatastoreContext datastoreContext = newShardDatastoreContextBuilder(shardName)
1444 .customRaftPolicyImplementation(DisableElectionsRaftPolicy.class.getName()).build();
1446 shardInfo = new ShardInformation(shardName, shardId, getPeerAddresses(shardName), datastoreContext,
1447 Shard.builder(), peerAddressResolver);
1448 shardInfo.setActiveMember(false);
1449 localShards.put(shardName, shardInfo);
1450 shardInfo.setActor(newShardActor(schemaContext, shardInfo));
1452 removeShardOnFailure = false;
1453 shardInfo = existingShardInfo;
1456 execAddShard(shardName, shardInfo, response, removeShardOnFailure, sender);
1459 private void execAddShard(final String shardName,
1460 final ShardInformation shardInfo,
1461 final RemotePrimaryShardFound response,
1462 final boolean removeShardOnFailure,
1463 final ActorRef sender) {
1465 final String localShardAddress =
1466 peerAddressResolver.getShardActorAddress(shardName, cluster.getCurrentMemberName());
1468 //inform ShardLeader to add this shard as a replica by sending an AddServer message
1469 LOG.debug("{}: Sending AddServer message to peer {} for shard {}", persistenceId(),
1470 response.getPrimaryPath(), shardInfo.getShardId());
1472 final Timeout addServerTimeout = new Timeout(shardInfo.getDatastoreContext()
1473 .getShardLeaderElectionTimeout().duration());
1474 final Future<Object> futureObj = ask(getContext().actorSelection(response.getPrimaryPath()),
1475 new AddServer(shardInfo.getShardId().toString(), localShardAddress, true), addServerTimeout);
1477 futureObj.onComplete(new OnComplete<Object>() {
1479 public void onComplete(Throwable failure, Object addServerResponse) {
1480 if (failure != null) {
1481 LOG.debug("{}: AddServer request to {} for {} failed", persistenceId(),
1482 response.getPrimaryPath(), shardName, failure);
1484 final String msg = String.format("AddServer request to leader %s for shard %s failed",
1485 response.getPrimaryPath(), shardName);
1486 self().tell(new ForwardedAddServerFailure(shardName, msg, failure, removeShardOnFailure), sender);
1488 self().tell(new ForwardedAddServerReply(shardInfo, (AddServerReply)addServerResponse,
1489 response.getPrimaryPath(), removeShardOnFailure), sender);
1492 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1495 private void onAddServerFailure(String shardName, String message, Throwable failure, ActorRef sender,
1496 boolean removeShardOnFailure) {
1497 shardReplicaOperationsInProgress.remove(shardName);
1499 if (removeShardOnFailure) {
1500 ShardInformation shardInfo = localShards.remove(shardName);
1501 if (shardInfo.getActor() != null) {
1502 shardInfo.getActor().tell(PoisonPill.getInstance(), getSelf());
1506 sender.tell(new Status.Failure(message == null ? failure :
1507 new RuntimeException(message, failure)), getSelf());
1510 private void onAddServerReply(ShardInformation shardInfo, AddServerReply replyMsg, ActorRef sender,
1511 String leaderPath, boolean removeShardOnFailure) {
1512 String shardName = shardInfo.getShardName();
1513 shardReplicaOperationsInProgress.remove(shardName);
1515 LOG.debug("{}: Received {} for shard {} from leader {}", persistenceId(), replyMsg, shardName, leaderPath);
1517 if (replyMsg.getStatus() == ServerChangeStatus.OK) {
1518 LOG.debug("{}: Leader shard successfully added the replica shard {}", persistenceId(), shardName);
1520 // Make the local shard voting capable
1521 shardInfo.setDatastoreContext(newShardDatastoreContext(shardName), getSelf());
1522 shardInfo.setActiveMember(true);
1525 sender.tell(new Status.Success(null), getSelf());
1526 } else if (replyMsg.getStatus() == ServerChangeStatus.ALREADY_EXISTS) {
1527 sendLocalReplicaAlreadyExistsReply(shardName, sender);
1529 LOG.warn("{}: Leader failed to add shard replica {} with status {}",
1530 persistenceId(), shardName, replyMsg.getStatus());
1532 Exception failure = getServerChangeException(AddServer.class, replyMsg.getStatus(), leaderPath,
1533 shardInfo.getShardId());
1535 onAddServerFailure(shardName, null, failure, sender, removeShardOnFailure);
1539 private static Exception getServerChangeException(Class<?> serverChange, ServerChangeStatus serverChangeStatus,
1540 String leaderPath, ShardIdentifier shardId) {
1542 switch (serverChangeStatus) {
1544 failure = new TimeoutException(String.format(
1545 "The shard leader %s timed out trying to replicate the initial data to the new shard %s."
1546 + "Possible causes - there was a problem replicating the data or shard leadership changed "
1547 + "while replicating the shard data", leaderPath, shardId.getShardName()));
1550 failure = createNoShardLeaderException(shardId);
1553 failure = new UnsupportedOperationException(String.format("%s request is not supported for shard %s",
1554 serverChange.getSimpleName(), shardId.getShardName()));
1557 failure = new RuntimeException(String.format(
1558 "%s request to leader %s for shard %s failed with status %s",
1559 serverChange.getSimpleName(), leaderPath, shardId.getShardName(), serverChangeStatus));
1564 private void onRemoveShardReplica(final RemoveShardReplica shardReplicaMsg) {
1565 LOG.debug("{}: onRemoveShardReplica: {}", persistenceId(), shardReplicaMsg);
1567 findPrimary(shardReplicaMsg.getShardName(), new AutoFindPrimaryFailureResponseHandler(getSender(),
1568 shardReplicaMsg.getShardName(), persistenceId(), getSelf()) {
1570 public void onRemotePrimaryShardFound(RemotePrimaryShardFound response) {
1571 doRemoveShardReplicaAsync(response.getPrimaryPath());
1575 public void onLocalPrimaryFound(LocalPrimaryShardFound response) {
1576 doRemoveShardReplicaAsync(response.getPrimaryPath());
1579 private void doRemoveShardReplicaAsync(final String primaryPath) {
1580 getSelf().tell((RunnableMessage) () -> removeShardReplica(shardReplicaMsg, getShardName(),
1581 primaryPath, getSender()), getTargetActor());
1586 private void onRemovePrefixShardReplica(final RemovePrefixShardReplica message) {
1587 LOG.debug("{}: onRemovePrefixShardReplica: {}", persistenceId(), message);
1589 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(),
1590 ClusterUtils.getCleanShardName(message.getShardPrefix()));
1591 final String shardName = shardId.getShardName();
1593 findPrimary(shardName, new AutoFindPrimaryFailureResponseHandler(getSender(),
1594 shardName, persistenceId(), getSelf()) {
1596 public void onRemotePrimaryShardFound(RemotePrimaryShardFound response) {
1597 doRemoveShardReplicaAsync(response.getPrimaryPath());
1601 public void onLocalPrimaryFound(LocalPrimaryShardFound response) {
1602 doRemoveShardReplicaAsync(response.getPrimaryPath());
1605 private void doRemoveShardReplicaAsync(final String primaryPath) {
1606 getSelf().tell((RunnableMessage) () -> removePrefixShardReplica(message, getShardName(),
1607 primaryPath, getSender()), getTargetActor());
1612 private void persistShardList() {
1613 List<String> shardList = new ArrayList<>(localShards.keySet());
1614 for (ShardInformation shardInfo : localShards.values()) {
1615 if (!shardInfo.isActiveMember()) {
1616 shardList.remove(shardInfo.getShardName());
1619 LOG.debug("{}: persisting the shard list {}", persistenceId(), shardList);
1620 saveSnapshot(updateShardManagerSnapshot(shardList, configuration.getAllPrefixShardConfigurations()));
1623 private ShardManagerSnapshot updateShardManagerSnapshot(
1624 final List<String> shardList,
1625 final Map<DOMDataTreeIdentifier, PrefixShardConfiguration> allPrefixShardConfigurations) {
1626 currentSnapshot = new ShardManagerSnapshot(shardList, allPrefixShardConfigurations);
1627 return currentSnapshot;
1630 private void applyShardManagerSnapshot(ShardManagerSnapshot snapshot) {
1631 currentSnapshot = snapshot;
1633 LOG.debug("{}: onSnapshotOffer: {}", persistenceId(), currentSnapshot);
1635 final MemberName currentMember = cluster.getCurrentMemberName();
1636 Set<String> configuredShardList =
1637 new HashSet<>(configuration.getMemberShardNames(currentMember));
1638 for (String shard : currentSnapshot.getShardList()) {
1639 if (!configuredShardList.contains(shard)) {
1640 // add the current member as a replica for the shard
1641 LOG.debug("{}: adding shard {}", persistenceId(), shard);
1642 configuration.addMemberReplicaForShard(shard, currentMember);
1644 configuredShardList.remove(shard);
1647 for (String shard : configuredShardList) {
1648 // remove the member as a replica for the shard
1649 LOG.debug("{}: removing shard {}", persistenceId(), shard);
1650 configuration.removeMemberReplicaForShard(shard, currentMember);
1654 private void onSaveSnapshotSuccess(SaveSnapshotSuccess successMessage) {
1655 LOG.debug("{} saved ShardManager snapshot successfully. Deleting the prev snapshot if available",
1657 deleteSnapshots(new SnapshotSelectionCriteria(scala.Long.MaxValue(), successMessage.metadata().timestamp() - 1,
1661 private void onChangeShardServersVotingStatus(final ChangeShardMembersVotingStatus changeMembersVotingStatus) {
1662 LOG.debug("{}: onChangeShardServersVotingStatus: {}", persistenceId(), changeMembersVotingStatus);
1664 String shardName = changeMembersVotingStatus.getShardName();
1665 Map<String, Boolean> serverVotingStatusMap = new HashMap<>();
1666 for (Entry<String, Boolean> e: changeMembersVotingStatus.getMeberVotingStatusMap().entrySet()) {
1667 serverVotingStatusMap.put(getShardIdentifier(MemberName.forName(e.getKey()), shardName).toString(),
1671 ChangeServersVotingStatus changeServersVotingStatus = new ChangeServersVotingStatus(serverVotingStatusMap);
1673 findLocalShard(shardName, getSender(),
1674 localShardFound -> changeShardMembersVotingStatus(changeServersVotingStatus, shardName,
1675 localShardFound.getPath(), getSender()));
1678 private void onFlipShardMembersVotingStatus(FlipShardMembersVotingStatus flipMembersVotingStatus) {
1679 LOG.debug("{}: onFlipShardMembersVotingStatus: {}", persistenceId(), flipMembersVotingStatus);
1681 ActorRef sender = getSender();
1682 final String shardName = flipMembersVotingStatus.getShardName();
1683 findLocalShard(shardName, sender, localShardFound -> {
1684 Future<Object> future = ask(localShardFound.getPath(), GetOnDemandRaftState.INSTANCE,
1685 Timeout.apply(30, TimeUnit.SECONDS));
1687 future.onComplete(new OnComplete<Object>() {
1689 public void onComplete(Throwable failure, Object response) {
1690 if (failure != null) {
1691 sender.tell(new Status.Failure(new RuntimeException(
1692 String.format("Failed to access local shard %s", shardName), failure)), self());
1696 OnDemandRaftState raftState = (OnDemandRaftState) response;
1697 Map<String, Boolean> serverVotingStatusMap = new HashMap<>();
1698 for (Entry<String, Boolean> e: raftState.getPeerVotingStates().entrySet()) {
1699 serverVotingStatusMap.put(e.getKey(), !e.getValue());
1702 serverVotingStatusMap.put(getShardIdentifier(cluster.getCurrentMemberName(), shardName)
1703 .toString(), !raftState.isVoting());
1705 changeShardMembersVotingStatus(new ChangeServersVotingStatus(serverVotingStatusMap),
1706 shardName, localShardFound.getPath(), sender);
1708 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1713 private void findLocalShard(FindLocalShard message) {
1714 LOG.debug("{}: findLocalShard : {}", persistenceId(), message.getShardName());
1716 final ShardInformation shardInformation = localShards.get(message.getShardName());
1718 if (shardInformation == null) {
1719 LOG.debug("{}: Local shard {} not found - shards present: {}",
1720 persistenceId(), message.getShardName(), localShards.keySet());
1722 getSender().tell(new LocalShardNotFound(message.getShardName()), getSelf());
1726 sendResponse(shardInformation, message.isWaitUntilInitialized(), false,
1727 () -> new LocalShardFound(shardInformation.getActor()));
1730 private void findLocalShard(final String shardName, final ActorRef sender,
1731 final Consumer<LocalShardFound> onLocalShardFound) {
1732 Timeout findLocalTimeout = new Timeout(datastoreContextFactory.getBaseDatastoreContext()
1733 .getShardInitializationTimeout().duration().$times(2));
1735 Future<Object> futureObj = ask(getSelf(), new FindLocalShard(shardName, true), findLocalTimeout);
1736 futureObj.onComplete(new OnComplete<Object>() {
1738 public void onComplete(Throwable failure, Object response) {
1739 if (failure != null) {
1740 LOG.debug("{}: Received failure from FindLocalShard for shard {}", persistenceId, shardName,
1742 sender.tell(new Status.Failure(new RuntimeException(
1743 String.format("Failed to find local shard %s", shardName), failure)), self());
1745 if (response instanceof LocalShardFound) {
1746 getSelf().tell((RunnableMessage) () -> onLocalShardFound.accept((LocalShardFound) response),
1748 } else if (response instanceof LocalShardNotFound) {
1749 String msg = String.format("Local shard %s does not exist", shardName);
1750 LOG.debug("{}: {}", persistenceId, msg);
1751 sender.tell(new Status.Failure(new IllegalArgumentException(msg)), self());
1753 String msg = String.format("Failed to find local shard %s: received response: %s",
1754 shardName, response);
1755 LOG.debug("{}: {}", persistenceId, msg);
1756 sender.tell(new Status.Failure(response instanceof Throwable ? (Throwable) response :
1757 new RuntimeException(msg)), self());
1761 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1764 private void changeShardMembersVotingStatus(ChangeServersVotingStatus changeServersVotingStatus,
1765 final String shardName, final ActorRef shardActorRef, final ActorRef sender) {
1766 if (isShardReplicaOperationInProgress(shardName, sender)) {
1770 shardReplicaOperationsInProgress.add(shardName);
1772 DatastoreContext datastoreContext = newShardDatastoreContextBuilder(shardName).build();
1773 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(), shardName);
1775 LOG.debug("{}: Sending ChangeServersVotingStatus message {} to local shard {}", persistenceId(),
1776 changeServersVotingStatus, shardActorRef.path());
1778 Timeout timeout = new Timeout(datastoreContext.getShardLeaderElectionTimeout().duration().$times(2));
1779 Future<Object> futureObj = ask(shardActorRef, changeServersVotingStatus, timeout);
1781 futureObj.onComplete(new OnComplete<Object>() {
1783 public void onComplete(Throwable failure, Object response) {
1784 shardReplicaOperationsInProgress.remove(shardName);
1785 if (failure != null) {
1786 String msg = String.format("ChangeServersVotingStatus request to local shard %s failed",
1787 shardActorRef.path());
1788 LOG.debug("{}: {}", persistenceId(), msg, failure);
1789 sender.tell(new Status.Failure(new RuntimeException(msg, failure)), self());
1791 LOG.debug("{}: Received {} from local shard {}", persistenceId(), response, shardActorRef.path());
1793 ServerChangeReply replyMsg = (ServerChangeReply) response;
1794 if (replyMsg.getStatus() == ServerChangeStatus.OK) {
1795 LOG.debug("{}: ChangeServersVotingStatus succeeded for shard {}", persistenceId(), shardName);
1796 sender.tell(new Status.Success(null), getSelf());
1797 } else if (replyMsg.getStatus() == ServerChangeStatus.INVALID_REQUEST) {
1798 sender.tell(new Status.Failure(new IllegalArgumentException(String.format(
1799 "The requested voting state change for shard %s is invalid. At least one member "
1800 + "must be voting", shardId.getShardName()))), getSelf());
1802 LOG.warn("{}: ChangeServersVotingStatus failed for shard {} with status {}",
1803 persistenceId(), shardName, replyMsg.getStatus());
1805 Exception error = getServerChangeException(ChangeServersVotingStatus.class,
1806 replyMsg.getStatus(), shardActorRef.path().toString(), shardId);
1807 sender.tell(new Status.Failure(error), getSelf());
1811 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1814 private static final class ForwardedAddServerReply {
1815 ShardInformation shardInfo;
1816 AddServerReply addServerReply;
1818 boolean removeShardOnFailure;
1820 ForwardedAddServerReply(ShardInformation shardInfo, AddServerReply addServerReply, String leaderPath,
1821 boolean removeShardOnFailure) {
1822 this.shardInfo = shardInfo;
1823 this.addServerReply = addServerReply;
1824 this.leaderPath = leaderPath;
1825 this.removeShardOnFailure = removeShardOnFailure;
1829 private static final class ForwardedAddServerFailure {
1831 String failureMessage;
1833 boolean removeShardOnFailure;
1835 ForwardedAddServerFailure(String shardName, String failureMessage, Throwable failure,
1836 boolean removeShardOnFailure) {
1837 this.shardName = shardName;
1838 this.failureMessage = failureMessage;
1839 this.failure = failure;
1840 this.removeShardOnFailure = removeShardOnFailure;
1844 static class OnShardInitialized {
1845 private final Runnable replyRunnable;
1846 private Cancellable timeoutSchedule;
1848 OnShardInitialized(Runnable replyRunnable) {
1849 this.replyRunnable = replyRunnable;
1852 Runnable getReplyRunnable() {
1853 return replyRunnable;
1856 Cancellable getTimeoutSchedule() {
1857 return timeoutSchedule;
1860 void setTimeoutSchedule(Cancellable timeoutSchedule) {
1861 this.timeoutSchedule = timeoutSchedule;
1865 static class OnShardReady extends OnShardInitialized {
1866 OnShardReady(Runnable replyRunnable) {
1867 super(replyRunnable);
1871 private interface RunnableMessage extends Runnable {
1875 * The FindPrimaryResponseHandler provides specific callback methods which are invoked when a response to the
1876 * a remote or local find primary message is processed.
1878 private interface FindPrimaryResponseHandler {
1880 * Invoked when a Failure message is received as a response.
1882 * @param failure the failure exception
1884 void onFailure(Throwable failure);
1887 * Invoked when a RemotePrimaryShardFound response is received.
1889 * @param response the response
1891 void onRemotePrimaryShardFound(RemotePrimaryShardFound response);
1894 * Invoked when a LocalPrimaryShardFound response is received.
1896 * @param response the response
1898 void onLocalPrimaryFound(LocalPrimaryShardFound response);
1901 * Invoked when an unknown response is received. This is another type of failure.
1903 * @param response the response
1905 void onUnknownResponse(Object response);
1909 * The AutoFindPrimaryFailureResponseHandler automatically processes Failure responses when finding a primary
1910 * replica and sends a wrapped Failure response to some targetActor.
1912 private abstract static class AutoFindPrimaryFailureResponseHandler implements FindPrimaryResponseHandler {
1913 private final ActorRef targetActor;
1914 private final String shardName;
1915 private final String persistenceId;
1916 private final ActorRef shardManagerActor;
1919 * Constructs an instance.
1921 * @param targetActor The actor to whom the Failure response should be sent when a FindPrimary failure occurs
1922 * @param shardName The name of the shard for which the primary replica had to be found
1923 * @param persistenceId The persistenceId for the ShardManager
1924 * @param shardManagerActor The ShardManager actor which triggered the call to FindPrimary
1926 protected AutoFindPrimaryFailureResponseHandler(ActorRef targetActor, String shardName, String persistenceId,
1927 ActorRef shardManagerActor) {
1928 this.targetActor = Preconditions.checkNotNull(targetActor);
1929 this.shardName = Preconditions.checkNotNull(shardName);
1930 this.persistenceId = Preconditions.checkNotNull(persistenceId);
1931 this.shardManagerActor = Preconditions.checkNotNull(shardManagerActor);
1934 public ActorRef getTargetActor() {
1938 public String getShardName() {
1943 public void onFailure(Throwable failure) {
1944 LOG.debug("{}: Received failure from FindPrimary for shard {}", persistenceId, shardName, failure);
1945 targetActor.tell(new Status.Failure(new RuntimeException(
1946 String.format("Failed to find leader for shard %s", shardName), failure)), shardManagerActor);
1950 public void onUnknownResponse(Object response) {
1951 String msg = String.format("Failed to find leader for shard %s: received response: %s",
1952 shardName, response);
1953 LOG.debug("{}: {}", persistenceId, msg);
1954 targetActor.tell(new Status.Failure(response instanceof Throwable ? (Throwable) response :
1955 new RuntimeException(msg)), shardManagerActor);
1960 * The WrappedShardResponse class wraps a response from a Shard.
1962 private static final class WrappedShardResponse {
1963 private final ShardIdentifier shardId;
1964 private final Object response;
1965 private final String leaderPath;
1967 WrappedShardResponse(ShardIdentifier shardId, Object response, String leaderPath) {
1968 this.shardId = shardId;
1969 this.response = response;
1970 this.leaderPath = leaderPath;
1973 ShardIdentifier getShardId() {
1977 Object getResponse() {
1981 String getLeaderPath() {
1986 private static final class ShardNotInitializedTimeout {
1987 private final ActorRef sender;
1988 private final ShardInformation shardInfo;
1989 private final OnShardInitialized onShardInitialized;
1991 ShardNotInitializedTimeout(ShardInformation shardInfo, OnShardInitialized onShardInitialized, ActorRef sender) {
1992 this.sender = sender;
1993 this.shardInfo = shardInfo;
1994 this.onShardInitialized = onShardInitialized;
1997 ActorRef getSender() {
2001 ShardInformation getShardInfo() {
2005 OnShardInitialized getOnShardInitialized() {
2006 return onShardInitialized;