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.common.actor.Dispatchers;
55 import org.opendaylight.controller.cluster.datastore.AbstractDataStore;
56 import org.opendaylight.controller.cluster.datastore.ClusterWrapper;
57 import org.opendaylight.controller.cluster.datastore.DatastoreContext;
58 import org.opendaylight.controller.cluster.datastore.DatastoreContext.Builder;
59 import org.opendaylight.controller.cluster.datastore.DatastoreContextFactory;
60 import org.opendaylight.controller.cluster.datastore.Shard;
61 import org.opendaylight.controller.cluster.datastore.config.Configuration;
62 import org.opendaylight.controller.cluster.datastore.config.ModuleShardConfiguration;
63 import org.opendaylight.controller.cluster.datastore.config.PrefixShardConfiguration;
64 import org.opendaylight.controller.cluster.datastore.exceptions.AlreadyExistsException;
65 import org.opendaylight.controller.cluster.datastore.exceptions.NoShardLeaderException;
66 import org.opendaylight.controller.cluster.datastore.exceptions.NotInitializedException;
67 import org.opendaylight.controller.cluster.datastore.exceptions.PrimaryNotFoundException;
68 import org.opendaylight.controller.cluster.datastore.identifiers.ShardIdentifier;
69 import org.opendaylight.controller.cluster.datastore.messages.ActorInitialized;
70 import org.opendaylight.controller.cluster.datastore.messages.AddPrefixShardReplica;
71 import org.opendaylight.controller.cluster.datastore.messages.AddShardReplica;
72 import org.opendaylight.controller.cluster.datastore.messages.ChangeShardMembersVotingStatus;
73 import org.opendaylight.controller.cluster.datastore.messages.CreateShard;
74 import org.opendaylight.controller.cluster.datastore.messages.FindLocalShard;
75 import org.opendaylight.controller.cluster.datastore.messages.FindPrimary;
76 import org.opendaylight.controller.cluster.datastore.messages.FlipShardMembersVotingStatus;
77 import org.opendaylight.controller.cluster.datastore.messages.GetShardRole;
78 import org.opendaylight.controller.cluster.datastore.messages.GetShardRoleReply;
79 import org.opendaylight.controller.cluster.datastore.messages.LocalPrimaryShardFound;
80 import org.opendaylight.controller.cluster.datastore.messages.LocalShardFound;
81 import org.opendaylight.controller.cluster.datastore.messages.LocalShardNotFound;
82 import org.opendaylight.controller.cluster.datastore.messages.RemoteFindPrimary;
83 import org.opendaylight.controller.cluster.datastore.messages.RemotePrimaryShardFound;
84 import org.opendaylight.controller.cluster.datastore.messages.RemovePrefixShardReplica;
85 import org.opendaylight.controller.cluster.datastore.messages.RemoveShardReplica;
86 import org.opendaylight.controller.cluster.datastore.messages.ShardLeaderStateChanged;
87 import org.opendaylight.controller.cluster.datastore.messages.UpdateSchemaContext;
88 import org.opendaylight.controller.cluster.datastore.persisted.DatastoreSnapshot;
89 import org.opendaylight.controller.cluster.datastore.persisted.ShardManagerSnapshot;
90 import org.opendaylight.controller.cluster.datastore.utils.ClusterUtils;
91 import org.opendaylight.controller.cluster.datastore.utils.CompositeOnComplete;
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(final 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(final 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(final Throwable failure, final 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(final WrappedShardResponse message) {
395 if (message.getResponse() instanceof RemoveServerReply) {
396 onRemoveServerReply(getSender(), message.getShardId(), (RemoveServerReply) message.getResponse(),
397 message.getLeaderPath());
401 private void onRemoveServerReply(final ActorRef originalSender, final ShardIdentifier shardId,
402 final RemoveServerReply replyMsg, final String leaderPath) {
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(final Throwable failure, final 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(final RemoveShardReplica contextMessage, final String shardName,
461 final String primaryPath, 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(final Throwable failure, final 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(final 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 // At any rate, invalidate primaryShardInfo cache
535 primaryShardInfoCache.remove(shardName);
537 shardActorsStopping.remove(shardName);
538 notifyOnCompleteTasks(failure, result);
539 }, ActorRef.noSender());
543 shardActorsStopping.put(shardName, onComplete);
544 stopFuture.onComplete(onComplete, new Dispatchers(context().system().dispatchers())
545 .getDispatcher(Dispatchers.DispatcherType.Client));
548 LOG.debug("{} : Local Shard replica for shard {} has been removed", persistenceId(), shardName);
552 private void onGetSnapshot() {
553 LOG.debug("{}: onGetSnapshot", persistenceId());
555 List<String> notInitialized = null;
556 for (ShardInformation shardInfo : localShards.values()) {
557 if (!shardInfo.isShardInitialized()) {
558 if (notInitialized == null) {
559 notInitialized = new ArrayList<>();
562 notInitialized.add(shardInfo.getShardName());
566 if (notInitialized != null) {
567 getSender().tell(new Status.Failure(new IllegalStateException(String.format(
568 "%d shard(s) %s are not initialized", notInitialized.size(), notInitialized))), getSelf());
572 ActorRef replyActor = getContext().actorOf(ShardManagerGetSnapshotReplyActor.props(
573 new ArrayList<>(localShards.keySet()), type, currentSnapshot , getSender(), persistenceId(),
574 datastoreContextFactory.getBaseDatastoreContext().getShardInitializationTimeout().duration()));
576 for (ShardInformation shardInfo: localShards.values()) {
577 shardInfo.getActor().tell(GetSnapshot.INSTANCE, replyActor);
581 @SuppressWarnings("checkstyle:IllegalCatch")
582 private void onCreateShard(final CreateShard createShard) {
583 LOG.debug("{}: onCreateShard: {}", persistenceId(), createShard);
587 String shardName = createShard.getModuleShardConfig().getShardName();
588 if (localShards.containsKey(shardName)) {
589 LOG.debug("{}: Shard {} already exists", persistenceId(), shardName);
590 reply = new Status.Success(String.format("Shard with name %s already exists", shardName));
592 doCreateShard(createShard);
593 reply = new Status.Success(null);
595 } catch (Exception e) {
596 LOG.error("{}: onCreateShard failed", persistenceId(), e);
597 reply = new Status.Failure(e);
600 if (getSender() != null && !getContext().system().deadLetters().equals(getSender())) {
601 getSender().tell(reply, getSelf());
605 private void onPrefixShardCreated(final PrefixShardCreated message) {
606 LOG.debug("{}: onPrefixShardCreated: {}", persistenceId(), message);
608 final PrefixShardConfiguration config = message.getConfiguration();
609 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(),
610 ClusterUtils.getCleanShardName(config.getPrefix().getRootIdentifier()));
611 final String shardName = shardId.getShardName();
613 if (isPreviousShardActorStopInProgress(shardName, message)) {
617 if (localShards.containsKey(shardName)) {
618 LOG.debug("{}: Received create for an already existing shard {}", persistenceId(), shardName);
619 final PrefixShardConfiguration existing =
620 configuration.getAllPrefixShardConfigurations().get(config.getPrefix());
622 if (existing != null && existing.equals(config)) {
623 // we don't have to do nothing here
628 doCreatePrefixShard(config, shardId, shardName);
631 private boolean isPreviousShardActorStopInProgress(final String shardName, final Object messageToDefer) {
632 final CompositeOnComplete<Boolean> stopOnComplete = shardActorsStopping.get(shardName);
633 if (stopOnComplete == null) {
637 LOG.debug("{} : Stop is in progress for shard {} - adding OnComplete callback to defer {}", persistenceId(),
638 shardName, messageToDefer);
639 final ActorRef sender = getSender();
640 stopOnComplete.addOnComplete(new OnComplete<Boolean>() {
642 public void onComplete(final Throwable failure, final Boolean result) {
643 LOG.debug("{} : Stop complete for shard {} - re-queing {}", persistenceId(), shardName, messageToDefer);
644 self().tell(messageToDefer, sender);
651 private void doCreatePrefixShard(final PrefixShardConfiguration config, final ShardIdentifier shardId,
652 final String shardName) {
653 configuration.addPrefixShardConfiguration(config);
655 final Builder builder = newShardDatastoreContextBuilder(shardName);
656 builder.logicalStoreType(LogicalDatastoreType.valueOf(config.getPrefix().getDatastoreType().name()))
657 .storeRoot(config.getPrefix().getRootIdentifier());
658 DatastoreContext shardDatastoreContext = builder.build();
660 final Map<String, String> peerAddresses = getPeerAddresses(shardName);
661 final boolean isActiveMember = true;
663 LOG.debug("{} doCreatePrefixShard: shardId: {}, memberNames: {}, peerAddresses: {}, isActiveMember: {}",
664 persistenceId(), shardId, config.getShardMemberNames(), peerAddresses, isActiveMember);
666 final ShardInformation info = new ShardInformation(shardName, shardId, peerAddresses,
667 shardDatastoreContext, Shard.builder(), peerAddressResolver);
668 info.setActiveMember(isActiveMember);
669 localShards.put(info.getShardName(), info);
671 if (schemaContext != null) {
672 info.setSchemaContext(schemaContext);
673 info.setActor(newShardActor(info));
677 private void onPrefixShardRemoved(final PrefixShardRemoved message) {
678 LOG.debug("{}: onPrefixShardRemoved : {}", persistenceId(), message);
680 final DOMDataTreeIdentifier prefix = message.getPrefix();
681 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(),
682 ClusterUtils.getCleanShardName(prefix.getRootIdentifier()));
684 configuration.removePrefixShardConfiguration(prefix);
685 removeShard(shardId);
688 private void doCreateShard(final CreateShard createShard) {
689 final ModuleShardConfiguration moduleShardConfig = createShard.getModuleShardConfig();
690 final String shardName = moduleShardConfig.getShardName();
692 configuration.addModuleShardConfiguration(moduleShardConfig);
694 DatastoreContext shardDatastoreContext = createShard.getDatastoreContext();
695 if (shardDatastoreContext == null) {
696 shardDatastoreContext = newShardDatastoreContext(shardName);
698 shardDatastoreContext = DatastoreContext.newBuilderFrom(shardDatastoreContext).shardPeerAddressResolver(
699 peerAddressResolver).build();
702 ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(), shardName);
704 boolean shardWasInRecoveredSnapshot = currentSnapshot != null
705 && currentSnapshot.getShardList().contains(shardName);
707 Map<String, String> peerAddresses;
708 boolean isActiveMember;
709 if (shardWasInRecoveredSnapshot || configuration.getMembersFromShardName(shardName)
710 .contains(cluster.getCurrentMemberName())) {
711 peerAddresses = getPeerAddresses(shardName);
712 isActiveMember = true;
714 // The local member is not in the static shard member configuration and the shard did not
715 // previously exist (ie !shardWasInRecoveredSnapshot). In this case we'll create
716 // the shard with no peers and with elections disabled so it stays as follower. A
717 // subsequent AddServer request will be needed to make it an active member.
718 isActiveMember = false;
719 peerAddresses = Collections.emptyMap();
720 shardDatastoreContext = DatastoreContext.newBuilderFrom(shardDatastoreContext)
721 .customRaftPolicyImplementation(DisableElectionsRaftPolicy.class.getName()).build();
724 LOG.debug("{} doCreateShard: shardId: {}, memberNames: {}, peerAddresses: {}, isActiveMember: {}",
725 persistenceId(), shardId, moduleShardConfig.getShardMemberNames(), peerAddresses,
728 ShardInformation info = new ShardInformation(shardName, shardId, peerAddresses,
729 shardDatastoreContext, createShard.getShardBuilder(), peerAddressResolver);
730 info.setActiveMember(isActiveMember);
731 localShards.put(info.getShardName(), info);
733 if (schemaContext != null) {
734 info.setSchemaContext(schemaContext);
735 info.setActor(newShardActor(info));
739 private DatastoreContext.Builder newShardDatastoreContextBuilder(final String shardName) {
740 return DatastoreContext.newBuilderFrom(datastoreContextFactory.getShardDatastoreContext(shardName))
741 .shardPeerAddressResolver(peerAddressResolver);
744 private DatastoreContext newShardDatastoreContext(final String shardName) {
745 return newShardDatastoreContextBuilder(shardName).build();
748 private void checkReady() {
749 if (isReadyWithLeaderId()) {
750 LOG.info("{}: All Shards are ready - data store {} is ready, available count is {}",
751 persistenceId(), type, waitTillReadyCountdownLatch.getCount());
753 waitTillReadyCountdownLatch.countDown();
757 private void onLeaderStateChanged(final ShardLeaderStateChanged leaderStateChanged) {
758 LOG.info("{}: Received LeaderStateChanged message: {}", persistenceId(), leaderStateChanged);
760 ShardInformation shardInformation = findShardInformation(leaderStateChanged.getMemberId());
761 if (shardInformation != null) {
762 shardInformation.setLocalDataTree(leaderStateChanged.getLocalShardDataTree());
763 shardInformation.setLeaderVersion(leaderStateChanged.getLeaderPayloadVersion());
764 if (shardInformation.setLeaderId(leaderStateChanged.getLeaderId())) {
765 primaryShardInfoCache.remove(shardInformation.getShardName());
770 LOG.debug("No shard found with member Id {}", leaderStateChanged.getMemberId());
774 private void onShardNotInitializedTimeout(final ShardNotInitializedTimeout message) {
775 ShardInformation shardInfo = message.getShardInfo();
777 LOG.debug("{}: Received ShardNotInitializedTimeout message for shard {}", persistenceId(),
778 shardInfo.getShardName());
780 shardInfo.removeOnShardInitialized(message.getOnShardInitialized());
782 if (!shardInfo.isShardInitialized()) {
783 LOG.debug("{}: Returning NotInitializedException for shard {}", persistenceId(), shardInfo.getShardName());
784 message.getSender().tell(createNotInitializedException(shardInfo.getShardId()), getSelf());
786 LOG.debug("{}: Returning NoShardLeaderException for shard {}", persistenceId(), shardInfo.getShardName());
787 message.getSender().tell(createNoShardLeaderException(shardInfo.getShardId()), getSelf());
791 private void onFollowerInitialSyncStatus(final FollowerInitialSyncUpStatus status) {
792 LOG.info("{} Received follower initial sync status for {} status sync done {}", persistenceId(),
793 status.getName(), status.isInitialSyncDone());
795 ShardInformation shardInformation = findShardInformation(status.getName());
797 if (shardInformation != null) {
798 shardInformation.setFollowerSyncStatus(status.isInitialSyncDone());
800 shardManagerMBean.setSyncStatus(isInSync());
805 private void onRoleChangeNotification(final RoleChangeNotification roleChanged) {
806 LOG.info("{}: Received role changed for {} from {} to {}", persistenceId(), roleChanged.getMemberId(),
807 roleChanged.getOldRole(), roleChanged.getNewRole());
809 ShardInformation shardInformation = findShardInformation(roleChanged.getMemberId());
810 if (shardInformation != null) {
811 shardInformation.setRole(roleChanged.getNewRole());
813 shardManagerMBean.setSyncStatus(isInSync());
818 private ShardInformation findShardInformation(final String memberId) {
819 for (ShardInformation info : localShards.values()) {
820 if (info.getShardId().toString().equals(memberId)) {
828 private boolean isReadyWithLeaderId() {
829 boolean isReady = true;
830 for (ShardInformation info : localShards.values()) {
831 if (!info.isShardReadyWithLeaderId()) {
839 private boolean isInSync() {
840 for (ShardInformation info : localShards.values()) {
841 if (!info.isInSync()) {
848 private void onActorInitialized(final Object message) {
849 final ActorRef sender = getSender();
851 if (sender == null) {
852 return; //why is a non-actor sending this message? Just ignore.
855 String actorName = sender.path().name();
856 //find shard name from actor name; actor name is stringified shardId
858 final ShardIdentifier shardId;
860 shardId = ShardIdentifier.fromShardIdString(actorName);
861 } catch (IllegalArgumentException e) {
862 LOG.debug("{}: ignoring actor {}", actorName, e);
866 markShardAsInitialized(shardId.getShardName());
869 private void markShardAsInitialized(final String shardName) {
870 LOG.debug("{}: Initializing shard [{}]", persistenceId(), shardName);
872 ShardInformation shardInformation = localShards.get(shardName);
873 if (shardInformation != null) {
874 shardInformation.setActorInitialized();
876 shardInformation.getActor().tell(new RegisterRoleChangeListener(), self());
881 protected void handleRecover(final Object message) throws Exception {
882 if (message instanceof RecoveryCompleted) {
883 onRecoveryCompleted();
884 } else if (message instanceof SnapshotOffer) {
885 applyShardManagerSnapshot((ShardManagerSnapshot)((SnapshotOffer) message).snapshot());
889 @SuppressWarnings("checkstyle:IllegalCatch")
890 private void onRecoveryCompleted() {
891 LOG.info("Recovery complete : {}", persistenceId());
893 if (currentSnapshot == null && restoreFromSnapshot != null
894 && restoreFromSnapshot.getShardManagerSnapshot() != null) {
895 ShardManagerSnapshot snapshot = restoreFromSnapshot.getShardManagerSnapshot();
897 LOG.debug("{}: Restoring from ShardManagerSnapshot: {}", persistenceId(), snapshot);
899 applyShardManagerSnapshot(snapshot);
905 private void sendResponse(final ShardInformation shardInformation, final boolean doWait,
906 final boolean wantShardReady, final Supplier<Object> messageSupplier) {
907 if (!shardInformation.isShardInitialized() || wantShardReady && !shardInformation.isShardReadyWithLeaderId()) {
909 final ActorRef sender = getSender();
910 final ActorRef self = self();
912 Runnable replyRunnable = () -> sender.tell(messageSupplier.get(), self);
914 OnShardInitialized onShardInitialized = wantShardReady ? new OnShardReady(replyRunnable) :
915 new OnShardInitialized(replyRunnable);
917 shardInformation.addOnShardInitialized(onShardInitialized);
919 FiniteDuration timeout = shardInformation.getDatastoreContext()
920 .getShardInitializationTimeout().duration();
921 if (shardInformation.isShardInitialized()) {
922 // If the shard is already initialized then we'll wait enough time for the shard to
923 // elect a leader, ie 2 times the election timeout.
924 timeout = FiniteDuration.create(shardInformation.getDatastoreContext().getShardRaftConfig()
925 .getElectionTimeOutInterval().toMillis() * 2, TimeUnit.MILLISECONDS);
928 LOG.debug("{}: Scheduling {} ms timer to wait for shard {}", persistenceId(), timeout.toMillis(),
929 shardInformation.getShardName());
931 Cancellable timeoutSchedule = getContext().system().scheduler().scheduleOnce(
933 new ShardNotInitializedTimeout(shardInformation, onShardInitialized, sender),
934 getContext().dispatcher(), getSelf());
936 onShardInitialized.setTimeoutSchedule(timeoutSchedule);
938 } else if (!shardInformation.isShardInitialized()) {
939 LOG.debug("{}: Returning NotInitializedException for shard {}", persistenceId(),
940 shardInformation.getShardName());
941 getSender().tell(createNotInitializedException(shardInformation.getShardId()), getSelf());
943 LOG.debug("{}: Returning NoShardLeaderException for shard {}", persistenceId(),
944 shardInformation.getShardName());
945 getSender().tell(createNoShardLeaderException(shardInformation.getShardId()), getSelf());
951 getSender().tell(messageSupplier.get(), getSelf());
954 private static NoShardLeaderException createNoShardLeaderException(final ShardIdentifier shardId) {
955 return new NoShardLeaderException(null, shardId.toString());
958 private static NotInitializedException createNotInitializedException(final ShardIdentifier shardId) {
959 return new NotInitializedException(String.format(
960 "Found primary shard %s but it's not initialized yet. Please try again later", shardId));
964 static MemberName memberToName(final Member member) {
965 return MemberName.forName(member.roles().iterator().next());
968 private void memberRemoved(final ClusterEvent.MemberRemoved message) {
969 MemberName memberName = memberToName(message.member());
971 LOG.info("{}: Received MemberRemoved: memberName: {}, address: {}", persistenceId(), memberName,
972 message.member().address());
974 peerAddressResolver.removePeerAddress(memberName);
976 for (ShardInformation info : localShards.values()) {
977 info.peerDown(memberName, getShardIdentifier(memberName, info.getShardName()).toString(), getSelf());
981 private void memberExited(final ClusterEvent.MemberExited message) {
982 MemberName memberName = memberToName(message.member());
984 LOG.info("{}: Received MemberExited: memberName: {}, address: {}", persistenceId(), memberName,
985 message.member().address());
987 peerAddressResolver.removePeerAddress(memberName);
989 for (ShardInformation info : localShards.values()) {
990 info.peerDown(memberName, getShardIdentifier(memberName, info.getShardName()).toString(), getSelf());
994 private void memberUp(final ClusterEvent.MemberUp message) {
995 MemberName memberName = memberToName(message.member());
997 LOG.info("{}: Received MemberUp: memberName: {}, address: {}", persistenceId(), memberName,
998 message.member().address());
1000 memberUp(memberName, message.member().address());
1003 private void memberUp(final MemberName memberName, final Address address) {
1004 addPeerAddress(memberName, address);
1008 private void memberWeaklyUp(final MemberWeaklyUp message) {
1009 MemberName memberName = memberToName(message.member());
1011 LOG.info("{}: Received MemberWeaklyUp: memberName: {}, address: {}", persistenceId(), memberName,
1012 message.member().address());
1014 memberUp(memberName, message.member().address());
1017 private void addPeerAddress(final MemberName memberName, final Address address) {
1018 peerAddressResolver.addPeerAddress(memberName, address);
1020 for (ShardInformation info : localShards.values()) {
1021 String shardName = info.getShardName();
1022 String peerId = getShardIdentifier(memberName, shardName).toString();
1023 info.updatePeerAddress(peerId, peerAddressResolver.getShardActorAddress(shardName, memberName), getSelf());
1025 info.peerUp(memberName, peerId, getSelf());
1029 private void memberReachable(final ClusterEvent.ReachableMember message) {
1030 MemberName memberName = memberToName(message.member());
1031 LOG.info("Received ReachableMember: memberName {}, address: {}", memberName, message.member().address());
1033 addPeerAddress(memberName, message.member().address());
1035 markMemberAvailable(memberName);
1038 private void memberUnreachable(final ClusterEvent.UnreachableMember message) {
1039 MemberName memberName = memberToName(message.member());
1040 LOG.info("Received UnreachableMember: memberName {}, address: {}", memberName, message.member().address());
1042 markMemberUnavailable(memberName);
1045 private void markMemberUnavailable(final MemberName memberName) {
1046 final String memberStr = memberName.getName();
1047 for (ShardInformation info : localShards.values()) {
1048 String leaderId = info.getLeaderId();
1049 // XXX: why are we using String#contains() here?
1050 if (leaderId != null && leaderId.contains(memberStr)) {
1051 LOG.debug("Marking Leader {} as unavailable.", leaderId);
1052 info.setLeaderAvailable(false);
1054 primaryShardInfoCache.remove(info.getShardName());
1057 info.peerDown(memberName, getShardIdentifier(memberName, info.getShardName()).toString(), getSelf());
1061 private void markMemberAvailable(final MemberName memberName) {
1062 final String memberStr = memberName.getName();
1063 for (ShardInformation info : localShards.values()) {
1064 String leaderId = info.getLeaderId();
1065 // XXX: why are we using String#contains() here?
1066 if (leaderId != null && leaderId.contains(memberStr)) {
1067 LOG.debug("Marking Leader {} as available.", leaderId);
1068 info.setLeaderAvailable(true);
1071 info.peerUp(memberName, getShardIdentifier(memberName, info.getShardName()).toString(), getSelf());
1075 private void onDatastoreContextFactory(final DatastoreContextFactory factory) {
1076 datastoreContextFactory = factory;
1077 for (ShardInformation info : localShards.values()) {
1078 info.setDatastoreContext(newShardDatastoreContext(info.getShardName()), getSelf());
1082 private void onGetLocalShardIds() {
1083 final List<String> response = new ArrayList<>(localShards.size());
1085 for (ShardInformation info : localShards.values()) {
1086 response.add(info.getShardId().toString());
1089 getSender().tell(new Status.Success(response), getSelf());
1092 private void onSwitchShardBehavior(final SwitchShardBehavior message) {
1093 final ShardIdentifier identifier = message.getShardId();
1095 if (identifier != null) {
1096 final ShardInformation info = localShards.get(identifier.getShardName());
1098 getSender().tell(new Status.Failure(
1099 new IllegalArgumentException("Shard " + identifier + " is not local")), getSelf());
1103 switchShardBehavior(info, new SwitchBehavior(message.getNewState(), message.getTerm()));
1105 for (ShardInformation info : localShards.values()) {
1106 switchShardBehavior(info, new SwitchBehavior(message.getNewState(), message.getTerm()));
1110 getSender().tell(new Status.Success(null), getSelf());
1113 private void switchShardBehavior(final ShardInformation info, final SwitchBehavior switchBehavior) {
1114 final ActorRef actor = info.getActor();
1115 if (actor != null) {
1116 actor.tell(switchBehavior, getSelf());
1118 LOG.warn("Could not switch the behavior of shard {} to {} - shard is not yet available",
1119 info.getShardName(), switchBehavior.getNewState());
1124 * Notifies all the local shards of a change in the schema context.
1126 * @param message the message to send
1128 private void updateSchemaContext(final Object message) {
1129 schemaContext = ((UpdateSchemaContext) message).getSchemaContext();
1131 LOG.debug("Got updated SchemaContext: # of modules {}", schemaContext.getAllModuleIdentifiers().size());
1133 for (ShardInformation info : localShards.values()) {
1134 info.setSchemaContext(schemaContext);
1136 if (info.getActor() == null) {
1137 LOG.debug("Creating Shard {}", info.getShardId());
1138 info.setActor(newShardActor(info));
1140 info.getActor().tell(message, getSelf());
1146 protected ClusterWrapper getCluster() {
1151 protected ActorRef newShardActor(final ShardInformation info) {
1152 return getContext().actorOf(info.newProps().withDispatcher(shardDispatcherPath),
1153 info.getShardId().toString());
1156 private void findPrimary(final FindPrimary message) {
1157 LOG.debug("{}: In findPrimary: {}", persistenceId(), message);
1159 final String shardName = message.getShardName();
1160 final boolean canReturnLocalShardState = !(message instanceof RemoteFindPrimary);
1162 // First see if the there is a local replica for the shard
1163 final ShardInformation info = localShards.get(shardName);
1164 if (info != null && info.isActiveMember()) {
1165 sendResponse(info, message.isWaitUntilReady(), true, () -> {
1166 String primaryPath = info.getSerializedLeaderActor();
1167 Object found = canReturnLocalShardState && info.isLeader()
1168 ? new LocalPrimaryShardFound(primaryPath, info.getLocalShardDataTree().get()) :
1169 new RemotePrimaryShardFound(primaryPath, info.getLeaderVersion());
1171 LOG.debug("{}: Found primary for {}: {}", persistenceId(), shardName, found);
1178 final Collection<String> visitedAddresses;
1179 if (message instanceof RemoteFindPrimary) {
1180 visitedAddresses = ((RemoteFindPrimary)message).getVisitedAddresses();
1182 visitedAddresses = new ArrayList<>(1);
1185 visitedAddresses.add(peerAddressResolver.getShardManagerActorPathBuilder(cluster.getSelfAddress()).toString());
1187 for (String address: peerAddressResolver.getShardManagerPeerActorAddresses()) {
1188 if (visitedAddresses.contains(address)) {
1192 LOG.debug("{}: findPrimary for {} forwarding to remote ShardManager {}, visitedAddresses: {}",
1193 persistenceId(), shardName, address, visitedAddresses);
1195 getContext().actorSelection(address).forward(new RemoteFindPrimary(shardName,
1196 message.isWaitUntilReady(), visitedAddresses), getContext());
1200 LOG.debug("{}: No shard found for {}", persistenceId(), shardName);
1202 getSender().tell(new PrimaryNotFoundException(
1203 String.format("No primary shard found for %s.", shardName)), getSelf());
1206 private void findPrimary(final String shardName, final FindPrimaryResponseHandler handler) {
1207 Timeout findPrimaryTimeout = new Timeout(datastoreContextFactory.getBaseDatastoreContext()
1208 .getShardInitializationTimeout().duration().$times(2));
1210 Future<Object> futureObj = ask(getSelf(), new FindPrimary(shardName, true), findPrimaryTimeout);
1211 futureObj.onComplete(new OnComplete<Object>() {
1213 public void onComplete(final Throwable failure, final Object response) {
1214 if (failure != null) {
1215 handler.onFailure(failure);
1217 if (response instanceof RemotePrimaryShardFound) {
1218 handler.onRemotePrimaryShardFound((RemotePrimaryShardFound) response);
1219 } else if (response instanceof LocalPrimaryShardFound) {
1220 handler.onLocalPrimaryFound((LocalPrimaryShardFound) response);
1222 handler.onUnknownResponse(response);
1226 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1230 * Construct the name of the shard actor given the name of the member on
1231 * which the shard resides and the name of the shard.
1233 * @param memberName the member name
1234 * @param shardName the shard name
1237 private ShardIdentifier getShardIdentifier(final MemberName memberName, final String shardName) {
1238 return peerAddressResolver.getShardIdentifier(memberName, shardName);
1242 * Create shards that are local to the member on which the ShardManager runs.
1244 private void createLocalShards() {
1245 MemberName memberName = this.cluster.getCurrentMemberName();
1246 Collection<String> memberShardNames = this.configuration.getMemberShardNames(memberName);
1248 Map<String, DatastoreSnapshot.ShardSnapshot> shardSnapshots = new HashMap<>();
1249 if (restoreFromSnapshot != null) {
1250 for (DatastoreSnapshot.ShardSnapshot snapshot: restoreFromSnapshot.getShardSnapshots()) {
1251 shardSnapshots.put(snapshot.getName(), snapshot);
1255 restoreFromSnapshot = null; // null out to GC
1257 for (String shardName : memberShardNames) {
1258 ShardIdentifier shardId = getShardIdentifier(memberName, shardName);
1260 LOG.debug("{}: Creating local shard: {}", persistenceId(), shardId);
1262 Map<String, String> peerAddresses = getPeerAddresses(shardName);
1263 localShards.put(shardName, new ShardInformation(shardName, shardId, peerAddresses,
1264 newShardDatastoreContext(shardName), Shard.builder().restoreFromSnapshot(
1265 shardSnapshots.get(shardName)), peerAddressResolver));
1270 * Given the name of the shard find the addresses of all it's peers.
1272 * @param shardName the shard name
1274 private Map<String, String> getPeerAddresses(final String shardName) {
1275 final Collection<MemberName> members = configuration.getMembersFromShardName(shardName);
1276 return getPeerAddresses(shardName, members);
1279 private Map<String, String> getPeerAddresses(final String shardName, final Collection<MemberName> members) {
1280 Map<String, String> peerAddresses = new HashMap<>();
1281 MemberName currentMemberName = this.cluster.getCurrentMemberName();
1283 for (MemberName memberName : members) {
1284 if (!currentMemberName.equals(memberName)) {
1285 ShardIdentifier shardId = getShardIdentifier(memberName, shardName);
1286 String address = peerAddressResolver.getShardActorAddress(shardName, memberName);
1287 peerAddresses.put(shardId.toString(), address);
1290 return peerAddresses;
1294 public SupervisorStrategy supervisorStrategy() {
1296 return new OneForOneStrategy(10, Duration.create("1 minute"),
1297 (Function<Throwable, Directive>) t -> {
1298 LOG.warn("Supervisor Strategy caught unexpected exception - resuming", t);
1299 return SupervisorStrategy.resume();
1304 public String persistenceId() {
1305 return persistenceId;
1309 ShardManagerInfoMBean getMBean() {
1310 return shardManagerMBean;
1313 private boolean isShardReplicaOperationInProgress(final String shardName, final ActorRef sender) {
1314 if (shardReplicaOperationsInProgress.contains(shardName)) {
1315 String msg = String.format("A shard replica operation for %s is already in progress", shardName);
1316 LOG.debug("{}: {}", persistenceId(), msg);
1317 sender.tell(new Status.Failure(new IllegalStateException(msg)), getSelf());
1324 private void onAddPrefixShardReplica(final AddPrefixShardReplica message) {
1325 LOG.debug("{}: onAddPrefixShardReplica: {}", persistenceId(), message);
1327 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(),
1328 ClusterUtils.getCleanShardName(message.getShardPrefix()));
1329 final String shardName = shardId.getShardName();
1331 // Create the localShard
1332 if (schemaContext == null) {
1333 String msg = String.format(
1334 "No SchemaContext is available in order to create a local shard instance for %s", shardName);
1335 LOG.debug("{}: {}", persistenceId(), msg);
1336 getSender().tell(new Status.Failure(new IllegalStateException(msg)), getSelf());
1340 findPrimary(shardName, new AutoFindPrimaryFailureResponseHandler(getSender(), shardName, persistenceId(),
1343 public void onRemotePrimaryShardFound(final RemotePrimaryShardFound response) {
1344 final RunnableMessage runnable = (RunnableMessage) () -> addPrefixShard(getShardName(),
1345 message.getShardPrefix(), response, getSender());
1346 if (!isPreviousShardActorStopInProgress(getShardName(), runnable)) {
1347 getSelf().tell(runnable, getTargetActor());
1352 public void onLocalPrimaryFound(final LocalPrimaryShardFound message) {
1353 sendLocalReplicaAlreadyExistsReply(getShardName(), getTargetActor());
1358 private void onAddShardReplica(final AddShardReplica shardReplicaMsg) {
1359 final String shardName = shardReplicaMsg.getShardName();
1361 LOG.debug("{}: onAddShardReplica: {}", persistenceId(), shardReplicaMsg);
1363 // verify the shard with the specified name is present in the cluster configuration
1364 if (!this.configuration.isShardConfigured(shardName)) {
1365 String msg = String.format("No module configuration exists for shard %s", shardName);
1366 LOG.debug("{}: {}", persistenceId(), msg);
1367 getSender().tell(new Status.Failure(new IllegalArgumentException(msg)), getSelf());
1371 // Create the localShard
1372 if (schemaContext == null) {
1373 String msg = String.format(
1374 "No SchemaContext is available in order to create a local shard instance for %s", shardName);
1375 LOG.debug("{}: {}", persistenceId(), msg);
1376 getSender().tell(new Status.Failure(new IllegalStateException(msg)), getSelf());
1380 findPrimary(shardName, new AutoFindPrimaryFailureResponseHandler(getSender(), shardName, persistenceId(),
1383 public void onRemotePrimaryShardFound(final RemotePrimaryShardFound response) {
1384 final RunnableMessage runnable = (RunnableMessage) () ->
1385 addShard(getShardName(), response, getSender());
1386 if (!isPreviousShardActorStopInProgress(getShardName(), runnable)) {
1387 getSelf().tell(runnable, getTargetActor());
1392 public void onLocalPrimaryFound(final LocalPrimaryShardFound message) {
1393 sendLocalReplicaAlreadyExistsReply(getShardName(), getTargetActor());
1398 private void sendLocalReplicaAlreadyExistsReply(final String shardName, final ActorRef sender) {
1399 String msg = String.format("Local shard %s already exists", shardName);
1400 LOG.debug("{}: {}", persistenceId(), msg);
1401 sender.tell(new Status.Failure(new AlreadyExistsException(msg)), getSelf());
1404 private void addPrefixShard(final String shardName, final YangInstanceIdentifier shardPrefix,
1405 final RemotePrimaryShardFound response, final ActorRef sender) {
1406 if (isShardReplicaOperationInProgress(shardName, sender)) {
1410 shardReplicaOperationsInProgress.add(shardName);
1412 final ShardInformation shardInfo;
1413 final boolean removeShardOnFailure;
1414 ShardInformation existingShardInfo = localShards.get(shardName);
1415 if (existingShardInfo == null) {
1416 removeShardOnFailure = true;
1417 ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(), shardName);
1419 final Builder builder = newShardDatastoreContextBuilder(shardName);
1420 builder.storeRoot(shardPrefix).customRaftPolicyImplementation(DisableElectionsRaftPolicy.class.getName());
1422 DatastoreContext datastoreContext = builder.build();
1424 shardInfo = new ShardInformation(shardName, shardId, getPeerAddresses(shardName), datastoreContext,
1425 Shard.builder(), peerAddressResolver);
1426 shardInfo.setActiveMember(false);
1427 shardInfo.setSchemaContext(schemaContext);
1428 localShards.put(shardName, shardInfo);
1429 shardInfo.setActor(newShardActor(shardInfo));
1431 removeShardOnFailure = false;
1432 shardInfo = existingShardInfo;
1435 execAddShard(shardName, shardInfo, response, removeShardOnFailure, sender);
1438 private void addShard(final String shardName, final RemotePrimaryShardFound response, final ActorRef sender) {
1439 if (isShardReplicaOperationInProgress(shardName, sender)) {
1443 shardReplicaOperationsInProgress.add(shardName);
1445 final ShardInformation shardInfo;
1446 final boolean removeShardOnFailure;
1447 ShardInformation existingShardInfo = localShards.get(shardName);
1448 if (existingShardInfo == null) {
1449 removeShardOnFailure = true;
1450 ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(), shardName);
1452 DatastoreContext datastoreContext = newShardDatastoreContextBuilder(shardName)
1453 .customRaftPolicyImplementation(DisableElectionsRaftPolicy.class.getName()).build();
1455 shardInfo = new ShardInformation(shardName, shardId, getPeerAddresses(shardName), datastoreContext,
1456 Shard.builder(), peerAddressResolver);
1457 shardInfo.setActiveMember(false);
1458 shardInfo.setSchemaContext(schemaContext);
1459 localShards.put(shardName, shardInfo);
1460 shardInfo.setActor(newShardActor(shardInfo));
1462 removeShardOnFailure = false;
1463 shardInfo = existingShardInfo;
1466 execAddShard(shardName, shardInfo, response, removeShardOnFailure, sender);
1469 private void execAddShard(final String shardName,
1470 final ShardInformation shardInfo,
1471 final RemotePrimaryShardFound response,
1472 final boolean removeShardOnFailure,
1473 final ActorRef sender) {
1475 final String localShardAddress =
1476 peerAddressResolver.getShardActorAddress(shardName, cluster.getCurrentMemberName());
1478 //inform ShardLeader to add this shard as a replica by sending an AddServer message
1479 LOG.debug("{}: Sending AddServer message to peer {} for shard {}", persistenceId(),
1480 response.getPrimaryPath(), shardInfo.getShardId());
1482 final Timeout addServerTimeout = new Timeout(shardInfo.getDatastoreContext()
1483 .getShardLeaderElectionTimeout().duration());
1484 final Future<Object> futureObj = ask(getContext().actorSelection(response.getPrimaryPath()),
1485 new AddServer(shardInfo.getShardId().toString(), localShardAddress, true), addServerTimeout);
1487 futureObj.onComplete(new OnComplete<Object>() {
1489 public void onComplete(final Throwable failure, final Object addServerResponse) {
1490 if (failure != null) {
1491 LOG.debug("{}: AddServer request to {} for {} failed", persistenceId(),
1492 response.getPrimaryPath(), shardName, failure);
1494 final String msg = String.format("AddServer request to leader %s for shard %s failed",
1495 response.getPrimaryPath(), shardName);
1496 self().tell(new ForwardedAddServerFailure(shardName, msg, failure, removeShardOnFailure), sender);
1498 self().tell(new ForwardedAddServerReply(shardInfo, (AddServerReply)addServerResponse,
1499 response.getPrimaryPath(), removeShardOnFailure), sender);
1502 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1505 private void onAddServerFailure(final String shardName, final String message, final Throwable failure,
1506 final ActorRef sender, final boolean removeShardOnFailure) {
1507 shardReplicaOperationsInProgress.remove(shardName);
1509 if (removeShardOnFailure) {
1510 ShardInformation shardInfo = localShards.remove(shardName);
1511 if (shardInfo.getActor() != null) {
1512 shardInfo.getActor().tell(PoisonPill.getInstance(), getSelf());
1516 sender.tell(new Status.Failure(message == null ? failure :
1517 new RuntimeException(message, failure)), getSelf());
1520 private void onAddServerReply(final ShardInformation shardInfo, final AddServerReply replyMsg,
1521 final ActorRef sender, final String leaderPath, final boolean removeShardOnFailure) {
1522 String shardName = shardInfo.getShardName();
1523 shardReplicaOperationsInProgress.remove(shardName);
1525 LOG.debug("{}: Received {} for shard {} from leader {}", persistenceId(), replyMsg, shardName, leaderPath);
1527 if (replyMsg.getStatus() == ServerChangeStatus.OK) {
1528 LOG.debug("{}: Leader shard successfully added the replica shard {}", persistenceId(), shardName);
1530 // Make the local shard voting capable
1531 shardInfo.setDatastoreContext(newShardDatastoreContext(shardName), getSelf());
1532 shardInfo.setActiveMember(true);
1535 sender.tell(new Status.Success(null), getSelf());
1536 } else if (replyMsg.getStatus() == ServerChangeStatus.ALREADY_EXISTS) {
1537 sendLocalReplicaAlreadyExistsReply(shardName, sender);
1539 LOG.warn("{}: Leader failed to add shard replica {} with status {}",
1540 persistenceId(), shardName, replyMsg.getStatus());
1542 Exception failure = getServerChangeException(AddServer.class, replyMsg.getStatus(), leaderPath,
1543 shardInfo.getShardId());
1545 onAddServerFailure(shardName, null, failure, sender, removeShardOnFailure);
1549 private static Exception getServerChangeException(final Class<?> serverChange,
1550 final ServerChangeStatus serverChangeStatus, final String leaderPath, final ShardIdentifier shardId) {
1551 switch (serverChangeStatus) {
1553 return new TimeoutException(String.format(
1554 "The shard leader %s timed out trying to replicate the initial data to the new shard %s."
1555 + "Possible causes - there was a problem replicating the data or shard leadership changed "
1556 + "while replicating the shard data", leaderPath, shardId.getShardName()));
1558 return createNoShardLeaderException(shardId);
1560 return new UnsupportedOperationException(String.format("%s request is not supported for shard %s",
1561 serverChange.getSimpleName(), shardId.getShardName()));
1563 return new RuntimeException(String.format("%s request to leader %s for shard %s failed with status %s",
1564 serverChange.getSimpleName(), leaderPath, shardId.getShardName(), serverChangeStatus));
1568 private void onRemoveShardReplica(final RemoveShardReplica shardReplicaMsg) {
1569 LOG.debug("{}: onRemoveShardReplica: {}", persistenceId(), shardReplicaMsg);
1571 findPrimary(shardReplicaMsg.getShardName(), new AutoFindPrimaryFailureResponseHandler(getSender(),
1572 shardReplicaMsg.getShardName(), persistenceId(), getSelf()) {
1574 public void onRemotePrimaryShardFound(final RemotePrimaryShardFound response) {
1575 doRemoveShardReplicaAsync(response.getPrimaryPath());
1579 public void onLocalPrimaryFound(final LocalPrimaryShardFound response) {
1580 doRemoveShardReplicaAsync(response.getPrimaryPath());
1583 private void doRemoveShardReplicaAsync(final String primaryPath) {
1584 getSelf().tell((RunnableMessage) () -> removeShardReplica(shardReplicaMsg, getShardName(),
1585 primaryPath, getSender()), getTargetActor());
1590 private void onRemovePrefixShardReplica(final RemovePrefixShardReplica message) {
1591 LOG.debug("{}: onRemovePrefixShardReplica: {}", persistenceId(), message);
1593 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(),
1594 ClusterUtils.getCleanShardName(message.getShardPrefix()));
1595 final String shardName = shardId.getShardName();
1597 findPrimary(shardName, new AutoFindPrimaryFailureResponseHandler(getSender(),
1598 shardName, persistenceId(), getSelf()) {
1600 public void onRemotePrimaryShardFound(final RemotePrimaryShardFound response) {
1601 doRemoveShardReplicaAsync(response.getPrimaryPath());
1605 public void onLocalPrimaryFound(final LocalPrimaryShardFound response) {
1606 doRemoveShardReplicaAsync(response.getPrimaryPath());
1609 private void doRemoveShardReplicaAsync(final String primaryPath) {
1610 getSelf().tell((RunnableMessage) () -> removePrefixShardReplica(message, getShardName(),
1611 primaryPath, getSender()), getTargetActor());
1616 private void persistShardList() {
1617 List<String> shardList = new ArrayList<>(localShards.keySet());
1618 for (ShardInformation shardInfo : localShards.values()) {
1619 if (!shardInfo.isActiveMember()) {
1620 shardList.remove(shardInfo.getShardName());
1623 LOG.debug("{}: persisting the shard list {}", persistenceId(), shardList);
1624 saveSnapshot(updateShardManagerSnapshot(shardList, configuration.getAllPrefixShardConfigurations()));
1627 private ShardManagerSnapshot updateShardManagerSnapshot(
1628 final List<String> shardList,
1629 final Map<DOMDataTreeIdentifier, PrefixShardConfiguration> allPrefixShardConfigurations) {
1630 currentSnapshot = new ShardManagerSnapshot(shardList, allPrefixShardConfigurations);
1631 return currentSnapshot;
1634 private void applyShardManagerSnapshot(final ShardManagerSnapshot snapshot) {
1635 currentSnapshot = snapshot;
1637 LOG.debug("{}: onSnapshotOffer: {}", persistenceId(), currentSnapshot);
1639 final MemberName currentMember = cluster.getCurrentMemberName();
1640 Set<String> configuredShardList =
1641 new HashSet<>(configuration.getMemberShardNames(currentMember));
1642 for (String shard : currentSnapshot.getShardList()) {
1643 if (!configuredShardList.contains(shard)) {
1644 // add the current member as a replica for the shard
1645 LOG.debug("{}: adding shard {}", persistenceId(), shard);
1646 configuration.addMemberReplicaForShard(shard, currentMember);
1648 configuredShardList.remove(shard);
1651 for (String shard : configuredShardList) {
1652 // remove the member as a replica for the shard
1653 LOG.debug("{}: removing shard {}", persistenceId(), shard);
1654 configuration.removeMemberReplicaForShard(shard, currentMember);
1658 private void onSaveSnapshotSuccess(final SaveSnapshotSuccess successMessage) {
1659 LOG.debug("{} saved ShardManager snapshot successfully. Deleting the prev snapshot if available",
1661 deleteSnapshots(new SnapshotSelectionCriteria(scala.Long.MaxValue(), successMessage.metadata().timestamp() - 1,
1665 private void onChangeShardServersVotingStatus(final ChangeShardMembersVotingStatus changeMembersVotingStatus) {
1666 LOG.debug("{}: onChangeShardServersVotingStatus: {}", persistenceId(), changeMembersVotingStatus);
1668 String shardName = changeMembersVotingStatus.getShardName();
1669 Map<String, Boolean> serverVotingStatusMap = new HashMap<>();
1670 for (Entry<String, Boolean> e: changeMembersVotingStatus.getMeberVotingStatusMap().entrySet()) {
1671 serverVotingStatusMap.put(getShardIdentifier(MemberName.forName(e.getKey()), shardName).toString(),
1675 ChangeServersVotingStatus changeServersVotingStatus = new ChangeServersVotingStatus(serverVotingStatusMap);
1677 findLocalShard(shardName, getSender(),
1678 localShardFound -> changeShardMembersVotingStatus(changeServersVotingStatus, shardName,
1679 localShardFound.getPath(), getSender()));
1682 private void onFlipShardMembersVotingStatus(final FlipShardMembersVotingStatus flipMembersVotingStatus) {
1683 LOG.debug("{}: onFlipShardMembersVotingStatus: {}", persistenceId(), flipMembersVotingStatus);
1685 ActorRef sender = getSender();
1686 final String shardName = flipMembersVotingStatus.getShardName();
1687 findLocalShard(shardName, sender, localShardFound -> {
1688 Future<Object> future = ask(localShardFound.getPath(), GetOnDemandRaftState.INSTANCE,
1689 Timeout.apply(30, TimeUnit.SECONDS));
1691 future.onComplete(new OnComplete<Object>() {
1693 public void onComplete(final Throwable failure, final Object response) {
1694 if (failure != null) {
1695 sender.tell(new Status.Failure(new RuntimeException(
1696 String.format("Failed to access local shard %s", shardName), failure)), self());
1700 OnDemandRaftState raftState = (OnDemandRaftState) response;
1701 Map<String, Boolean> serverVotingStatusMap = new HashMap<>();
1702 for (Entry<String, Boolean> e: raftState.getPeerVotingStates().entrySet()) {
1703 serverVotingStatusMap.put(e.getKey(), !e.getValue());
1706 serverVotingStatusMap.put(getShardIdentifier(cluster.getCurrentMemberName(), shardName)
1707 .toString(), !raftState.isVoting());
1709 changeShardMembersVotingStatus(new ChangeServersVotingStatus(serverVotingStatusMap),
1710 shardName, localShardFound.getPath(), sender);
1712 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1717 private void findLocalShard(final FindLocalShard message) {
1718 LOG.debug("{}: findLocalShard : {}", persistenceId(), message.getShardName());
1720 final ShardInformation shardInformation = localShards.get(message.getShardName());
1722 if (shardInformation == null) {
1723 LOG.debug("{}: Local shard {} not found - shards present: {}",
1724 persistenceId(), message.getShardName(), localShards.keySet());
1726 getSender().tell(new LocalShardNotFound(message.getShardName()), getSelf());
1730 sendResponse(shardInformation, message.isWaitUntilInitialized(), false,
1731 () -> new LocalShardFound(shardInformation.getActor()));
1734 private void findLocalShard(final String shardName, final ActorRef sender,
1735 final Consumer<LocalShardFound> onLocalShardFound) {
1736 Timeout findLocalTimeout = new Timeout(datastoreContextFactory.getBaseDatastoreContext()
1737 .getShardInitializationTimeout().duration().$times(2));
1739 Future<Object> futureObj = ask(getSelf(), new FindLocalShard(shardName, true), findLocalTimeout);
1740 futureObj.onComplete(new OnComplete<Object>() {
1742 public void onComplete(final Throwable failure, final Object response) {
1743 if (failure != null) {
1744 LOG.debug("{}: Received failure from FindLocalShard for shard {}", persistenceId, shardName,
1746 sender.tell(new Status.Failure(new RuntimeException(
1747 String.format("Failed to find local shard %s", shardName), failure)), self());
1749 if (response instanceof LocalShardFound) {
1750 getSelf().tell((RunnableMessage) () -> onLocalShardFound.accept((LocalShardFound) response),
1752 } else if (response instanceof LocalShardNotFound) {
1753 String msg = String.format("Local shard %s does not exist", shardName);
1754 LOG.debug("{}: {}", persistenceId, msg);
1755 sender.tell(new Status.Failure(new IllegalArgumentException(msg)), self());
1757 String msg = String.format("Failed to find local shard %s: received response: %s",
1758 shardName, response);
1759 LOG.debug("{}: {}", persistenceId, msg);
1760 sender.tell(new Status.Failure(response instanceof Throwable ? (Throwable) response :
1761 new RuntimeException(msg)), self());
1765 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1768 private void changeShardMembersVotingStatus(final ChangeServersVotingStatus changeServersVotingStatus,
1769 final String shardName, final ActorRef shardActorRef, final ActorRef sender) {
1770 if (isShardReplicaOperationInProgress(shardName, sender)) {
1774 shardReplicaOperationsInProgress.add(shardName);
1776 DatastoreContext datastoreContext = newShardDatastoreContextBuilder(shardName).build();
1777 final ShardIdentifier shardId = getShardIdentifier(cluster.getCurrentMemberName(), shardName);
1779 LOG.debug("{}: Sending ChangeServersVotingStatus message {} to local shard {}", persistenceId(),
1780 changeServersVotingStatus, shardActorRef.path());
1782 Timeout timeout = new Timeout(datastoreContext.getShardLeaderElectionTimeout().duration().$times(2));
1783 Future<Object> futureObj = ask(shardActorRef, changeServersVotingStatus, timeout);
1785 futureObj.onComplete(new OnComplete<Object>() {
1787 public void onComplete(final Throwable failure, final Object response) {
1788 shardReplicaOperationsInProgress.remove(shardName);
1789 if (failure != null) {
1790 String msg = String.format("ChangeServersVotingStatus request to local shard %s failed",
1791 shardActorRef.path());
1792 LOG.debug("{}: {}", persistenceId(), msg, failure);
1793 sender.tell(new Status.Failure(new RuntimeException(msg, failure)), self());
1795 LOG.debug("{}: Received {} from local shard {}", persistenceId(), response, shardActorRef.path());
1797 ServerChangeReply replyMsg = (ServerChangeReply) response;
1798 if (replyMsg.getStatus() == ServerChangeStatus.OK) {
1799 LOG.debug("{}: ChangeServersVotingStatus succeeded for shard {}", persistenceId(), shardName);
1800 sender.tell(new Status.Success(null), getSelf());
1801 } else if (replyMsg.getStatus() == ServerChangeStatus.INVALID_REQUEST) {
1802 sender.tell(new Status.Failure(new IllegalArgumentException(String.format(
1803 "The requested voting state change for shard %s is invalid. At least one member "
1804 + "must be voting", shardId.getShardName()))), getSelf());
1806 LOG.warn("{}: ChangeServersVotingStatus failed for shard {} with status {}",
1807 persistenceId(), shardName, replyMsg.getStatus());
1809 Exception error = getServerChangeException(ChangeServersVotingStatus.class,
1810 replyMsg.getStatus(), shardActorRef.path().toString(), shardId);
1811 sender.tell(new Status.Failure(error), getSelf());
1815 }, new Dispatchers(context().system().dispatchers()).getDispatcher(Dispatchers.DispatcherType.Client));
1818 private static final class ForwardedAddServerReply {
1819 ShardInformation shardInfo;
1820 AddServerReply addServerReply;
1822 boolean removeShardOnFailure;
1824 ForwardedAddServerReply(final ShardInformation shardInfo, final AddServerReply addServerReply,
1825 final String leaderPath, final boolean removeShardOnFailure) {
1826 this.shardInfo = shardInfo;
1827 this.addServerReply = addServerReply;
1828 this.leaderPath = leaderPath;
1829 this.removeShardOnFailure = removeShardOnFailure;
1833 private static final class ForwardedAddServerFailure {
1835 String failureMessage;
1837 boolean removeShardOnFailure;
1839 ForwardedAddServerFailure(final String shardName, final String failureMessage, final Throwable failure,
1840 final boolean removeShardOnFailure) {
1841 this.shardName = shardName;
1842 this.failureMessage = failureMessage;
1843 this.failure = failure;
1844 this.removeShardOnFailure = removeShardOnFailure;
1848 static class OnShardInitialized {
1849 private final Runnable replyRunnable;
1850 private Cancellable timeoutSchedule;
1852 OnShardInitialized(final Runnable replyRunnable) {
1853 this.replyRunnable = replyRunnable;
1856 Runnable getReplyRunnable() {
1857 return replyRunnable;
1860 Cancellable getTimeoutSchedule() {
1861 return timeoutSchedule;
1864 void setTimeoutSchedule(final Cancellable timeoutSchedule) {
1865 this.timeoutSchedule = timeoutSchedule;
1869 static class OnShardReady extends OnShardInitialized {
1870 OnShardReady(final Runnable replyRunnable) {
1871 super(replyRunnable);
1875 private interface RunnableMessage extends Runnable {
1879 * The FindPrimaryResponseHandler provides specific callback methods which are invoked when a response to the
1880 * a remote or local find primary message is processed.
1882 private interface FindPrimaryResponseHandler {
1884 * Invoked when a Failure message is received as a response.
1886 * @param failure the failure exception
1888 void onFailure(Throwable failure);
1891 * Invoked when a RemotePrimaryShardFound response is received.
1893 * @param response the response
1895 void onRemotePrimaryShardFound(RemotePrimaryShardFound response);
1898 * Invoked when a LocalPrimaryShardFound response is received.
1900 * @param response the response
1902 void onLocalPrimaryFound(LocalPrimaryShardFound response);
1905 * Invoked when an unknown response is received. This is another type of failure.
1907 * @param response the response
1909 void onUnknownResponse(Object response);
1913 * The AutoFindPrimaryFailureResponseHandler automatically processes Failure responses when finding a primary
1914 * replica and sends a wrapped Failure response to some targetActor.
1916 private abstract static class AutoFindPrimaryFailureResponseHandler implements FindPrimaryResponseHandler {
1917 private final ActorRef targetActor;
1918 private final String shardName;
1919 private final String persistenceId;
1920 private final ActorRef shardManagerActor;
1923 * Constructs an instance.
1925 * @param targetActor The actor to whom the Failure response should be sent when a FindPrimary failure occurs
1926 * @param shardName The name of the shard for which the primary replica had to be found
1927 * @param persistenceId The persistenceId for the ShardManager
1928 * @param shardManagerActor The ShardManager actor which triggered the call to FindPrimary
1930 protected AutoFindPrimaryFailureResponseHandler(final ActorRef targetActor, final String shardName,
1931 final String persistenceId, final ActorRef shardManagerActor) {
1932 this.targetActor = Preconditions.checkNotNull(targetActor);
1933 this.shardName = Preconditions.checkNotNull(shardName);
1934 this.persistenceId = Preconditions.checkNotNull(persistenceId);
1935 this.shardManagerActor = Preconditions.checkNotNull(shardManagerActor);
1938 public ActorRef getTargetActor() {
1942 public String getShardName() {
1947 public void onFailure(final Throwable failure) {
1948 LOG.debug("{}: Received failure from FindPrimary for shard {}", persistenceId, shardName, failure);
1949 targetActor.tell(new Status.Failure(new RuntimeException(
1950 String.format("Failed to find leader for shard %s", shardName), failure)), shardManagerActor);
1954 public void onUnknownResponse(final Object response) {
1955 String msg = String.format("Failed to find leader for shard %s: received response: %s",
1956 shardName, response);
1957 LOG.debug("{}: {}", persistenceId, msg);
1958 targetActor.tell(new Status.Failure(response instanceof Throwable ? (Throwable) response :
1959 new RuntimeException(msg)), shardManagerActor);
1964 * The WrappedShardResponse class wraps a response from a Shard.
1966 private static final class WrappedShardResponse {
1967 private final ShardIdentifier shardId;
1968 private final Object response;
1969 private final String leaderPath;
1971 WrappedShardResponse(final ShardIdentifier shardId, final Object response, final String leaderPath) {
1972 this.shardId = shardId;
1973 this.response = response;
1974 this.leaderPath = leaderPath;
1977 ShardIdentifier getShardId() {
1981 Object getResponse() {
1985 String getLeaderPath() {
1990 private static final class ShardNotInitializedTimeout {
1991 private final ActorRef sender;
1992 private final ShardInformation shardInfo;
1993 private final OnShardInitialized onShardInitialized;
1995 ShardNotInitializedTimeout(final ShardInformation shardInfo, final OnShardInitialized onShardInitialized,
1996 final ActorRef sender) {
1997 this.sender = sender;
1998 this.shardInfo = shardInfo;
1999 this.onShardInitialized = onShardInitialized;
2002 ActorRef getSender() {
2006 ShardInformation getShardInfo() {
2010 OnShardInitialized getOnShardInitialized() {
2011 return onShardInitialized;