opendaylight/md-sal/sal-akka-raft/src/main/java/org/opendaylight/controller/cluster/raft/RaftActor.java

   1 /*
   2  * Copyright (c) 2014 Cisco Systems, Inc. and others.  All rights reserved.
   3  *
   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
   7  */
   8
   9 package org.opendaylight.controller.cluster.raft;
  10
  11 import akka.actor.ActorRef;
  12 import akka.actor.ActorSelection;
  13 import akka.japi.Procedure;
  14 import akka.persistence.RecoveryCompleted;
  15 import akka.persistence.SaveSnapshotFailure;
  16 import akka.persistence.SaveSnapshotSuccess;
  17 import akka.persistence.SnapshotOffer;
  18 import akka.persistence.SnapshotSelectionCriteria;
  19 import com.google.common.annotations.VisibleForTesting;
  20 import com.google.common.base.Objects;
  21 import com.google.common.base.Optional;
  22 import com.google.common.base.Stopwatch;
  23 import com.google.common.collect.ImmutableMap;
  24 import com.google.common.collect.Lists;
  25 import java.io.Serializable;
  26 import java.util.Collection;
  27 import java.util.List;
  28 import java.util.Map;
  29 import java.util.concurrent.TimeUnit;
  30 import org.apache.commons.lang3.time.DurationFormatUtils;
  31 import org.opendaylight.controller.cluster.DataPersistenceProvider;
  32 import org.opendaylight.controller.cluster.DelegatingPersistentDataProvider;
  33 import org.opendaylight.controller.cluster.NonPersistentDataProvider;
  34 import org.opendaylight.controller.cluster.PersistentDataProvider;
  35 import org.opendaylight.controller.cluster.common.actor.AbstractUntypedPersistentActor;
  36 import org.opendaylight.controller.cluster.notifications.LeaderStateChanged;
  37 import org.opendaylight.controller.cluster.notifications.RoleChanged;
  38 import org.opendaylight.controller.cluster.raft.base.messages.ApplyJournalEntries;
  39 import org.opendaylight.controller.cluster.raft.base.messages.ApplyLogEntries;
  40 import org.opendaylight.controller.cluster.raft.base.messages.ApplySnapshot;
  41 import org.opendaylight.controller.cluster.raft.base.messages.ApplyState;
  42 import org.opendaylight.controller.cluster.raft.base.messages.CaptureSnapshot;
  43 import org.opendaylight.controller.cluster.raft.base.messages.CaptureSnapshotReply;
  44 import org.opendaylight.controller.cluster.raft.base.messages.Replicate;
  45 import org.opendaylight.controller.cluster.raft.behaviors.AbstractLeader;
  46 import org.opendaylight.controller.cluster.raft.behaviors.AbstractRaftActorBehavior;
  47 import org.opendaylight.controller.cluster.raft.behaviors.Follower;
  48 import org.opendaylight.controller.cluster.raft.behaviors.RaftActorBehavior;
  49 import org.opendaylight.controller.cluster.raft.client.messages.FindLeader;
  50 import org.opendaylight.controller.cluster.raft.client.messages.FindLeaderReply;
  51 import org.opendaylight.controller.cluster.raft.client.messages.FollowerInfo;
  52 import org.opendaylight.controller.cluster.raft.client.messages.GetOnDemandRaftState;
  53 import org.opendaylight.controller.cluster.raft.client.messages.OnDemandRaftState;
  54 import org.opendaylight.controller.cluster.raft.protobuff.client.messages.Payload;
  55 import org.slf4j.Logger;
  56 import org.slf4j.LoggerFactory;
  57
  58 /**
  59  * RaftActor encapsulates a state machine that needs to be kept synchronized
  60  * in a cluster. It implements the RAFT algorithm as described in the paper
  61  * <a href='https://ramcloud.stanford.edu/wiki/download/attachments/11370504/raft.pdf'>
  62  * In Search of an Understandable Consensus Algorithm</a>
  63  * <p/>
  64  * RaftActor has 3 states and each state has a certain behavior associated
  65  * with it. A Raft actor can behave as,
  66  * <ul>
  67  * <li> A Leader </li>
  68  * <li> A Follower (or) </li>
  69  * <li> A Candidate </li>
  70  * </ul>
  71  * <p/>
  72  * <p/>
  73  * A RaftActor MUST be a Leader in order to accept requests from clients to
  74  * change the state of it's encapsulated state machine. Once a RaftActor becomes
  75  * a Leader it is also responsible for ensuring that all followers ultimately
  76  * have the same log and therefore the same state machine as itself.
  77  * <p/>
  78  * <p/>
  79  * The current behavior of a RaftActor determines how election for leadership
  80  * is initiated and how peer RaftActors react to request for votes.
  81  * <p/>
  82  * <p/>
  83  * Each RaftActor also needs to know the current election term. It uses this
  84  * information for a couple of things. One is to simply figure out who it
  85  * voted for in the last election. Another is to figure out if the message
  86  * it received to update it's state is stale.
  87  * <p/>
  88  * <p/>
  89  * The RaftActor uses akka-persistence to store it's replicated log.
  90  * Furthermore through it's behaviors a Raft Actor determines
  91  * <p/>
  92  * <ul>
  93  * <li> when a log entry should be persisted </li>
  94  * <li> when a log entry should be applied to the state machine (and) </li>
  95  * <li> when a snapshot should be saved </li>
  96  * </ul>
  97  */
  98 public abstract class RaftActor extends AbstractUntypedPersistentActor {
  99
 100     private static final long APPLY_STATE_DELAY_THRESHOLD_IN_NANOS = TimeUnit.MILLISECONDS.toNanos(50L); // 50 millis
 101
 102     private static final Procedure<ApplyJournalEntries> APPLY_JOURNAL_ENTRIES_PERSIST_CALLBACK =
 103             new Procedure<ApplyJournalEntries>() {
 104                 @Override
 105                 public void apply(ApplyJournalEntries param) throws Exception {
 106                 }
 107             };
 108     private static final String COMMIT_SNAPSHOT = "commit_snapshot";
 109
 110     protected final Logger LOG = LoggerFactory.getLogger(getClass());
 111
 112     /**
 113      * The current state determines the current behavior of a RaftActor
 114      * A Raft Actor always starts off in the Follower State
 115      */
 116     private RaftActorBehavior currentBehavior;
 117
 118     /**
 119      * This context should NOT be passed directly to any other actor it is
 120      * only to be consumed by the RaftActorBehaviors
 121      */
 122     private final RaftActorContextImpl context;
 123
 124     private final DelegatingPersistentDataProvider delegatingPersistenceProvider = new DelegatingPersistentDataProvider(null);
 125
 126     private final Procedure<Void> createSnapshotProcedure = new CreateSnapshotProcedure();
 127
 128     /**
 129      * The in-memory journal
 130      */
 131     private ReplicatedLogImpl replicatedLog = new ReplicatedLogImpl();
 132
 133     private Stopwatch recoveryTimer;
 134
 135     private int currentRecoveryBatchCount;
 136
 137     private final BehaviorStateHolder reusableBehaviorStateHolder = new BehaviorStateHolder();
 138
 139     public RaftActor(String id, Map<String, String> peerAddresses) {
 140         this(id, peerAddresses, Optional.<ConfigParams>absent());
 141     }
 142
 143     public RaftActor(String id, Map<String, String> peerAddresses,
 144          Optional<ConfigParams> configParams) {
 145
 146         context = new RaftActorContextImpl(this.getSelf(),
 147             this.getContext(), id, new ElectionTermImpl(),
 148             -1, -1, replicatedLog, peerAddresses,
 149             (configParams.isPresent() ? configParams.get(): new DefaultConfigParamsImpl()),
 150             LOG);
 151     }
 152
 153     private void initRecoveryTimer() {
 154         if(recoveryTimer == null) {
 155             recoveryTimer = Stopwatch.createStarted();
 156         }
 157     }
 158
 159     @Override
 160     public void preStart() throws Exception {
 161         LOG.info("Starting recovery for {} with journal batch size {}", persistenceId(),
 162                 context.getConfigParams().getJournalRecoveryLogBatchSize());
 163
 164         super.preStart();
 165     }
 166
 167     @Override
 168     public void postStop() {
 169         if(currentBehavior != null) {
 170             try {
 171                 currentBehavior.close();
 172             } catch (Exception e) {
 173                 LOG.debug("{}: Error closing behavior {}", persistenceId(), currentBehavior.state());
 174             }
 175         }
 176
 177         super.postStop();
 178     }
 179
 180     @Override
 181     public void handleRecover(Object message) {
 182         if(persistence().isRecoveryApplicable()) {
 183             if (message instanceof SnapshotOffer) {
 184                 onRecoveredSnapshot((SnapshotOffer) message);
 185             } else if (message instanceof ReplicatedLogEntry) {
 186                 onRecoveredJournalLogEntry((ReplicatedLogEntry) message);
 187             } else if (message instanceof ApplyLogEntries) {
 188                 // Handle this message for backwards compatibility with pre-Lithium versions.
 189                 onRecoveredApplyLogEntries(((ApplyLogEntries) message).getToIndex());
 190             } else if (message instanceof ApplyJournalEntries) {
 191                 onRecoveredApplyLogEntries(((ApplyJournalEntries) message).getToIndex());
 192             } else if (message instanceof DeleteEntries) {
 193                 replicatedLog.removeFrom(((DeleteEntries) message).getFromIndex());
 194             } else if (message instanceof UpdateElectionTerm) {
 195                 context.getTermInformation().update(((UpdateElectionTerm) message).getCurrentTerm(),
 196                         ((UpdateElectionTerm) message).getVotedFor());
 197             } else if (message instanceof RecoveryCompleted) {
 198                 onRecoveryCompletedMessage();
 199             }
 200         } else {
 201             if (message instanceof RecoveryCompleted) {
 202                 // Delete all the messages from the akka journal so that we do not end up with consistency issues
 203                 // Note I am not using the dataPersistenceProvider and directly using the akka api here
 204                 deleteMessages(lastSequenceNr());
 205
 206                 // Delete all the akka snapshots as they will not be needed
 207                 deleteSnapshots(new SnapshotSelectionCriteria(scala.Long.MaxValue(), scala.Long.MaxValue()));
 208
 209                 onRecoveryComplete();
 210
 211                 initializeBehavior();
 212             }
 213         }
 214     }
 215
 216     private void onRecoveredSnapshot(SnapshotOffer offer) {
 217         if(LOG.isDebugEnabled()) {
 218             LOG.debug("{}: SnapshotOffer called..", persistenceId());
 219         }
 220
 221         initRecoveryTimer();
 222
 223         Snapshot snapshot = (Snapshot) offer.snapshot();
 224
 225         // Create a replicated log with the snapshot information
 226         // The replicated log can be used later on to retrieve this snapshot
 227         // when we need to install it on a peer
 228         replicatedLog = new ReplicatedLogImpl(snapshot);
 229
 230         context.setReplicatedLog(replicatedLog);
 231         context.setLastApplied(snapshot.getLastAppliedIndex());
 232         context.setCommitIndex(snapshot.getLastAppliedIndex());
 233
 234         Stopwatch timer = Stopwatch.createStarted();
 235
 236         // Apply the snapshot to the actors state
 237         applyRecoverySnapshot(snapshot.getState());
 238
 239         timer.stop();
 240         LOG.info("Recovery snapshot applied for {} in {}: snapshotIndex={}, snapshotTerm={}, journal-size=" +
 241                 replicatedLog.size(), persistenceId(), timer.toString(),
 242                 replicatedLog.getSnapshotIndex(), replicatedLog.getSnapshotTerm());
 243     }
 244
 245     private void onRecoveredJournalLogEntry(ReplicatedLogEntry logEntry) {
 246         if(LOG.isDebugEnabled()) {
 247             LOG.debug("{}: Received ReplicatedLogEntry for recovery: {}", persistenceId(), logEntry.getIndex());
 248         }
 249
 250         replicatedLog.append(logEntry);
 251     }
 252
 253     private void onRecoveredApplyLogEntries(long toIndex) {
 254         if(LOG.isDebugEnabled()) {
 255             LOG.debug("{}: Received ApplyLogEntries for recovery, applying to state: {} to {}",
 256                     persistenceId(), context.getLastApplied() + 1, toIndex);
 257         }
 258
 259         for (long i = context.getLastApplied() + 1; i <= toIndex; i++) {
 260             batchRecoveredLogEntry(replicatedLog.get(i));
 261         }
 262
 263         context.setLastApplied(toIndex);
 264         context.setCommitIndex(toIndex);
 265     }
 266
 267     private void batchRecoveredLogEntry(ReplicatedLogEntry logEntry) {
 268         initRecoveryTimer();
 269
 270         int batchSize = context.getConfigParams().getJournalRecoveryLogBatchSize();
 271         if(currentRecoveryBatchCount == 0) {
 272             startLogRecoveryBatch(batchSize);
 273         }
 274
 275         appendRecoveredLogEntry(logEntry.getData());
 276
 277         if(++currentRecoveryBatchCount >= batchSize) {
 278             endCurrentLogRecoveryBatch();
 279         }
 280     }
 281
 282     private void endCurrentLogRecoveryBatch() {
 283         applyCurrentLogRecoveryBatch();
 284         currentRecoveryBatchCount = 0;
 285     }
 286
 287     private void onRecoveryCompletedMessage() {
 288         if(currentRecoveryBatchCount > 0) {
 289             endCurrentLogRecoveryBatch();
 290         }
 291
 292         onRecoveryComplete();
 293
 294         String recoveryTime = "";
 295         if(recoveryTimer != null) {
 296             recoveryTimer.stop();
 297             recoveryTime = " in " + recoveryTimer.toString();
 298             recoveryTimer = null;
 299         }
 300
 301         LOG.info(
 302             "Recovery completed" + recoveryTime + " - Switching actor to Follower - " +
 303                 "Persistence Id =  " + persistenceId() +
 304                 " Last index in log={}, snapshotIndex={}, snapshotTerm={}, " +
 305                 "journal-size={}",
 306             replicatedLog.lastIndex(), replicatedLog.getSnapshotIndex(),
 307             replicatedLog.getSnapshotTerm(), replicatedLog.size());
 308
 309         initializeBehavior();
 310     }
 311
 312     protected void initializeBehavior(){
 313         changeCurrentBehavior(new Follower(context));
 314     }
 315
 316     protected void changeCurrentBehavior(RaftActorBehavior newBehavior){
 317         reusableBehaviorStateHolder.init(currentBehavior);
 318         currentBehavior = newBehavior;
 319         handleBehaviorChange(reusableBehaviorStateHolder, currentBehavior);
 320     }
 321
 322     @Override public void handleCommand(Object message) {
 323         if (message instanceof ApplyState){
 324             ApplyState applyState = (ApplyState) message;
 325
 326             long elapsedTime = (System.nanoTime() - applyState.getStartTime());
 327             if(elapsedTime >= APPLY_STATE_DELAY_THRESHOLD_IN_NANOS){
 328                 LOG.warn("ApplyState took more time than expected. Elapsed Time = {} ms ApplyState = {}",
 329                         TimeUnit.NANOSECONDS.toMillis(elapsedTime), applyState);
 330             }
 331
 332             if(LOG.isDebugEnabled()) {
 333                 LOG.debug("{}: Applying state for log index {} data {}",
 334                     persistenceId(), applyState.getReplicatedLogEntry().getIndex(),
 335                     applyState.getReplicatedLogEntry().getData());
 336             }
 337
 338             applyState(applyState.getClientActor(), applyState.getIdentifier(),
 339                 applyState.getReplicatedLogEntry().getData());
 340
 341         } else if (message instanceof ApplyJournalEntries){
 342             ApplyJournalEntries applyEntries = (ApplyJournalEntries) message;
 343             if(LOG.isDebugEnabled()) {
 344                 LOG.debug("{}: Persisting ApplyLogEntries with index={}", persistenceId(), applyEntries.getToIndex());
 345             }
 346
 347             persistence().persist(applyEntries, APPLY_JOURNAL_ENTRIES_PERSIST_CALLBACK);
 348
 349         } else if(message instanceof ApplySnapshot ) {
 350             Snapshot snapshot = ((ApplySnapshot) message).getSnapshot();
 351
 352             if(LOG.isDebugEnabled()) {
 353                 LOG.debug("{}: ApplySnapshot called on Follower Actor " +
 354                         "snapshotIndex:{}, snapshotTerm:{}", persistenceId(), snapshot.getLastAppliedIndex(),
 355                     snapshot.getLastAppliedTerm()
 356                 );
 357             }
 358
 359             applySnapshot(snapshot.getState());
 360
 361             //clears the followers log, sets the snapshot index to ensure adjusted-index works
 362             replicatedLog = new ReplicatedLogImpl(snapshot);
 363             context.setReplicatedLog(replicatedLog);
 364             context.setLastApplied(snapshot.getLastAppliedIndex());
 365
 366         } else if (message instanceof FindLeader) {
 367             getSender().tell(
 368                 new FindLeaderReply(getLeaderAddress()),
 369                 getSelf()
 370             );
 371
 372         } else if (message instanceof SaveSnapshotSuccess) {
 373             SaveSnapshotSuccess success = (SaveSnapshotSuccess) message;
 374             LOG.info("{}: SaveSnapshotSuccess received for snapshot", persistenceId());
 375
 376             long sequenceNumber = success.metadata().sequenceNr();
 377
 378             commitSnapshot(sequenceNumber);
 379
 380         } else if (message instanceof SaveSnapshotFailure) {
 381             SaveSnapshotFailure saveSnapshotFailure = (SaveSnapshotFailure) message;
 382
 383             LOG.error("{}: SaveSnapshotFailure received for snapshot Cause:",
 384                     persistenceId(), saveSnapshotFailure.cause());
 385
 386             context.getSnapshotManager().rollback();
 387
 388         } else if (message instanceof CaptureSnapshot) {
 389             LOG.debug("{}: CaptureSnapshot received by actor: {}", persistenceId(), message);
 390
 391             context.getSnapshotManager().create(createSnapshotProcedure);
 392
 393         } else if (message instanceof CaptureSnapshotReply) {
 394             handleCaptureSnapshotReply(((CaptureSnapshotReply) message).getSnapshot());
 395         } else if(message instanceof GetOnDemandRaftState) {
 396             onGetOnDemandRaftStats();
 397         } else if (message.equals(COMMIT_SNAPSHOT)) {
 398             commitSnapshot(-1);
 399         } else {
 400             reusableBehaviorStateHolder.init(currentBehavior);
 401
 402             currentBehavior = currentBehavior.handleMessage(getSender(), message);
 403
 404             handleBehaviorChange(reusableBehaviorStateHolder, currentBehavior);
 405         }
 406     }
 407
 408     private void onGetOnDemandRaftStats() {
 409         // Debugging message to retrieve raft stats.
 410
 411         OnDemandRaftState.Builder builder = OnDemandRaftState.builder()
 412                 .commitIndex(context.getCommitIndex())
 413                 .currentTerm(context.getTermInformation().getCurrentTerm())
 414                 .inMemoryJournalDataSize(replicatedLog.dataSize())
 415                 .inMemoryJournalLogSize(replicatedLog.size())
 416                 .isSnapshotCaptureInitiated(context.getSnapshotManager().isCapturing())
 417                 .lastApplied(context.getLastApplied())
 418                 .lastIndex(replicatedLog.lastIndex())
 419                 .lastTerm(replicatedLog.lastTerm())
 420                 .leader(getLeaderId())
 421                 .raftState(currentBehavior.state().toString())
 422                 .replicatedToAllIndex(currentBehavior.getReplicatedToAllIndex())
 423                 .snapshotIndex(replicatedLog.getSnapshotIndex())
 424                 .snapshotTerm(replicatedLog.getSnapshotTerm())
 425                 .votedFor(context.getTermInformation().getVotedFor())
 426                 .peerAddresses(ImmutableMap.copyOf(context.getPeerAddresses()));
 427
 428         ReplicatedLogEntry lastLogEntry = getLastLogEntry();
 429         if (lastLogEntry != null) {
 430             builder.lastLogIndex(lastLogEntry.getIndex());
 431             builder.lastLogTerm(lastLogEntry.getTerm());
 432         }
 433
 434         if(currentBehavior instanceof AbstractLeader) {
 435             AbstractLeader leader = (AbstractLeader)currentBehavior;
 436             Collection<String> followerIds = leader.getFollowerIds();
 437             List<FollowerInfo> followerInfoList = Lists.newArrayListWithCapacity(followerIds.size());
 438             for(String id: followerIds) {
 439                 final FollowerLogInformation info = leader.getFollower(id);
 440                 followerInfoList.add(new FollowerInfo(id, info.getNextIndex(), info.getMatchIndex(),
 441                         info.isFollowerActive(), DurationFormatUtils.formatDurationHMS(info.timeSinceLastActivity())));
 442             }
 443
 444             builder.followerInfoList(followerInfoList);
 445         }
 446
 447         sender().tell(builder.build(), self());
 448
 449     }
 450
 451     private void handleBehaviorChange(BehaviorStateHolder oldBehaviorState, RaftActorBehavior currentBehavior) {
 452         RaftActorBehavior oldBehavior = oldBehaviorState.getBehavior();
 453
 454         if (oldBehavior != currentBehavior){
 455             onStateChanged();
 456         }
 457
 458         String oldBehaviorLeaderId = oldBehavior == null ? null : oldBehaviorState.getLeaderId();
 459         String oldBehaviorStateName = oldBehavior == null ? null : oldBehavior.state().name();
 460
 461         // it can happen that the state has not changed but the leader has changed.
 462         Optional<ActorRef> roleChangeNotifier = getRoleChangeNotifier();
 463         if(!Objects.equal(oldBehaviorLeaderId, currentBehavior.getLeaderId())) {
 464             if(roleChangeNotifier.isPresent()) {
 465                 roleChangeNotifier.get().tell(new LeaderStateChanged(getId(), currentBehavior.getLeaderId()), getSelf());
 466             }
 467
 468             onLeaderChanged(oldBehaviorLeaderId, currentBehavior.getLeaderId());
 469         }
 470
 471         if (roleChangeNotifier.isPresent() &&
 472                 (oldBehavior == null || (oldBehavior.state() != currentBehavior.state()))) {
 473             roleChangeNotifier.get().tell(new RoleChanged(getId(), oldBehaviorStateName ,
 474                     currentBehavior.state().name()), getSelf());
 475         }
 476     }
 477
 478     /**
 479      * When a derived RaftActor needs to persist something it must call
 480      * persistData.
 481      *
 482      * @param clientActor
 483      * @param identifier
 484      * @param data
 485      */
 486     protected void persistData(final ActorRef clientActor, final String identifier,
 487         final Payload data) {
 488
 489         ReplicatedLogEntry replicatedLogEntry = new ReplicatedLogImplEntry(
 490             context.getReplicatedLog().lastIndex() + 1,
 491             context.getTermInformation().getCurrentTerm(), data);
 492
 493         if(LOG.isDebugEnabled()) {
 494             LOG.debug("{}: Persist data {}", persistenceId(), replicatedLogEntry);
 495         }
 496
 497         final RaftActorContext raftContext = getRaftActorContext();
 498
 499         replicatedLog
 500                 .appendAndPersist(replicatedLogEntry, new Procedure<ReplicatedLogEntry>() {
 501                     @Override
 502                     public void apply(ReplicatedLogEntry replicatedLogEntry) throws Exception {
 503                         if(!hasFollowers()){
 504                             // Increment the Commit Index and the Last Applied values
 505                             raftContext.setCommitIndex(replicatedLogEntry.getIndex());
 506                             raftContext.setLastApplied(replicatedLogEntry.getIndex());
 507
 508                             // Apply the state immediately
 509                             applyState(clientActor, identifier, data);
 510
 511                             // Send a ApplyJournalEntries message so that we write the fact that we applied
 512                             // the state to durable storage
 513                             self().tell(new ApplyJournalEntries(replicatedLogEntry.getIndex()), self());
 514
 515                             context.getSnapshotManager().trimLog(context.getLastApplied(), currentBehavior);
 516
 517                         } else if (clientActor != null) {
 518                             // Send message for replication
 519                             currentBehavior.handleMessage(getSelf(),
 520                                     new Replicate(clientActor, identifier,
 521                                             replicatedLogEntry)
 522                             );
 523                         }
 524
 525                     }
 526                 });    }
 527
 528     protected String getId() {
 529         return context.getId();
 530     }
 531
 532     /**
 533      * Derived actors can call the isLeader method to check if the current
 534      * RaftActor is the Leader or not
 535      *
 536      * @return true it this RaftActor is a Leader false otherwise
 537      */
 538     protected boolean isLeader() {
 539         return context.getId().equals(currentBehavior.getLeaderId());
 540     }
 541
 542     /**
 543      * Derived actor can call getLeader if they need a reference to the Leader.
 544      * This would be useful for example in forwarding a request to an actor
 545      * which is the leader
 546      *
 547      * @return A reference to the leader if known, null otherwise
 548      */
 549     protected ActorSelection getLeader(){
 550         String leaderAddress = getLeaderAddress();
 551
 552         if(leaderAddress == null){
 553             return null;
 554         }
 555
 556         return context.actorSelection(leaderAddress);
 557     }
 558
 559     /**
 560      *
 561      * @return the current leader's id
 562      */
 563     protected String getLeaderId(){
 564         return currentBehavior.getLeaderId();
 565     }
 566
 567     protected RaftState getRaftState() {
 568         return currentBehavior.state();
 569     }
 570
 571     protected ReplicatedLogEntry getLastLogEntry() {
 572         return replicatedLog.last();
 573     }
 574
 575     protected Long getCurrentTerm(){
 576         return context.getTermInformation().getCurrentTerm();
 577     }
 578
 579     protected Long getCommitIndex(){
 580         return context.getCommitIndex();
 581     }
 582
 583     protected Long getLastApplied(){
 584         return context.getLastApplied();
 585     }
 586
 587     protected RaftActorContext getRaftActorContext() {
 588         return context;
 589     }
 590
 591     protected void updateConfigParams(ConfigParams configParams) {
 592         context.setConfigParams(configParams);
 593     }
 594
 595     public final DataPersistenceProvider persistence() {
 596         return delegatingPersistenceProvider.getDelegate();
 597     }
 598
 599     public void setPersistence(DataPersistenceProvider provider) {
 600         delegatingPersistenceProvider.setDelegate(provider);
 601     }
 602
 603     protected void setPersistence(boolean persistent) {
 604         if(persistent) {
 605             setPersistence(new PersistentDataProvider(this));
 606         } else {
 607             setPersistence(new NonPersistentDataProvider() {
 608                 /**
 609                  * The way snapshotting works is,
 610                  * <ol>
 611                  * <li> RaftActor calls createSnapshot on the Shard
 612                  * <li> Shard sends a CaptureSnapshotReply and RaftActor then calls saveSnapshot
 613                  * <li> When saveSnapshot is invoked on the akka-persistence API it uses the SnapshotStore to save
 614                  * the snapshot. The SnapshotStore sends SaveSnapshotSuccess or SaveSnapshotFailure. When the
 615                  * RaftActor gets SaveSnapshot success it commits the snapshot to the in-memory journal. This
 616                  * commitSnapshot is mimicking what is done in SaveSnapshotSuccess.
 617                  * </ol>
 618                  */
 619                 @Override
 620                 public void saveSnapshot(Object o) {
 621                     // Make saving Snapshot successful
 622                     // Committing the snapshot here would end up calling commit in the creating state which would
 623                     // be a state violation. That's why now we send a message to commit the snapshot.
 624                     self().tell(COMMIT_SNAPSHOT, self());
 625                 }
 626             });
 627         }
 628     }
 629
 630     /**
 631      * setPeerAddress sets the address of a known peer at a later time.
 632      * <p>
 633      * This is to account for situations where a we know that a peer
 634      * exists but we do not know an address up-front. This may also be used in
 635      * situations where a known peer starts off in a different location and we
 636      * need to change it's address
 637      * <p>
 638      * Note that if the peerId does not match the list of peers passed to
 639      * this actor during construction an IllegalStateException will be thrown.
 640      *
 641      * @param peerId
 642      * @param peerAddress
 643      */
 644     protected void setPeerAddress(String peerId, String peerAddress){
 645         context.setPeerAddress(peerId, peerAddress);
 646     }
 647
 648     protected void commitSnapshot(long sequenceNumber) {
 649         context.getSnapshotManager().commit(persistence(), sequenceNumber);
 650     }
 651
 652     /**
 653      * The applyState method will be called by the RaftActor when some data
 654      * needs to be applied to the actor's state
 655      *
 656      * @param clientActor A reference to the client who sent this message. This
 657      *                    is the same reference that was passed to persistData
 658      *                    by the derived actor. clientActor may be null when
 659      *                    the RaftActor is behaving as a follower or during
 660      *                    recovery.
 661      * @param identifier  The identifier of the persisted data. This is also
 662      *                    the same identifier that was passed to persistData by
 663      *                    the derived actor. identifier may be null when
 664      *                    the RaftActor is behaving as a follower or during
 665      *                    recovery
 666      * @param data        A piece of data that was persisted by the persistData call.
 667      *                    This should NEVER be null.
 668      */
 669     protected abstract void applyState(ActorRef clientActor, String identifier,
 670         Object data);
 671
 672     /**
 673      * This method is called during recovery at the start of a batch of state entries. Derived
 674      * classes should perform any initialization needed to start a batch.
 675      */
 676     protected abstract void startLogRecoveryBatch(int maxBatchSize);
 677
 678     /**
 679      * This method is called during recovery to append state data to the current batch. This method
 680      * is called 1 or more times after {@link #startLogRecoveryBatch}.
 681      *
 682      * @param data the state data
 683      */
 684     protected abstract void appendRecoveredLogEntry(Payload data);
 685
 686     /**
 687      * This method is called during recovery to reconstruct the state of the actor.
 688      *
 689      * @param snapshotBytes A snapshot of the state of the actor
 690      */
 691     protected abstract void applyRecoverySnapshot(byte[] snapshotBytes);
 692
 693     /**
 694      * This method is called during recovery at the end of a batch to apply the current batched
 695      * log entries. This method is called after {@link #appendRecoveredLogEntry}.
 696      */
 697     protected abstract void applyCurrentLogRecoveryBatch();
 698
 699     /**
 700      * This method is called when recovery is complete.
 701      */
 702     protected abstract void onRecoveryComplete();
 703
 704     /**
 705      * This method will be called by the RaftActor when a snapshot needs to be
 706      * created. The derived actor should respond with its current state.
 707      * <p/>
 708      * During recovery the state that is returned by the derived actor will
 709      * be passed back to it by calling the applySnapshot  method
 710      *
 711      * @return The current state of the actor
 712      */
 713     protected abstract void createSnapshot();
 714
 715     /**
 716      * This method can be called at any other point during normal
 717      * operations when the derived actor is out of sync with it's peers
 718      * and the only way to bring it in sync is by applying a snapshot
 719      *
 720      * @param snapshotBytes A snapshot of the state of the actor
 721      */
 722     protected abstract void applySnapshot(byte[] snapshotBytes);
 723
 724     /**
 725      * This method will be called by the RaftActor when the state of the
 726      * RaftActor changes. The derived actor can then use methods like
 727      * isLeader or getLeader to do something useful
 728      */
 729     protected abstract void onStateChanged();
 730
 731     /**
 732      * Notifier Actor for this RaftActor to notify when a role change happens
 733      * @return ActorRef - ActorRef of the notifier or Optional.absent if none.
 734      */
 735     protected abstract Optional<ActorRef> getRoleChangeNotifier();
 736
 737     protected void onLeaderChanged(String oldLeader, String newLeader){};
 738
 739     private String getLeaderAddress(){
 740         if(isLeader()){
 741             return getSelf().path().toString();
 742         }
 743         String leaderId = currentBehavior.getLeaderId();
 744         if (leaderId == null) {
 745             return null;
 746         }
 747         String peerAddress = context.getPeerAddress(leaderId);
 748         if(LOG.isDebugEnabled()) {
 749             LOG.debug("{}: getLeaderAddress leaderId = {} peerAddress = {}",
 750                     persistenceId(), leaderId, peerAddress);
 751         }
 752
 753         return peerAddress;
 754     }
 755
 756     private void handleCaptureSnapshotReply(byte[] snapshotBytes) {
 757         LOG.debug("{}: CaptureSnapshotReply received by actor: snapshot size {}", persistenceId(), snapshotBytes.length);
 758
 759         context.getSnapshotManager().persist(persistence(), snapshotBytes, currentBehavior, getTotalMemory());
 760     }
 761
 762     protected long getTotalMemory() {
 763         return Runtime.getRuntime().totalMemory();
 764     }
 765
 766     protected boolean hasFollowers(){
 767         return getRaftActorContext().getPeerAddresses().keySet().size() > 0;
 768     }
 769
 770     private class ReplicatedLogImpl extends AbstractReplicatedLogImpl {
 771         private static final int DATA_SIZE_DIVIDER = 5;
 772         private long dataSizeSinceLastSnapshot = 0L;
 773
 774
 775         public ReplicatedLogImpl(Snapshot snapshot) {
 776             super(snapshot.getLastAppliedIndex(), snapshot.getLastAppliedTerm(),
 777                 snapshot.getUnAppliedEntries());
 778         }
 779
 780         public ReplicatedLogImpl() {
 781             super();
 782         }
 783
 784         @Override public void removeFromAndPersist(long logEntryIndex) {
 785             int adjustedIndex = adjustedIndex(logEntryIndex);
 786
 787             if (adjustedIndex < 0) {
 788                 return;
 789             }
 790
 791             // FIXME: Maybe this should be done after the command is saved
 792             journal.subList(adjustedIndex , journal.size()).clear();
 793
 794             persistence().persist(new DeleteEntries(adjustedIndex), new Procedure<DeleteEntries>() {
 795
 796                 @Override
 797                 public void apply(DeleteEntries param)
 798                         throws Exception {
 799                     //FIXME : Doing nothing for now
 800                     dataSize = 0;
 801                     for (ReplicatedLogEntry entry : journal) {
 802                         dataSize += entry.size();
 803                     }
 804                 }
 805             });
 806         }
 807
 808         @Override public void appendAndPersist(
 809             final ReplicatedLogEntry replicatedLogEntry) {
 810             appendAndPersist(replicatedLogEntry, null);
 811         }
 812
 813         public void appendAndPersist(
 814             final ReplicatedLogEntry replicatedLogEntry,
 815             final Procedure<ReplicatedLogEntry> callback)  {
 816
 817             if(LOG.isDebugEnabled()) {
 818                 LOG.debug("{}: Append log entry and persist {} ", persistenceId(), replicatedLogEntry);
 819             }
 820
 821             // FIXME : By adding the replicated log entry to the in-memory journal we are not truly ensuring durability of the logs
 822             journal.add(replicatedLogEntry);
 823
 824             // When persisting events with persist it is guaranteed that the
 825             // persistent actor will not receive further commands between the
 826             // persist call and the execution(s) of the associated event
 827             // handler. This also holds for multiple persist calls in context
 828             // of a single command.
 829             persistence().persist(replicatedLogEntry,
 830                 new Procedure<ReplicatedLogEntry>() {
 831                     @Override
 832                     public void apply(ReplicatedLogEntry evt) throws Exception {
 833                         int logEntrySize = replicatedLogEntry.size();
 834
 835                         dataSize += logEntrySize;
 836                         long dataSizeForCheck = dataSize;
 837
 838                         dataSizeSinceLastSnapshot += logEntrySize;
 839
 840                         if (!hasFollowers()) {
 841                             // When we do not have followers we do not maintain an in-memory log
 842                             // due to this the journalSize will never become anything close to the
 843                             // snapshot batch count. In fact will mostly be 1.
 844                             // Similarly since the journal's dataSize depends on the entries in the
 845                             // journal the journal's dataSize will never reach a value close to the
 846                             // memory threshold.
 847                             // By maintaining the dataSize outside the journal we are tracking essentially
 848                             // what we have written to the disk however since we no longer are in
 849                             // need of doing a snapshot just for the sake of freeing up memory we adjust
 850                             // the real size of data by the DATA_SIZE_DIVIDER so that we do not snapshot as often
 851                             // as if we were maintaining a real snapshot
 852                             dataSizeForCheck = dataSizeSinceLastSnapshot / DATA_SIZE_DIVIDER;
 853                         }
 854                         long journalSize = replicatedLogEntry.getIndex() + 1;
 855                         long dataThreshold = getTotalMemory() *
 856                                 context.getConfigParams().getSnapshotDataThresholdPercentage() / 100;
 857
 858                         if ((journalSize % context.getConfigParams().getSnapshotBatchCount() == 0
 859                                 || dataSizeForCheck > dataThreshold)) {
 860
 861                             boolean started = context.getSnapshotManager().capture(replicatedLogEntry,
 862                                     currentBehavior.getReplicatedToAllIndex());
 863
 864                             if(started){
 865                                 dataSizeSinceLastSnapshot = 0;
 866                             }
 867
 868                         }
 869
 870                         if (callback != null){
 871                             callback.apply(replicatedLogEntry);
 872                         }
 873                     }
 874                 }
 875             );
 876         }
 877
 878     }
 879
 880     static class DeleteEntries implements Serializable {
 881         private static final long serialVersionUID = 1L;
 882         private final int fromIndex;
 883
 884         public DeleteEntries(int fromIndex) {
 885             this.fromIndex = fromIndex;
 886         }
 887
 888         public int getFromIndex() {
 889             return fromIndex;
 890         }
 891     }
 892
 893
 894     private class ElectionTermImpl implements ElectionTerm {
 895         /**
 896          * Identifier of the actor whose election term information this is
 897          */
 898         private long currentTerm = 0;
 899         private String votedFor = null;
 900
 901         @Override
 902         public long getCurrentTerm() {
 903             return currentTerm;
 904         }
 905
 906         @Override
 907         public String getVotedFor() {
 908             return votedFor;
 909         }
 910
 911         @Override public void update(long currentTerm, String votedFor) {
 912             if(LOG.isDebugEnabled()) {
 913                 LOG.debug("{}: Set currentTerm={}, votedFor={}", persistenceId(), currentTerm, votedFor);
 914             }
 915             this.currentTerm = currentTerm;
 916             this.votedFor = votedFor;
 917         }
 918
 919         @Override
 920         public void updateAndPersist(long currentTerm, String votedFor){
 921             update(currentTerm, votedFor);
 922             // FIXME : Maybe first persist then update the state
 923             persistence().persist(new UpdateElectionTerm(this.currentTerm, this.votedFor), new Procedure<UpdateElectionTerm>(){
 924
 925                 @Override public void apply(UpdateElectionTerm param)
 926                     throws Exception {
 927
 928                 }
 929             });
 930         }
 931     }
 932
 933     static class UpdateElectionTerm implements Serializable {
 934         private static final long serialVersionUID = 1L;
 935         private final long currentTerm;
 936         private final String votedFor;
 937
 938         public UpdateElectionTerm(long currentTerm, String votedFor) {
 939             this.currentTerm = currentTerm;
 940             this.votedFor = votedFor;
 941         }
 942
 943         public long getCurrentTerm() {
 944             return currentTerm;
 945         }
 946
 947         public String getVotedFor() {
 948             return votedFor;
 949         }
 950     }
 951
 952     private class CreateSnapshotProcedure implements Procedure<Void> {
 953
 954         @Override
 955         public void apply(Void aVoid) throws Exception {
 956             createSnapshot();
 957         }
 958     }
 959
 960     @VisibleForTesting
 961     void setCurrentBehavior(AbstractRaftActorBehavior behavior) {
 962         currentBehavior = behavior;
 963     }
 964
 965     protected RaftActorBehavior getCurrentBehavior() {
 966         return currentBehavior;
 967     }
 968
 969     private static class BehaviorStateHolder {
 970         private RaftActorBehavior behavior;
 971         private String leaderId;
 972
 973         void init(RaftActorBehavior behavior) {
 974             this.behavior = behavior;
 975             this.leaderId = behavior != null ? behavior.getLeaderId() : null;
 976         }
 977
 978         RaftActorBehavior getBehavior() {
 979             return behavior;
 980         }
 981
 982         String getLeaderId() {
 983             return leaderId;
 984         }
 985     }
 986 }