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.event.Logging;
  14 import akka.event.LoggingAdapter;
  15 import akka.japi.Procedure;
  16 import akka.persistence.RecoveryCompleted;
  17 import akka.persistence.SaveSnapshotFailure;
  18 import akka.persistence.SaveSnapshotSuccess;
  19 import akka.persistence.SnapshotOffer;
  20 import akka.persistence.SnapshotSelectionCriteria;
  21 import com.google.common.annotations.VisibleForTesting;
  22 import com.google.common.base.Optional;
  23 import com.google.common.base.Stopwatch;
  24 import com.google.protobuf.ByteString;
  25 import java.io.Serializable;
  26 import java.util.Map;
  27 import org.opendaylight.controller.cluster.DataPersistenceProvider;
  28 import org.opendaylight.controller.cluster.common.actor.AbstractUntypedPersistentActor;
  29 import org.opendaylight.controller.cluster.notifications.RoleChanged;
  30 import org.opendaylight.controller.cluster.raft.base.messages.ApplyLogEntries;
  31 import org.opendaylight.controller.cluster.raft.base.messages.ApplySnapshot;
  32 import org.opendaylight.controller.cluster.raft.base.messages.ApplyState;
  33 import org.opendaylight.controller.cluster.raft.base.messages.CaptureSnapshot;
  34 import org.opendaylight.controller.cluster.raft.base.messages.CaptureSnapshotReply;
  35 import org.opendaylight.controller.cluster.raft.base.messages.Replicate;
  36 import org.opendaylight.controller.cluster.raft.base.messages.SendHeartBeat;
  37 import org.opendaylight.controller.cluster.raft.base.messages.SendInstallSnapshot;
  38 import org.opendaylight.controller.cluster.raft.behaviors.AbstractRaftActorBehavior;
  39 import org.opendaylight.controller.cluster.raft.behaviors.Follower;
  40 import org.opendaylight.controller.cluster.raft.behaviors.RaftActorBehavior;
  41 import org.opendaylight.controller.cluster.raft.client.messages.FindLeader;
  42 import org.opendaylight.controller.cluster.raft.client.messages.FindLeaderReply;
  43 import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
  44 import org.opendaylight.controller.cluster.raft.protobuff.client.messages.Payload;
  45 import org.opendaylight.controller.protobuff.messages.cluster.raft.AppendEntriesMessages;
  46
  47 /**
  48  * RaftActor encapsulates a state machine that needs to be kept synchronized
  49  * in a cluster. It implements the RAFT algorithm as described in the paper
  50  * <a href='https://ramcloud.stanford.edu/wiki/download/attachments/11370504/raft.pdf'>
  51  * In Search of an Understandable Consensus Algorithm</a>
  52  * <p/>
  53  * RaftActor has 3 states and each state has a certain behavior associated
  54  * with it. A Raft actor can behave as,
  55  * <ul>
  56  * <li> A Leader </li>
  57  * <li> A Follower (or) </li>
  58  * <li> A Candidate </li>
  59  * </ul>
  60  * <p/>
  61  * <p/>
  62  * A RaftActor MUST be a Leader in order to accept requests from clients to
  63  * change the state of it's encapsulated state machine. Once a RaftActor becomes
  64  * a Leader it is also responsible for ensuring that all followers ultimately
  65  * have the same log and therefore the same state machine as itself.
  66  * <p/>
  67  * <p/>
  68  * The current behavior of a RaftActor determines how election for leadership
  69  * is initiated and how peer RaftActors react to request for votes.
  70  * <p/>
  71  * <p/>
  72  * Each RaftActor also needs to know the current election term. It uses this
  73  * information for a couple of things. One is to simply figure out who it
  74  * voted for in the last election. Another is to figure out if the message
  75  * it received to update it's state is stale.
  76  * <p/>
  77  * <p/>
  78  * The RaftActor uses akka-persistence to store it's replicated log.
  79  * Furthermore through it's behaviors a Raft Actor determines
  80  * <p/>
  81  * <ul>
  82  * <li> when a log entry should be persisted </li>
  83  * <li> when a log entry should be applied to the state machine (and) </li>
  84  * <li> when a snapshot should be saved </li>
  85  * </ul>
  86  */
  87 public abstract class RaftActor extends AbstractUntypedPersistentActor {
  88     protected final LoggingAdapter LOG =
  89         Logging.getLogger(getContext().system(), this);
  90
  91     /**
  92      * The current state determines the current behavior of a RaftActor
  93      * A Raft Actor always starts off in the Follower State
  94      */
  95     private RaftActorBehavior currentBehavior;
  96
  97     /**
  98      * This context should NOT be passed directly to any other actor it is
  99      * only to be consumed by the RaftActorBehaviors
 100      */
 101     private final RaftActorContext context;
 102
 103     /**
 104      * The in-memory journal
 105      */
 106     private ReplicatedLogImpl replicatedLog = new ReplicatedLogImpl();
 107
 108     private CaptureSnapshot captureSnapshot = null;
 109
 110     private volatile boolean hasSnapshotCaptureInitiated = false;
 111
 112     private Stopwatch recoveryTimer;
 113
 114     private int currentRecoveryBatchCount;
 115
 116
 117
 118     public RaftActor(String id, Map<String, String> peerAddresses) {
 119         this(id, peerAddresses, Optional.<ConfigParams>absent());
 120     }
 121
 122     public RaftActor(String id, Map<String, String> peerAddresses,
 123          Optional<ConfigParams> configParams) {
 124
 125         context = new RaftActorContextImpl(this.getSelf(),
 126             this.getContext(), id, new ElectionTermImpl(),
 127             -1, -1, replicatedLog, peerAddresses,
 128             (configParams.isPresent() ? configParams.get(): new DefaultConfigParamsImpl()),
 129             LOG);
 130     }
 131
 132     private void initRecoveryTimer() {
 133         if(recoveryTimer == null) {
 134             recoveryTimer = new Stopwatch();
 135             recoveryTimer.start();
 136         }
 137     }
 138
 139     @Override
 140     public void preStart() throws Exception {
 141         LOG.info("Starting recovery for {} with journal batch size {}", persistenceId(),
 142                 context.getConfigParams().getJournalRecoveryLogBatchSize());
 143
 144         super.preStart();
 145     }
 146
 147     @Override
 148     public void handleRecover(Object message) {
 149         if(persistence().isRecoveryApplicable()) {
 150             if (message instanceof SnapshotOffer) {
 151                 onRecoveredSnapshot((SnapshotOffer) message);
 152             } else if (message instanceof ReplicatedLogEntry) {
 153                 onRecoveredJournalLogEntry((ReplicatedLogEntry) message);
 154             } else if (message instanceof ApplyLogEntries) {
 155                 onRecoveredApplyLogEntries((ApplyLogEntries) message);
 156             } else if (message instanceof DeleteEntries) {
 157                 replicatedLog.removeFrom(((DeleteEntries) message).getFromIndex());
 158             } else if (message instanceof UpdateElectionTerm) {
 159                 context.getTermInformation().update(((UpdateElectionTerm) message).getCurrentTerm(),
 160                         ((UpdateElectionTerm) message).getVotedFor());
 161             } else if (message instanceof RecoveryCompleted) {
 162                 onRecoveryCompletedMessage();
 163             }
 164         } else {
 165             if (message instanceof RecoveryCompleted) {
 166                 // Delete all the messages from the akka journal so that we do not end up with consistency issues
 167                 // Note I am not using the dataPersistenceProvider and directly using the akka api here
 168                 deleteMessages(lastSequenceNr());
 169
 170                 // Delete all the akka snapshots as they will not be needed
 171                 deleteSnapshots(new SnapshotSelectionCriteria(scala.Long.MaxValue(), scala.Long.MaxValue()));
 172
 173                 onRecoveryComplete();
 174
 175                 RaftActorBehavior oldBehavior = currentBehavior;
 176                 currentBehavior = new Follower(context);
 177                 handleBehaviorChange(oldBehavior, currentBehavior);
 178             }
 179         }
 180     }
 181
 182     private void onRecoveredSnapshot(SnapshotOffer offer) {
 183         if(LOG.isDebugEnabled()) {
 184             LOG.debug("SnapshotOffer called..");
 185         }
 186
 187         initRecoveryTimer();
 188
 189         Snapshot snapshot = (Snapshot) offer.snapshot();
 190
 191         // Create a replicated log with the snapshot information
 192         // The replicated log can be used later on to retrieve this snapshot
 193         // when we need to install it on a peer
 194         replicatedLog = new ReplicatedLogImpl(snapshot);
 195
 196         context.setReplicatedLog(replicatedLog);
 197         context.setLastApplied(snapshot.getLastAppliedIndex());
 198         context.setCommitIndex(snapshot.getLastAppliedIndex());
 199
 200         Stopwatch timer = new Stopwatch();
 201         timer.start();
 202
 203         // Apply the snapshot to the actors state
 204         applyRecoverySnapshot(ByteString.copyFrom(snapshot.getState()));
 205
 206         timer.stop();
 207         LOG.info("Recovery snapshot applied for {} in {}: snapshotIndex={}, snapshotTerm={}, journal-size=" +
 208                 replicatedLog.size(), persistenceId(), timer.toString(),
 209                 replicatedLog.snapshotIndex, replicatedLog.snapshotTerm);
 210     }
 211
 212     private void onRecoveredJournalLogEntry(ReplicatedLogEntry logEntry) {
 213         if(LOG.isDebugEnabled()) {
 214             LOG.debug("Received ReplicatedLogEntry for recovery: {}", logEntry.getIndex());
 215         }
 216
 217         replicatedLog.append(logEntry);
 218     }
 219
 220     private void onRecoveredApplyLogEntries(ApplyLogEntries ale) {
 221         if(LOG.isDebugEnabled()) {
 222             LOG.debug("Received ApplyLogEntries for recovery, applying to state: {} to {}",
 223                     context.getLastApplied() + 1, ale.getToIndex());
 224         }
 225
 226         for (long i = context.getLastApplied() + 1; i <= ale.getToIndex(); i++) {
 227             batchRecoveredLogEntry(replicatedLog.get(i));
 228         }
 229
 230         context.setLastApplied(ale.getToIndex());
 231         context.setCommitIndex(ale.getToIndex());
 232     }
 233
 234     private void batchRecoveredLogEntry(ReplicatedLogEntry logEntry) {
 235         initRecoveryTimer();
 236
 237         int batchSize = context.getConfigParams().getJournalRecoveryLogBatchSize();
 238         if(currentRecoveryBatchCount == 0) {
 239             startLogRecoveryBatch(batchSize);
 240         }
 241
 242         appendRecoveredLogEntry(logEntry.getData());
 243
 244         if(++currentRecoveryBatchCount >= batchSize) {
 245             endCurrentLogRecoveryBatch();
 246         }
 247     }
 248
 249     private void endCurrentLogRecoveryBatch() {
 250         applyCurrentLogRecoveryBatch();
 251         currentRecoveryBatchCount = 0;
 252     }
 253
 254     private void onRecoveryCompletedMessage() {
 255         if(currentRecoveryBatchCount > 0) {
 256             endCurrentLogRecoveryBatch();
 257         }
 258
 259         onRecoveryComplete();
 260
 261         String recoveryTime = "";
 262         if(recoveryTimer != null) {
 263             recoveryTimer.stop();
 264             recoveryTime = " in " + recoveryTimer.toString();
 265             recoveryTimer = null;
 266         }
 267
 268         LOG.info(
 269             "Recovery completed" + recoveryTime + " - Switching actor to Follower - " +
 270                 "Persistence Id =  " + persistenceId() +
 271                 " Last index in log={}, snapshotIndex={}, snapshotTerm={}, " +
 272                 "journal-size={}",
 273             replicatedLog.lastIndex(), replicatedLog.snapshotIndex,
 274             replicatedLog.snapshotTerm, replicatedLog.size());
 275
 276         RaftActorBehavior oldBehavior = currentBehavior;
 277         currentBehavior = new Follower(context);
 278         handleBehaviorChange(oldBehavior, currentBehavior);
 279     }
 280
 281     @Override public void handleCommand(Object message) {
 282         if (message instanceof ApplyState){
 283             ApplyState applyState = (ApplyState) message;
 284
 285             if(LOG.isDebugEnabled()) {
 286                 LOG.debug("Applying state for log index {} data {}",
 287                     applyState.getReplicatedLogEntry().getIndex(),
 288                     applyState.getReplicatedLogEntry().getData());
 289             }
 290
 291             applyState(applyState.getClientActor(), applyState.getIdentifier(),
 292                 applyState.getReplicatedLogEntry().getData());
 293
 294         } else if (message instanceof ApplyLogEntries){
 295             ApplyLogEntries ale = (ApplyLogEntries) message;
 296             if(LOG.isDebugEnabled()) {
 297                 LOG.debug("Persisting ApplyLogEntries with index={}", ale.getToIndex());
 298             }
 299             persistence().persist(new ApplyLogEntries(ale.getToIndex()), new Procedure<ApplyLogEntries>() {
 300                 @Override
 301                 public void apply(ApplyLogEntries param) throws Exception {
 302                 }
 303             });
 304
 305         } else if(message instanceof ApplySnapshot ) {
 306             Snapshot snapshot = ((ApplySnapshot) message).getSnapshot();
 307
 308             if(LOG.isDebugEnabled()) {
 309                 LOG.debug("ApplySnapshot called on Follower Actor " +
 310                         "snapshotIndex:{}, snapshotTerm:{}", snapshot.getLastAppliedIndex(),
 311                     snapshot.getLastAppliedTerm()
 312                 );
 313             }
 314             applySnapshot(ByteString.copyFrom(snapshot.getState()));
 315
 316             //clears the followers log, sets the snapshot index to ensure adjusted-index works
 317             replicatedLog = new ReplicatedLogImpl(snapshot);
 318             context.setReplicatedLog(replicatedLog);
 319             context.setLastApplied(snapshot.getLastAppliedIndex());
 320
 321         } else if (message instanceof FindLeader) {
 322             getSender().tell(
 323                 new FindLeaderReply(getLeaderAddress()),
 324                 getSelf()
 325             );
 326
 327         } else if (message instanceof SaveSnapshotSuccess) {
 328             SaveSnapshotSuccess success = (SaveSnapshotSuccess) message;
 329             LOG.info("SaveSnapshotSuccess received for snapshot");
 330
 331             long sequenceNumber = success.metadata().sequenceNr();
 332
 333             commitSnapshot(sequenceNumber);
 334
 335         } else if (message instanceof SaveSnapshotFailure) {
 336             SaveSnapshotFailure saveSnapshotFailure = (SaveSnapshotFailure) message;
 337
 338             LOG.info("saveSnapshotFailure.metadata():{}", saveSnapshotFailure.metadata().toString());
 339             LOG.error(saveSnapshotFailure.cause(), "SaveSnapshotFailure received for snapshot Cause:");
 340
 341             context.getReplicatedLog().snapshotRollback();
 342
 343             LOG.info("Replicated Log rollbacked. Snapshot will be attempted in the next cycle." +
 344                 "snapshotIndex:{}, snapshotTerm:{}, log-size:{}",
 345                 context.getReplicatedLog().getSnapshotIndex(),
 346                 context.getReplicatedLog().getSnapshotTerm(),
 347                 context.getReplicatedLog().size());
 348
 349         } else if (message instanceof CaptureSnapshot) {
 350             LOG.info("CaptureSnapshot received by actor");
 351             CaptureSnapshot cs = (CaptureSnapshot)message;
 352             captureSnapshot = cs;
 353             createSnapshot();
 354
 355         } else if (message instanceof CaptureSnapshotReply){
 356             LOG.info("CaptureSnapshotReply received by actor");
 357             CaptureSnapshotReply csr = (CaptureSnapshotReply) message;
 358
 359             ByteString stateInBytes = csr.getSnapshot();
 360             LOG.info("CaptureSnapshotReply stateInBytes size:{}", stateInBytes.size());
 361             handleCaptureSnapshotReply(stateInBytes);
 362
 363         } else {
 364             if (!(message instanceof AppendEntriesMessages.AppendEntries)
 365                 && !(message instanceof AppendEntriesReply) && !(message instanceof SendHeartBeat)) {
 366                 if(LOG.isDebugEnabled()) {
 367                     LOG.debug("onReceiveCommand: message: {}", message.getClass());
 368                 }
 369             }
 370
 371             RaftActorBehavior oldBehavior = currentBehavior;
 372             currentBehavior = currentBehavior.handleMessage(getSender(), message);
 373
 374             handleBehaviorChange(oldBehavior, currentBehavior);
 375         }
 376     }
 377
 378     private void handleBehaviorChange(RaftActorBehavior oldBehavior, RaftActorBehavior currentBehavior) {
 379         if (oldBehavior != currentBehavior){
 380             onStateChanged();
 381         }
 382         if (oldBehavior != null) {
 383             // it can happen that the state has not changed but the leader has changed.
 384             onLeaderChanged(oldBehavior.getLeaderId(), currentBehavior.getLeaderId());
 385
 386             if (getRoleChangeNotifier().isPresent() && oldBehavior.state() != currentBehavior.state()) {
 387                 // we do not want to notify when the behavior/role is set for the first time (i.e follower)
 388                 getRoleChangeNotifier().get().tell(new RoleChanged(getId(), oldBehavior.state().name(),
 389                     currentBehavior.state().name()), getSelf());
 390             }
 391         }
 392     }
 393
 394     /**
 395      * When a derived RaftActor needs to persist something it must call
 396      * persistData.
 397      *
 398      * @param clientActor
 399      * @param identifier
 400      * @param data
 401      */
 402     protected void persistData(final ActorRef clientActor, final String identifier,
 403         final Payload data) {
 404
 405         ReplicatedLogEntry replicatedLogEntry = new ReplicatedLogImplEntry(
 406             context.getReplicatedLog().lastIndex() + 1,
 407             context.getTermInformation().getCurrentTerm(), data);
 408
 409         if(LOG.isDebugEnabled()) {
 410             LOG.debug("Persist data {}", replicatedLogEntry);
 411         }
 412
 413         final RaftActorContext raftContext = getRaftActorContext();
 414
 415         replicatedLog
 416                 .appendAndPersist(replicatedLogEntry, new Procedure<ReplicatedLogEntry>() {
 417                     @Override
 418                     public void apply(ReplicatedLogEntry replicatedLogEntry) throws Exception {
 419                         if(!hasFollowers()){
 420                             // Increment the Commit Index and the Last Applied values
 421                             raftContext.setCommitIndex(replicatedLogEntry.getIndex());
 422                             raftContext.setLastApplied(replicatedLogEntry.getIndex());
 423
 424                             // Apply the state immediately
 425                             applyState(clientActor, identifier, data);
 426
 427                             // Send a ApplyLogEntries message so that we write the fact that we applied
 428                             // the state to durable storage
 429                             self().tell(new ApplyLogEntries((int) replicatedLogEntry.getIndex()), self());
 430
 431                             // Check if the "real" snapshot capture has been initiated. If no then do the fake snapshot
 432                             if(!hasSnapshotCaptureInitiated){
 433                                 raftContext.getReplicatedLog().snapshotPreCommit(raftContext.getLastApplied(),
 434                                         raftContext.getTermInformation().getCurrentTerm());
 435                                 raftContext.getReplicatedLog().snapshotCommit();
 436                             } else {
 437                                 LOG.debug("Skipping fake snapshotting for {} because real snapshotting is in progress", getId());
 438                             }
 439                         } else if (clientActor != null) {
 440                             // Send message for replication
 441                             currentBehavior.handleMessage(getSelf(),
 442                                     new Replicate(clientActor, identifier,
 443                                             replicatedLogEntry)
 444                             );
 445                         }
 446
 447                     }
 448                 });    }
 449
 450     protected String getId() {
 451         return context.getId();
 452     }
 453
 454     /**
 455      * Derived actors can call the isLeader method to check if the current
 456      * RaftActor is the Leader or not
 457      *
 458      * @return true it this RaftActor is a Leader false otherwise
 459      */
 460     protected boolean isLeader() {
 461         return context.getId().equals(currentBehavior.getLeaderId());
 462     }
 463
 464     /**
 465      * Derived actor can call getLeader if they need a reference to the Leader.
 466      * This would be useful for example in forwarding a request to an actor
 467      * which is the leader
 468      *
 469      * @return A reference to the leader if known, null otherwise
 470      */
 471     protected ActorSelection getLeader(){
 472         String leaderAddress = getLeaderAddress();
 473
 474         if(leaderAddress == null){
 475             return null;
 476         }
 477
 478         return context.actorSelection(leaderAddress);
 479     }
 480
 481     /**
 482      *
 483      * @return the current leader's id
 484      */
 485     protected String getLeaderId(){
 486         return currentBehavior.getLeaderId();
 487     }
 488
 489     protected RaftState getRaftState() {
 490         return currentBehavior.state();
 491     }
 492
 493     protected ReplicatedLogEntry getLastLogEntry() {
 494         return replicatedLog.last();
 495     }
 496
 497     protected Long getCurrentTerm(){
 498         return context.getTermInformation().getCurrentTerm();
 499     }
 500
 501     protected Long getCommitIndex(){
 502         return context.getCommitIndex();
 503     }
 504
 505     protected Long getLastApplied(){
 506         return context.getLastApplied();
 507     }
 508
 509     protected RaftActorContext getRaftActorContext() {
 510         return context;
 511     }
 512
 513     /**
 514      * setPeerAddress sets the address of a known peer at a later time.
 515      * <p>
 516      * This is to account for situations where a we know that a peer
 517      * exists but we do not know an address up-front. This may also be used in
 518      * situations where a known peer starts off in a different location and we
 519      * need to change it's address
 520      * <p>
 521      * Note that if the peerId does not match the list of peers passed to
 522      * this actor during construction an IllegalStateException will be thrown.
 523      *
 524      * @param peerId
 525      * @param peerAddress
 526      */
 527     protected void setPeerAddress(String peerId, String peerAddress){
 528         context.setPeerAddress(peerId, peerAddress);
 529     }
 530
 531     protected void commitSnapshot(long sequenceNumber) {
 532         context.getReplicatedLog().snapshotCommit();
 533
 534         // TODO: Not sure if we want to be this aggressive with trimming stuff
 535         trimPersistentData(sequenceNumber);
 536     }
 537
 538     /**
 539      * The applyState method will be called by the RaftActor when some data
 540      * needs to be applied to the actor's state
 541      *
 542      * @param clientActor A reference to the client who sent this message. This
 543      *                    is the same reference that was passed to persistData
 544      *                    by the derived actor. clientActor may be null when
 545      *                    the RaftActor is behaving as a follower or during
 546      *                    recovery.
 547      * @param identifier  The identifier of the persisted data. This is also
 548      *                    the same identifier that was passed to persistData by
 549      *                    the derived actor. identifier may be null when
 550      *                    the RaftActor is behaving as a follower or during
 551      *                    recovery
 552      * @param data        A piece of data that was persisted by the persistData call.
 553      *                    This should NEVER be null.
 554      */
 555     protected abstract void applyState(ActorRef clientActor, String identifier,
 556         Object data);
 557
 558     /**
 559      * This method is called during recovery at the start of a batch of state entries. Derived
 560      * classes should perform any initialization needed to start a batch.
 561      */
 562     protected abstract void startLogRecoveryBatch(int maxBatchSize);
 563
 564     /**
 565      * This method is called during recovery to append state data to the current batch. This method
 566      * is called 1 or more times after {@link #startLogRecoveryBatch}.
 567      *
 568      * @param data the state data
 569      */
 570     protected abstract void appendRecoveredLogEntry(Payload data);
 571
 572     /**
 573      * This method is called during recovery to reconstruct the state of the actor.
 574      *
 575      * @param snapshot A snapshot of the state of the actor
 576      */
 577     protected abstract void applyRecoverySnapshot(ByteString snapshot);
 578
 579     /**
 580      * This method is called during recovery at the end of a batch to apply the current batched
 581      * log entries. This method is called after {@link #appendRecoveredLogEntry}.
 582      */
 583     protected abstract void applyCurrentLogRecoveryBatch();
 584
 585     /**
 586      * This method is called when recovery is complete.
 587      */
 588     protected abstract void onRecoveryComplete();
 589
 590     /**
 591      * This method will be called by the RaftActor when a snapshot needs to be
 592      * created. The derived actor should respond with its current state.
 593      * <p/>
 594      * During recovery the state that is returned by the derived actor will
 595      * be passed back to it by calling the applySnapshot  method
 596      *
 597      * @return The current state of the actor
 598      */
 599     protected abstract void createSnapshot();
 600
 601     /**
 602      * This method can be called at any other point during normal
 603      * operations when the derived actor is out of sync with it's peers
 604      * and the only way to bring it in sync is by applying a snapshot
 605      *
 606      * @param snapshot A snapshot of the state of the actor
 607      */
 608     protected abstract void applySnapshot(ByteString snapshot);
 609
 610     /**
 611      * This method will be called by the RaftActor when the state of the
 612      * RaftActor changes. The derived actor can then use methods like
 613      * isLeader or getLeader to do something useful
 614      */
 615     protected abstract void onStateChanged();
 616
 617     protected abstract DataPersistenceProvider persistence();
 618
 619     /**
 620      * Notifier Actor for this RaftActor to notify when a role change happens
 621      * @return ActorRef - ActorRef of the notifier or Optional.absent if none.
 622      */
 623     protected abstract Optional<ActorRef> getRoleChangeNotifier();
 624
 625     protected void onLeaderChanged(String oldLeader, String newLeader){};
 626
 627     private void trimPersistentData(long sequenceNumber) {
 628         // Trim akka snapshots
 629         // FIXME : Not sure how exactly the SnapshotSelectionCriteria is applied
 630         // For now guessing that it is ANDed.
 631         persistence().deleteSnapshots(new SnapshotSelectionCriteria(
 632             sequenceNumber - context.getConfigParams().getSnapshotBatchCount(), 43200000));
 633
 634         // Trim akka journal
 635         persistence().deleteMessages(sequenceNumber);
 636     }
 637
 638     private String getLeaderAddress(){
 639         if(isLeader()){
 640             return getSelf().path().toString();
 641         }
 642         String leaderId = currentBehavior.getLeaderId();
 643         if (leaderId == null) {
 644             return null;
 645         }
 646         String peerAddress = context.getPeerAddress(leaderId);
 647         if(LOG.isDebugEnabled()) {
 648             LOG.debug("getLeaderAddress leaderId = {} peerAddress = {}",
 649                     leaderId, peerAddress);
 650         }
 651
 652         return peerAddress;
 653     }
 654
 655     private void handleCaptureSnapshotReply(ByteString stateInBytes) {
 656         // create a snapshot object from the state provided and save it
 657         // when snapshot is saved async, SaveSnapshotSuccess is raised.
 658
 659         Snapshot sn = Snapshot.create(stateInBytes.toByteArray(),
 660             context.getReplicatedLog().getFrom(captureSnapshot.getLastAppliedIndex() + 1),
 661             captureSnapshot.getLastIndex(), captureSnapshot.getLastTerm(),
 662             captureSnapshot.getLastAppliedIndex(), captureSnapshot.getLastAppliedTerm());
 663
 664         persistence().saveSnapshot(sn);
 665
 666         LOG.info("Persisting of snapshot done:{}", sn.getLogMessage());
 667
 668         //be greedy and remove entries from in-mem journal which are in the snapshot
 669         // and update snapshotIndex and snapshotTerm without waiting for the success,
 670
 671         context.getReplicatedLog().snapshotPreCommit(
 672             captureSnapshot.getLastAppliedIndex(),
 673             captureSnapshot.getLastAppliedTerm());
 674
 675         LOG.info("Removed in-memory snapshotted entries, adjusted snaphsotIndex:{} " +
 676             "and term:{}", captureSnapshot.getLastAppliedIndex(),
 677             captureSnapshot.getLastAppliedTerm());
 678
 679         if (isLeader() && captureSnapshot.isInstallSnapshotInitiated()) {
 680             // this would be call straight to the leader and won't initiate in serialization
 681             currentBehavior.handleMessage(getSelf(), new SendInstallSnapshot(stateInBytes));
 682         }
 683
 684         captureSnapshot = null;
 685         hasSnapshotCaptureInitiated = false;
 686     }
 687
 688     protected boolean hasFollowers(){
 689         return getRaftActorContext().getPeerAddresses().keySet().size() > 0;
 690     }
 691
 692     private class ReplicatedLogImpl extends AbstractReplicatedLogImpl {
 693
 694         private static final int DATA_SIZE_DIVIDER = 5;
 695         private long dataSizeSinceLastSnapshot = 0;
 696
 697         public ReplicatedLogImpl(Snapshot snapshot) {
 698             super(snapshot.getLastAppliedIndex(), snapshot.getLastAppliedTerm(),
 699                 snapshot.getUnAppliedEntries());
 700         }
 701
 702         public ReplicatedLogImpl() {
 703             super();
 704         }
 705
 706         @Override public void removeFromAndPersist(long logEntryIndex) {
 707             int adjustedIndex = adjustedIndex(logEntryIndex);
 708
 709             if (adjustedIndex < 0) {
 710                 return;
 711             }
 712
 713             // FIXME: Maybe this should be done after the command is saved
 714             journal.subList(adjustedIndex , journal.size()).clear();
 715
 716             persistence().persist(new DeleteEntries(adjustedIndex), new Procedure<DeleteEntries>(){
 717
 718                 @Override public void apply(DeleteEntries param)
 719                     throws Exception {
 720                     //FIXME : Doing nothing for now
 721                     dataSize = 0;
 722                     for(ReplicatedLogEntry entry : journal){
 723                         dataSize += entry.size();
 724                     }
 725                 }
 726             });
 727         }
 728
 729         @Override public void appendAndPersist(
 730             final ReplicatedLogEntry replicatedLogEntry) {
 731             appendAndPersist(replicatedLogEntry, null);
 732         }
 733
 734         @Override
 735         public int dataSize() {
 736             return dataSize;
 737         }
 738
 739         public void appendAndPersist(
 740             final ReplicatedLogEntry replicatedLogEntry,
 741             final Procedure<ReplicatedLogEntry> callback)  {
 742
 743             if(LOG.isDebugEnabled()) {
 744                 LOG.debug("Append log entry and persist {} ", replicatedLogEntry);
 745             }
 746
 747             // FIXME : By adding the replicated log entry to the in-memory journal we are not truly ensuring durability of the logs
 748             journal.add(replicatedLogEntry);
 749
 750             // When persisting events with persist it is guaranteed that the
 751             // persistent actor will not receive further commands between the
 752             // persist call and the execution(s) of the associated event
 753             // handler. This also holds for multiple persist calls in context
 754             // of a single command.
 755             persistence().persist(replicatedLogEntry,
 756                 new Procedure<ReplicatedLogEntry>() {
 757                     @Override
 758                     public void apply(ReplicatedLogEntry evt) throws Exception {
 759                         int logEntrySize = replicatedLogEntry.size();
 760
 761                         dataSize += logEntrySize;
 762                         long dataSizeForCheck = dataSize;
 763
 764                         dataSizeSinceLastSnapshot += logEntrySize;
 765                         long journalSize = lastIndex()+1;
 766
 767                         if(!hasFollowers()) {
 768                             // When we do not have followers we do not maintain an in-memory log
 769                             // due to this the journalSize will never become anything close to the
 770                             // snapshot batch count. In fact will mostly be 1.
 771                             // Similarly since the journal's dataSize depends on the entries in the
 772                             // journal the journal's dataSize will never reach a value close to the
 773                             // memory threshold.
 774                             // By maintaining the dataSize outside the journal we are tracking essentially
 775                             // what we have written to the disk however since we no longer are in
 776                             // need of doing a snapshot just for the sake of freeing up memory we adjust
 777                             // the real size of data by the DATA_SIZE_DIVIDER so that we do not snapshot as often
 778                             // as if we were maintaining a real snapshot
 779                             dataSizeForCheck = dataSizeSinceLastSnapshot / DATA_SIZE_DIVIDER;
 780                         }
 781
 782                         long dataThreshold = Runtime.getRuntime().totalMemory() *
 783                                 getRaftActorContext().getConfigParams().getSnapshotDataThresholdPercentage() / 100;
 784
 785                         // when a snaphsot is being taken, captureSnapshot != null
 786                         if (hasSnapshotCaptureInitiated == false &&
 787                                 ( journalSize % context.getConfigParams().getSnapshotBatchCount() == 0 ||
 788                                         dataSizeForCheck > dataThreshold)) {
 789
 790                             dataSizeSinceLastSnapshot = 0;
 791
 792                             LOG.info("Initiating Snapshot Capture..");
 793                             long lastAppliedIndex = -1;
 794                             long lastAppliedTerm = -1;
 795
 796                             ReplicatedLogEntry lastAppliedEntry = get(context.getLastApplied());
 797                             if (!hasFollowers()) {
 798                                 lastAppliedIndex = replicatedLogEntry.getIndex();
 799                                 lastAppliedTerm = replicatedLogEntry.getTerm();
 800                             } else if (lastAppliedEntry != null) {
 801                                 lastAppliedIndex = lastAppliedEntry.getIndex();
 802                                 lastAppliedTerm = lastAppliedEntry.getTerm();
 803                             }
 804
 805                             if(LOG.isDebugEnabled()) {
 806                                 LOG.debug("Snapshot Capture logSize: {}", journal.size());
 807                                 LOG.debug("Snapshot Capture lastApplied:{} ",
 808                                     context.getLastApplied());
 809                                 LOG.debug("Snapshot Capture lastAppliedIndex:{}", lastAppliedIndex);
 810                                 LOG.debug("Snapshot Capture lastAppliedTerm:{}", lastAppliedTerm);
 811                             }
 812
 813                             // send a CaptureSnapshot to self to make the expensive operation async.
 814                             getSelf().tell(new CaptureSnapshot(
 815                                 lastIndex(), lastTerm(), lastAppliedIndex, lastAppliedTerm),
 816                                 null);
 817                             hasSnapshotCaptureInitiated = true;
 818                         }
 819                         if(callback != null){
 820                             callback.apply(replicatedLogEntry);
 821                         }
 822                     }
 823                 }
 824             );
 825         }
 826
 827     }
 828
 829     static class DeleteEntries implements Serializable {
 830         private static final long serialVersionUID = 1L;
 831         private final int fromIndex;
 832
 833         public DeleteEntries(int fromIndex) {
 834             this.fromIndex = fromIndex;
 835         }
 836
 837         public int getFromIndex() {
 838             return fromIndex;
 839         }
 840     }
 841
 842
 843     private class ElectionTermImpl implements ElectionTerm {
 844         /**
 845          * Identifier of the actor whose election term information this is
 846          */
 847         private long currentTerm = 0;
 848         private String votedFor = null;
 849
 850         @Override
 851         public long getCurrentTerm() {
 852             return currentTerm;
 853         }
 854
 855         @Override
 856         public String getVotedFor() {
 857             return votedFor;
 858         }
 859
 860         @Override public void update(long currentTerm, String votedFor) {
 861             if(LOG.isDebugEnabled()) {
 862                 LOG.debug("Set currentTerm={}, votedFor={}", currentTerm, votedFor);
 863             }
 864             this.currentTerm = currentTerm;
 865             this.votedFor = votedFor;
 866         }
 867
 868         @Override
 869         public void updateAndPersist(long currentTerm, String votedFor){
 870             update(currentTerm, votedFor);
 871             // FIXME : Maybe first persist then update the state
 872             persistence().persist(new UpdateElectionTerm(this.currentTerm, this.votedFor), new Procedure<UpdateElectionTerm>(){
 873
 874                 @Override public void apply(UpdateElectionTerm param)
 875                     throws Exception {
 876
 877                 }
 878             });
 879         }
 880     }
 881
 882     static class UpdateElectionTerm implements Serializable {
 883         private static final long serialVersionUID = 1L;
 884         private final long currentTerm;
 885         private final String votedFor;
 886
 887         public UpdateElectionTerm(long currentTerm, String votedFor) {
 888             this.currentTerm = currentTerm;
 889             this.votedFor = votedFor;
 890         }
 891
 892         public long getCurrentTerm() {
 893             return currentTerm;
 894         }
 895
 896         public String getVotedFor() {
 897             return votedFor;
 898         }
 899     }
 900
 901     protected class NonPersistentRaftDataProvider extends NonPersistentDataProvider {
 902
 903         public NonPersistentRaftDataProvider(){
 904
 905         }
 906
 907         /**
 908          * The way snapshotting works is,
 909          * <ol>
 910          * <li> RaftActor calls createSnapshot on the Shard
 911          * <li> Shard sends a CaptureSnapshotReply and RaftActor then calls saveSnapshot
 912          * <li> When saveSnapshot is invoked on the akka-persistence API it uses the SnapshotStore to save the snapshot.
 913          * The SnapshotStore sends SaveSnapshotSuccess or SaveSnapshotFailure. When the RaftActor gets SaveSnapshot
 914          * success it commits the snapshot to the in-memory journal. This commitSnapshot is mimicking what is done
 915          * in SaveSnapshotSuccess.
 916          * </ol>
 917          * @param o
 918          */
 919         @Override
 920         public void saveSnapshot(Object o) {
 921             // Make saving Snapshot successful
 922             commitSnapshot(-1L);
 923         }
 924     }
 925
 926     @VisibleForTesting
 927     void setCurrentBehavior(AbstractRaftActorBehavior behavior) {
 928         currentBehavior = behavior;
 929     }
 930
 931     protected RaftActorBehavior getCurrentBehavior() {
 932         return currentBehavior;
 933     }
 934
 935 }