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.Optional;
  21 import com.google.common.base.Stopwatch;
  22 import com.google.protobuf.ByteString;
  23 import java.io.Serializable;
  24 import java.util.Map;
  25 import org.opendaylight.controller.cluster.DataPersistenceProvider;
  26 import org.opendaylight.controller.cluster.common.actor.AbstractUntypedPersistentActor;
  27 import org.opendaylight.controller.cluster.notifications.RoleChanged;
  28 import org.opendaylight.controller.cluster.raft.base.messages.ApplyLogEntries;
  29 import org.opendaylight.controller.cluster.raft.base.messages.ApplySnapshot;
  30 import org.opendaylight.controller.cluster.raft.base.messages.ApplyState;
  31 import org.opendaylight.controller.cluster.raft.base.messages.CaptureSnapshot;
  32 import org.opendaylight.controller.cluster.raft.base.messages.CaptureSnapshotReply;
  33 import org.opendaylight.controller.cluster.raft.base.messages.Replicate;
  34 import org.opendaylight.controller.cluster.raft.base.messages.SendHeartBeat;
  35 import org.opendaylight.controller.cluster.raft.base.messages.SendInstallSnapshot;
  36 import org.opendaylight.controller.cluster.raft.behaviors.AbstractRaftActorBehavior;
  37 import org.opendaylight.controller.cluster.raft.behaviors.Follower;
  38 import org.opendaylight.controller.cluster.raft.behaviors.RaftActorBehavior;
  39 import org.opendaylight.controller.cluster.raft.client.messages.FindLeader;
  40 import org.opendaylight.controller.cluster.raft.client.messages.FindLeaderReply;
  41 import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
  42 import org.opendaylight.controller.cluster.raft.protobuff.client.messages.Payload;
  43 import org.opendaylight.controller.protobuff.messages.cluster.raft.AppendEntriesMessages;
  44 import org.slf4j.Logger;
  45 import org.slf4j.LoggerFactory;
  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 Logger LOG = LoggerFactory.getLogger(getClass());
  89
  90     /**
  91      * The current state determines the current behavior of a RaftActor
  92      * A Raft Actor always starts off in the Follower State
  93      */
  94     private RaftActorBehavior currentBehavior;
  95
  96     /**
  97      * This context should NOT be passed directly to any other actor it is
  98      * only to be consumed by the RaftActorBehaviors
  99      */
 100     private final RaftActorContext context;
 101
 102     /**
 103      * The in-memory journal
 104      */
 105     private ReplicatedLogImpl replicatedLog = new ReplicatedLogImpl();
 106
 107     private CaptureSnapshot captureSnapshot = null;
 108
 109     private Stopwatch recoveryTimer;
 110
 111     private int currentRecoveryBatchCount;
 112
 113     public RaftActor(String id, Map<String, String> peerAddresses) {
 114         this(id, peerAddresses, Optional.<ConfigParams>absent());
 115     }
 116
 117     public RaftActor(String id, Map<String, String> peerAddresses,
 118          Optional<ConfigParams> configParams) {
 119
 120         context = new RaftActorContextImpl(this.getSelf(),
 121             this.getContext(), id, new ElectionTermImpl(),
 122             -1, -1, replicatedLog, peerAddresses,
 123             (configParams.isPresent() ? configParams.get(): new DefaultConfigParamsImpl()),
 124             LOG);
 125     }
 126
 127     private void initRecoveryTimer() {
 128         if(recoveryTimer == null) {
 129             recoveryTimer = new Stopwatch();
 130             recoveryTimer.start();
 131         }
 132     }
 133
 134     @Override
 135     public void preStart() throws Exception {
 136         LOG.info("Starting recovery for {} with journal batch size {}", persistenceId(),
 137                 context.getConfigParams().getJournalRecoveryLogBatchSize());
 138
 139         super.preStart();
 140     }
 141
 142     @Override
 143     public void handleRecover(Object message) {
 144         if(persistence().isRecoveryApplicable()) {
 145             if (message instanceof SnapshotOffer) {
 146                 onRecoveredSnapshot((SnapshotOffer) message);
 147             } else if (message instanceof ReplicatedLogEntry) {
 148                 onRecoveredJournalLogEntry((ReplicatedLogEntry) message);
 149             } else if (message instanceof ApplyLogEntries) {
 150                 onRecoveredApplyLogEntries((ApplyLogEntries) message);
 151             } else if (message instanceof DeleteEntries) {
 152                 replicatedLog.removeFrom(((DeleteEntries) message).getFromIndex());
 153             } else if (message instanceof UpdateElectionTerm) {
 154                 context.getTermInformation().update(((UpdateElectionTerm) message).getCurrentTerm(),
 155                         ((UpdateElectionTerm) message).getVotedFor());
 156             } else if (message instanceof RecoveryCompleted) {
 157                 onRecoveryCompletedMessage();
 158             }
 159         } else {
 160             if (message instanceof RecoveryCompleted) {
 161                 // Delete all the messages from the akka journal so that we do not end up with consistency issues
 162                 // Note I am not using the dataPersistenceProvider and directly using the akka api here
 163                 deleteMessages(lastSequenceNr());
 164
 165                 // Delete all the akka snapshots as they will not be needed
 166                 deleteSnapshots(new SnapshotSelectionCriteria(scala.Long.MaxValue(), scala.Long.MaxValue()));
 167
 168                 onRecoveryComplete();
 169
 170                 initializeBehavior();
 171             }
 172         }
 173     }
 174
 175     private void onRecoveredSnapshot(SnapshotOffer offer) {
 176         if(LOG.isDebugEnabled()) {
 177             LOG.debug("{}: SnapshotOffer called..", persistenceId());
 178         }
 179
 180         initRecoveryTimer();
 181
 182         Snapshot snapshot = (Snapshot) offer.snapshot();
 183
 184         // Create a replicated log with the snapshot information
 185         // The replicated log can be used later on to retrieve this snapshot
 186         // when we need to install it on a peer
 187         replicatedLog = new ReplicatedLogImpl(snapshot);
 188
 189         context.setReplicatedLog(replicatedLog);
 190         context.setLastApplied(snapshot.getLastAppliedIndex());
 191         context.setCommitIndex(snapshot.getLastAppliedIndex());
 192
 193         Stopwatch timer = new Stopwatch();
 194         timer.start();
 195
 196         // Apply the snapshot to the actors state
 197         applyRecoverySnapshot(snapshot.getState());
 198
 199         timer.stop();
 200         LOG.info("Recovery snapshot applied for {} in {}: snapshotIndex={}, snapshotTerm={}, journal-size=" +
 201                 replicatedLog.size(), persistenceId(), timer.toString(),
 202                 replicatedLog.getSnapshotIndex(), replicatedLog.getSnapshotTerm());
 203     }
 204
 205     private void onRecoveredJournalLogEntry(ReplicatedLogEntry logEntry) {
 206         if(LOG.isDebugEnabled()) {
 207             LOG.debug("{}: Received ReplicatedLogEntry for recovery: {}", persistenceId(), logEntry.getIndex());
 208         }
 209
 210         replicatedLog.append(logEntry);
 211     }
 212
 213     private void onRecoveredApplyLogEntries(ApplyLogEntries ale) {
 214         if(LOG.isDebugEnabled()) {
 215             LOG.debug("{}: Received ApplyLogEntries for recovery, applying to state: {} to {}",
 216                     persistenceId(), context.getLastApplied() + 1, ale.getToIndex());
 217         }
 218
 219         for (long i = context.getLastApplied() + 1; i <= ale.getToIndex(); i++) {
 220             batchRecoveredLogEntry(replicatedLog.get(i));
 221         }
 222
 223         context.setLastApplied(ale.getToIndex());
 224         context.setCommitIndex(ale.getToIndex());
 225     }
 226
 227     private void batchRecoveredLogEntry(ReplicatedLogEntry logEntry) {
 228         initRecoveryTimer();
 229
 230         int batchSize = context.getConfigParams().getJournalRecoveryLogBatchSize();
 231         if(currentRecoveryBatchCount == 0) {
 232             startLogRecoveryBatch(batchSize);
 233         }
 234
 235         appendRecoveredLogEntry(logEntry.getData());
 236
 237         if(++currentRecoveryBatchCount >= batchSize) {
 238             endCurrentLogRecoveryBatch();
 239         }
 240     }
 241
 242     private void endCurrentLogRecoveryBatch() {
 243         applyCurrentLogRecoveryBatch();
 244         currentRecoveryBatchCount = 0;
 245     }
 246
 247     private void onRecoveryCompletedMessage() {
 248         if(currentRecoveryBatchCount > 0) {
 249             endCurrentLogRecoveryBatch();
 250         }
 251
 252         onRecoveryComplete();
 253
 254         String recoveryTime = "";
 255         if(recoveryTimer != null) {
 256             recoveryTimer.stop();
 257             recoveryTime = " in " + recoveryTimer.toString();
 258             recoveryTimer = null;
 259         }
 260
 261         LOG.info(
 262             "Recovery completed" + recoveryTime + " - Switching actor to Follower - " +
 263                 "Persistence Id =  " + persistenceId() +
 264                 " Last index in log={}, snapshotIndex={}, snapshotTerm={}, " +
 265                 "journal-size={}",
 266             replicatedLog.lastIndex(), replicatedLog.getSnapshotIndex(),
 267             replicatedLog.getSnapshotTerm(), replicatedLog.size());
 268
 269         initializeBehavior();
 270     }
 271
 272     protected void initializeBehavior(){
 273         changeCurrentBehavior(new Follower(context));
 274     }
 275
 276     protected void changeCurrentBehavior(RaftActorBehavior newBehavior){
 277         RaftActorBehavior oldBehavior = currentBehavior;
 278         currentBehavior = newBehavior;
 279         handleBehaviorChange(oldBehavior, currentBehavior);
 280     }
 281
 282     @Override public void handleCommand(Object message) {
 283         if (message instanceof ApplyState){
 284             ApplyState applyState = (ApplyState) message;
 285
 286             if(LOG.isDebugEnabled()) {
 287                 LOG.debug("{}: Applying state for log index {} data {}",
 288                     persistenceId(), applyState.getReplicatedLogEntry().getIndex(),
 289                     applyState.getReplicatedLogEntry().getData());
 290             }
 291
 292             applyState(applyState.getClientActor(), applyState.getIdentifier(),
 293                 applyState.getReplicatedLogEntry().getData());
 294
 295         } else if (message instanceof ApplyLogEntries){
 296             ApplyLogEntries ale = (ApplyLogEntries) message;
 297             if(LOG.isDebugEnabled()) {
 298                 LOG.debug("{}: Persisting ApplyLogEntries with index={}", persistenceId(), ale.getToIndex());
 299             }
 300             persistence().persist(new ApplyLogEntries(ale.getToIndex()), new Procedure<ApplyLogEntries>() {
 301                 @Override
 302                 public void apply(ApplyLogEntries param) throws Exception {
 303                 }
 304             });
 305
 306         } else if(message instanceof ApplySnapshot ) {
 307             Snapshot snapshot = ((ApplySnapshot) message).getSnapshot();
 308
 309             if(LOG.isDebugEnabled()) {
 310                 LOG.debug("{}: ApplySnapshot called on Follower Actor " +
 311                         "snapshotIndex:{}, snapshotTerm:{}", persistenceId(), snapshot.getLastAppliedIndex(),
 312                     snapshot.getLastAppliedTerm()
 313                 );
 314             }
 315
 316             applySnapshot(snapshot.getState());
 317
 318             //clears the followers log, sets the snapshot index to ensure adjusted-index works
 319             replicatedLog = new ReplicatedLogImpl(snapshot);
 320             context.setReplicatedLog(replicatedLog);
 321             context.setLastApplied(snapshot.getLastAppliedIndex());
 322
 323         } else if (message instanceof FindLeader) {
 324             getSender().tell(
 325                 new FindLeaderReply(getLeaderAddress()),
 326                 getSelf()
 327             );
 328
 329         } else if (message instanceof SaveSnapshotSuccess) {
 330             SaveSnapshotSuccess success = (SaveSnapshotSuccess) message;
 331             LOG.info("{}: SaveSnapshotSuccess received for snapshot", persistenceId());
 332
 333             long sequenceNumber = success.metadata().sequenceNr();
 334
 335             commitSnapshot(sequenceNumber);
 336
 337         } else if (message instanceof SaveSnapshotFailure) {
 338             SaveSnapshotFailure saveSnapshotFailure = (SaveSnapshotFailure) message;
 339
 340             LOG.error("{}: SaveSnapshotFailure received for snapshot Cause:",
 341                     persistenceId(), saveSnapshotFailure.cause());
 342
 343             context.getReplicatedLog().snapshotRollback();
 344
 345             LOG.info("{}: Replicated Log rollbacked. Snapshot will be attempted in the next cycle." +
 346                 "snapshotIndex:{}, snapshotTerm:{}, log-size:{}", persistenceId(),
 347                 context.getReplicatedLog().getSnapshotIndex(),
 348                 context.getReplicatedLog().getSnapshotTerm(),
 349                 context.getReplicatedLog().size());
 350
 351         } else if (message instanceof CaptureSnapshot) {
 352             LOG.info("{}: CaptureSnapshot received by actor", persistenceId());
 353
 354             if(captureSnapshot == null) {
 355                 captureSnapshot = (CaptureSnapshot)message;
 356                 createSnapshot();
 357             }
 358
 359         } else if (message instanceof CaptureSnapshotReply){
 360             handleCaptureSnapshotReply(((CaptureSnapshotReply) message).getSnapshot());
 361
 362         } else {
 363             if (!(message instanceof AppendEntriesMessages.AppendEntries)
 364                 && !(message instanceof AppendEntriesReply) && !(message instanceof SendHeartBeat)) {
 365                 if(LOG.isDebugEnabled()) {
 366                     LOG.debug("{}: onReceiveCommand: message: {}", persistenceId(), message.getClass());
 367                 }
 368             }
 369
 370             RaftActorBehavior oldBehavior = currentBehavior;
 371             currentBehavior = currentBehavior.handleMessage(getSender(), message);
 372
 373             handleBehaviorChange(oldBehavior, currentBehavior);
 374         }
 375     }
 376
 377     private void handleBehaviorChange(RaftActorBehavior oldBehavior, RaftActorBehavior currentBehavior) {
 378         if (oldBehavior != currentBehavior){
 379             onStateChanged();
 380         }
 381
 382         String oldBehaviorLeaderId = oldBehavior == null? null : oldBehavior.getLeaderId();
 383         String oldBehaviorState = oldBehavior == null? null : oldBehavior.state().name();
 384
 385         // it can happen that the state has not changed but the leader has changed.
 386         onLeaderChanged(oldBehaviorLeaderId, currentBehavior.getLeaderId());
 387
 388         if (getRoleChangeNotifier().isPresent() &&
 389                 (oldBehavior == null || (oldBehavior.state() != currentBehavior.state()))) {
 390             getRoleChangeNotifier().get().tell(
 391                     new RoleChanged(getId(), oldBehaviorState , currentBehavior.state().name()),
 392                     getSelf());
 393         }
 394     }
 395
 396     /**
 397      * When a derived RaftActor needs to persist something it must call
 398      * persistData.
 399      *
 400      * @param clientActor
 401      * @param identifier
 402      * @param data
 403      */
 404     protected void persistData(final ActorRef clientActor, final String identifier,
 405         final Payload data) {
 406
 407         ReplicatedLogEntry replicatedLogEntry = new ReplicatedLogImplEntry(
 408             context.getReplicatedLog().lastIndex() + 1,
 409             context.getTermInformation().getCurrentTerm(), data);
 410
 411         if(LOG.isDebugEnabled()) {
 412             LOG.debug("{}: Persist data {}", persistenceId(), replicatedLogEntry);
 413         }
 414
 415         final RaftActorContext raftContext = getRaftActorContext();
 416
 417         replicatedLog
 418                 .appendAndPersist(replicatedLogEntry, new Procedure<ReplicatedLogEntry>() {
 419                     @Override
 420                     public void apply(ReplicatedLogEntry replicatedLogEntry) throws Exception {
 421                         if(!hasFollowers()){
 422                             // Increment the Commit Index and the Last Applied values
 423                             raftContext.setCommitIndex(replicatedLogEntry.getIndex());
 424                             raftContext.setLastApplied(replicatedLogEntry.getIndex());
 425
 426                             // Apply the state immediately
 427                             applyState(clientActor, identifier, data);
 428
 429                             // Send a ApplyLogEntries message so that we write the fact that we applied
 430                             // the state to durable storage
 431                             self().tell(new ApplyLogEntries((int) replicatedLogEntry.getIndex()), self());
 432
 433                             // Check if the "real" snapshot capture has been initiated. If no then do the fake snapshot
 434                             if(!context.isSnapshotCaptureInitiated()){
 435                                 raftContext.getReplicatedLog().snapshotPreCommit(raftContext.getLastApplied(),
 436                                         raftContext.getTermInformation().getCurrentTerm());
 437                                 raftContext.getReplicatedLog().snapshotCommit();
 438                             } else {
 439                                 LOG.debug("{}: Skipping fake snapshotting for {} because real snapshotting is in progress",
 440                                         persistenceId(), getId());
 441                             }
 442                         } else if (clientActor != null) {
 443                             // Send message for replication
 444                             currentBehavior.handleMessage(getSelf(),
 445                                     new Replicate(clientActor, identifier,
 446                                             replicatedLogEntry)
 447                             );
 448                         }
 449
 450                     }
 451                 });    }
 452
 453     protected String getId() {
 454         return context.getId();
 455     }
 456
 457     /**
 458      * Derived actors can call the isLeader method to check if the current
 459      * RaftActor is the Leader or not
 460      *
 461      * @return true it this RaftActor is a Leader false otherwise
 462      */
 463     protected boolean isLeader() {
 464         return context.getId().equals(currentBehavior.getLeaderId());
 465     }
 466
 467     /**
 468      * Derived actor can call getLeader if they need a reference to the Leader.
 469      * This would be useful for example in forwarding a request to an actor
 470      * which is the leader
 471      *
 472      * @return A reference to the leader if known, null otherwise
 473      */
 474     protected ActorSelection getLeader(){
 475         String leaderAddress = getLeaderAddress();
 476
 477         if(leaderAddress == null){
 478             return null;
 479         }
 480
 481         return context.actorSelection(leaderAddress);
 482     }
 483
 484     /**
 485      *
 486      * @return the current leader's id
 487      */
 488     protected String getLeaderId(){
 489         return currentBehavior.getLeaderId();
 490     }
 491
 492     protected RaftState getRaftState() {
 493         return currentBehavior.state();
 494     }
 495
 496     protected ReplicatedLogEntry getLastLogEntry() {
 497         return replicatedLog.last();
 498     }
 499
 500     protected Long getCurrentTerm(){
 501         return context.getTermInformation().getCurrentTerm();
 502     }
 503
 504     protected Long getCommitIndex(){
 505         return context.getCommitIndex();
 506     }
 507
 508     protected Long getLastApplied(){
 509         return context.getLastApplied();
 510     }
 511
 512     protected RaftActorContext getRaftActorContext() {
 513         return context;
 514     }
 515
 516     /**
 517      * setPeerAddress sets the address of a known peer at a later time.
 518      * <p>
 519      * This is to account for situations where a we know that a peer
 520      * exists but we do not know an address up-front. This may also be used in
 521      * situations where a known peer starts off in a different location and we
 522      * need to change it's address
 523      * <p>
 524      * Note that if the peerId does not match the list of peers passed to
 525      * this actor during construction an IllegalStateException will be thrown.
 526      *
 527      * @param peerId
 528      * @param peerAddress
 529      */
 530     protected void setPeerAddress(String peerId, String peerAddress){
 531         context.setPeerAddress(peerId, peerAddress);
 532     }
 533
 534     protected void commitSnapshot(long sequenceNumber) {
 535         context.getReplicatedLog().snapshotCommit();
 536
 537         // TODO: Not sure if we want to be this aggressive with trimming stuff
 538         trimPersistentData(sequenceNumber);
 539     }
 540
 541     /**
 542      * The applyState method will be called by the RaftActor when some data
 543      * needs to be applied to the actor's state
 544      *
 545      * @param clientActor A reference to the client who sent this message. This
 546      *                    is the same reference that was passed to persistData
 547      *                    by the derived actor. clientActor may be null when
 548      *                    the RaftActor is behaving as a follower or during
 549      *                    recovery.
 550      * @param identifier  The identifier of the persisted data. This is also
 551      *                    the same identifier that was passed to persistData by
 552      *                    the derived actor. identifier may be null when
 553      *                    the RaftActor is behaving as a follower or during
 554      *                    recovery
 555      * @param data        A piece of data that was persisted by the persistData call.
 556      *                    This should NEVER be null.
 557      */
 558     protected abstract void applyState(ActorRef clientActor, String identifier,
 559         Object data);
 560
 561     /**
 562      * This method is called during recovery at the start of a batch of state entries. Derived
 563      * classes should perform any initialization needed to start a batch.
 564      */
 565     protected abstract void startLogRecoveryBatch(int maxBatchSize);
 566
 567     /**
 568      * This method is called during recovery to append state data to the current batch. This method
 569      * is called 1 or more times after {@link #startLogRecoveryBatch}.
 570      *
 571      * @param data the state data
 572      */
 573     protected abstract void appendRecoveredLogEntry(Payload data);
 574
 575     /**
 576      * This method is called during recovery to reconstruct the state of the actor.
 577      *
 578      * @param snapshot A snapshot of the state of the actor
 579      */
 580     protected abstract void applyRecoverySnapshot(byte[] snapshotBytes);
 581
 582     /**
 583      * This method is called during recovery at the end of a batch to apply the current batched
 584      * log entries. This method is called after {@link #appendRecoveredLogEntry}.
 585      */
 586     protected abstract void applyCurrentLogRecoveryBatch();
 587
 588     /**
 589      * This method is called when recovery is complete.
 590      */
 591     protected abstract void onRecoveryComplete();
 592
 593     /**
 594      * This method will be called by the RaftActor when a snapshot needs to be
 595      * created. The derived actor should respond with its current state.
 596      * <p/>
 597      * During recovery the state that is returned by the derived actor will
 598      * be passed back to it by calling the applySnapshot  method
 599      *
 600      * @return The current state of the actor
 601      */
 602     protected abstract void createSnapshot();
 603
 604     /**
 605      * This method can be called at any other point during normal
 606      * operations when the derived actor is out of sync with it's peers
 607      * and the only way to bring it in sync is by applying a snapshot
 608      *
 609      * @param snapshotBytes A snapshot of the state of the actor
 610      */
 611     protected abstract void applySnapshot(byte[] snapshotBytes);
 612
 613     /**
 614      * This method will be called by the RaftActor when the state of the
 615      * RaftActor changes. The derived actor can then use methods like
 616      * isLeader or getLeader to do something useful
 617      */
 618     protected abstract void onStateChanged();
 619
 620     protected abstract DataPersistenceProvider persistence();
 621
 622     /**
 623      * Notifier Actor for this RaftActor to notify when a role change happens
 624      * @return ActorRef - ActorRef of the notifier or Optional.absent if none.
 625      */
 626     protected abstract Optional<ActorRef> getRoleChangeNotifier();
 627
 628     protected void onLeaderChanged(String oldLeader, String newLeader){};
 629
 630     private void trimPersistentData(long sequenceNumber) {
 631         // Trim akka snapshots
 632         // FIXME : Not sure how exactly the SnapshotSelectionCriteria is applied
 633         // For now guessing that it is ANDed.
 634         persistence().deleteSnapshots(new SnapshotSelectionCriteria(
 635             sequenceNumber - context.getConfigParams().getSnapshotBatchCount(), 43200000));
 636
 637         // Trim akka journal
 638         persistence().deleteMessages(sequenceNumber);
 639     }
 640
 641     private String getLeaderAddress(){
 642         if(isLeader()){
 643             return getSelf().path().toString();
 644         }
 645         String leaderId = currentBehavior.getLeaderId();
 646         if (leaderId == null) {
 647             return null;
 648         }
 649         String peerAddress = context.getPeerAddress(leaderId);
 650         if(LOG.isDebugEnabled()) {
 651             LOG.debug("{}: getLeaderAddress leaderId = {} peerAddress = {}",
 652                     persistenceId(), leaderId, peerAddress);
 653         }
 654
 655         return peerAddress;
 656     }
 657
 658     private void handleCaptureSnapshotReply(byte[] snapshotBytes) {
 659         LOG.info("{}: CaptureSnapshotReply received by actor: snapshot size {}", persistenceId(), snapshotBytes.length);
 660
 661         // create a snapshot object from the state provided and save it
 662         // when snapshot is saved async, SaveSnapshotSuccess is raised.
 663
 664         Snapshot sn = Snapshot.create(snapshotBytes,
 665             context.getReplicatedLog().getFrom(captureSnapshot.getLastAppliedIndex() + 1),
 666             captureSnapshot.getLastIndex(), captureSnapshot.getLastTerm(),
 667             captureSnapshot.getLastAppliedIndex(), captureSnapshot.getLastAppliedTerm());
 668
 669         persistence().saveSnapshot(sn);
 670
 671         LOG.info("{}: Persisting of snapshot done:{}", persistenceId(), sn.getLogMessage());
 672
 673         //be greedy and remove entries from in-mem journal which are in the snapshot
 674         // and update snapshotIndex and snapshotTerm without waiting for the success,
 675
 676         context.getReplicatedLog().snapshotPreCommit(
 677             captureSnapshot.getLastAppliedIndex(),
 678             captureSnapshot.getLastAppliedTerm());
 679
 680         LOG.info("{}: Removed in-memory snapshotted entries, adjusted snaphsotIndex:{} " +
 681             "and term:{}", persistenceId(), captureSnapshot.getLastAppliedIndex(),
 682             captureSnapshot.getLastAppliedTerm());
 683
 684         if (isLeader() && captureSnapshot.isInstallSnapshotInitiated()) {
 685             // this would be call straight to the leader and won't initiate in serialization
 686             currentBehavior.handleMessage(getSelf(), new SendInstallSnapshot(
 687                     ByteString.copyFrom(snapshotBytes)));
 688         }
 689
 690         captureSnapshot = null;
 691         context.setSnapshotCaptureInitiated(false);
 692     }
 693
 694     protected boolean hasFollowers(){
 695         return getRaftActorContext().getPeerAddresses().keySet().size() > 0;
 696     }
 697
 698     private class ReplicatedLogImpl extends AbstractReplicatedLogImpl {
 699
 700         private static final int DATA_SIZE_DIVIDER = 5;
 701         private long dataSizeSinceLastSnapshot = 0;
 702
 703         public ReplicatedLogImpl(Snapshot snapshot) {
 704             super(snapshot.getLastAppliedIndex(), snapshot.getLastAppliedTerm(),
 705                 snapshot.getUnAppliedEntries());
 706         }
 707
 708         public ReplicatedLogImpl() {
 709             super();
 710         }
 711
 712         @Override public void removeFromAndPersist(long logEntryIndex) {
 713             int adjustedIndex = adjustedIndex(logEntryIndex);
 714
 715             if (adjustedIndex < 0) {
 716                 return;
 717             }
 718
 719             // FIXME: Maybe this should be done after the command is saved
 720             journal.subList(adjustedIndex , journal.size()).clear();
 721
 722             persistence().persist(new DeleteEntries(adjustedIndex), new Procedure<DeleteEntries>(){
 723
 724                 @Override public void apply(DeleteEntries param)
 725                     throws Exception {
 726                     //FIXME : Doing nothing for now
 727                     dataSize = 0;
 728                     for(ReplicatedLogEntry entry : journal){
 729                         dataSize += entry.size();
 730                     }
 731                 }
 732             });
 733         }
 734
 735         @Override public void appendAndPersist(
 736             final ReplicatedLogEntry replicatedLogEntry) {
 737             appendAndPersist(replicatedLogEntry, null);
 738         }
 739
 740         @Override
 741         public int dataSize() {
 742             return dataSize;
 743         }
 744
 745         public void appendAndPersist(
 746             final ReplicatedLogEntry replicatedLogEntry,
 747             final Procedure<ReplicatedLogEntry> callback)  {
 748
 749             if(LOG.isDebugEnabled()) {
 750                 LOG.debug("{}: Append log entry and persist {} ", persistenceId(), replicatedLogEntry);
 751             }
 752
 753             // FIXME : By adding the replicated log entry to the in-memory journal we are not truly ensuring durability of the logs
 754             journal.add(replicatedLogEntry);
 755
 756             // When persisting events with persist it is guaranteed that the
 757             // persistent actor will not receive further commands between the
 758             // persist call and the execution(s) of the associated event
 759             // handler. This also holds for multiple persist calls in context
 760             // of a single command.
 761             persistence().persist(replicatedLogEntry,
 762                 new Procedure<ReplicatedLogEntry>() {
 763                     @Override
 764                     public void apply(ReplicatedLogEntry evt) throws Exception {
 765                         int logEntrySize = replicatedLogEntry.size();
 766
 767                         dataSize += logEntrySize;
 768                         long dataSizeForCheck = dataSize;
 769
 770                         dataSizeSinceLastSnapshot += logEntrySize;
 771                         long journalSize = lastIndex()+1;
 772
 773                         if(!hasFollowers()) {
 774                             // When we do not have followers we do not maintain an in-memory log
 775                             // due to this the journalSize will never become anything close to the
 776                             // snapshot batch count. In fact will mostly be 1.
 777                             // Similarly since the journal's dataSize depends on the entries in the
 778                             // journal the journal's dataSize will never reach a value close to the
 779                             // memory threshold.
 780                             // By maintaining the dataSize outside the journal we are tracking essentially
 781                             // what we have written to the disk however since we no longer are in
 782                             // need of doing a snapshot just for the sake of freeing up memory we adjust
 783                             // the real size of data by the DATA_SIZE_DIVIDER so that we do not snapshot as often
 784                             // as if we were maintaining a real snapshot
 785                             dataSizeForCheck = dataSizeSinceLastSnapshot / DATA_SIZE_DIVIDER;
 786                         }
 787
 788                         long dataThreshold = Runtime.getRuntime().totalMemory() *
 789                                 getRaftActorContext().getConfigParams().getSnapshotDataThresholdPercentage() / 100;
 790
 791                         // when a snaphsot is being taken, captureSnapshot != null
 792                         if (!context.isSnapshotCaptureInitiated() &&
 793                                 ( journalSize % context.getConfigParams().getSnapshotBatchCount() == 0 ||
 794                                         dataSizeForCheck > dataThreshold)) {
 795
 796                             dataSizeSinceLastSnapshot = 0;
 797
 798                             LOG.info("{}: Initiating Snapshot Capture..", persistenceId());
 799                             long lastAppliedIndex = -1;
 800                             long lastAppliedTerm = -1;
 801
 802                             ReplicatedLogEntry lastAppliedEntry = get(context.getLastApplied());
 803                             if (!hasFollowers()) {
 804                                 lastAppliedIndex = replicatedLogEntry.getIndex();
 805                                 lastAppliedTerm = replicatedLogEntry.getTerm();
 806                             } else if (lastAppliedEntry != null) {
 807                                 lastAppliedIndex = lastAppliedEntry.getIndex();
 808                                 lastAppliedTerm = lastAppliedEntry.getTerm();
 809                             }
 810
 811                             if(LOG.isDebugEnabled()) {
 812                                 LOG.debug("{}: Snapshot Capture logSize: {}", persistenceId(), journal.size());
 813                                 LOG.debug("{}: Snapshot Capture lastApplied:{} ",
 814                                         persistenceId(), context.getLastApplied());
 815                                 LOG.debug("{}: Snapshot Capture lastAppliedIndex:{}", persistenceId(),
 816                                         lastAppliedIndex);
 817                                 LOG.debug("{}: Snapshot Capture lastAppliedTerm:{}", persistenceId(),
 818                                         lastAppliedTerm);
 819                             }
 820
 821                             // send a CaptureSnapshot to self to make the expensive operation async.
 822                             getSelf().tell(new CaptureSnapshot(
 823                                 lastIndex(), lastTerm(), lastAppliedIndex, lastAppliedTerm),
 824                                 null);
 825                             context.setSnapshotCaptureInitiated(true);
 826                         }
 827                         if(callback != null){
 828                             callback.apply(replicatedLogEntry);
 829                         }
 830                     }
 831                 }
 832             );
 833         }
 834
 835     }
 836
 837     static class DeleteEntries implements Serializable {
 838         private static final long serialVersionUID = 1L;
 839         private final int fromIndex;
 840
 841         public DeleteEntries(int fromIndex) {
 842             this.fromIndex = fromIndex;
 843         }
 844
 845         public int getFromIndex() {
 846             return fromIndex;
 847         }
 848     }
 849
 850
 851     private class ElectionTermImpl implements ElectionTerm {
 852         /**
 853          * Identifier of the actor whose election term information this is
 854          */
 855         private long currentTerm = 0;
 856         private String votedFor = null;
 857
 858         @Override
 859         public long getCurrentTerm() {
 860             return currentTerm;
 861         }
 862
 863         @Override
 864         public String getVotedFor() {
 865             return votedFor;
 866         }
 867
 868         @Override public void update(long currentTerm, String votedFor) {
 869             if(LOG.isDebugEnabled()) {
 870                 LOG.debug("{}: Set currentTerm={}, votedFor={}", persistenceId(), currentTerm, votedFor);
 871             }
 872             this.currentTerm = currentTerm;
 873             this.votedFor = votedFor;
 874         }
 875
 876         @Override
 877         public void updateAndPersist(long currentTerm, String votedFor){
 878             update(currentTerm, votedFor);
 879             // FIXME : Maybe first persist then update the state
 880             persistence().persist(new UpdateElectionTerm(this.currentTerm, this.votedFor), new Procedure<UpdateElectionTerm>(){
 881
 882                 @Override public void apply(UpdateElectionTerm param)
 883                     throws Exception {
 884
 885                 }
 886             });
 887         }
 888     }
 889
 890     static class UpdateElectionTerm implements Serializable {
 891         private static final long serialVersionUID = 1L;
 892         private final long currentTerm;
 893         private final String votedFor;
 894
 895         public UpdateElectionTerm(long currentTerm, String votedFor) {
 896             this.currentTerm = currentTerm;
 897             this.votedFor = votedFor;
 898         }
 899
 900         public long getCurrentTerm() {
 901             return currentTerm;
 902         }
 903
 904         public String getVotedFor() {
 905             return votedFor;
 906         }
 907     }
 908
 909     protected class NonPersistentRaftDataProvider extends NonPersistentDataProvider {
 910
 911         public NonPersistentRaftDataProvider(){
 912
 913         }
 914
 915         /**
 916          * The way snapshotting works is,
 917          * <ol>
 918          * <li> RaftActor calls createSnapshot on the Shard
 919          * <li> Shard sends a CaptureSnapshotReply and RaftActor then calls saveSnapshot
 920          * <li> When saveSnapshot is invoked on the akka-persistence API it uses the SnapshotStore to save the snapshot.
 921          * The SnapshotStore sends SaveSnapshotSuccess or SaveSnapshotFailure. When the RaftActor gets SaveSnapshot
 922          * success it commits the snapshot to the in-memory journal. This commitSnapshot is mimicking what is done
 923          * in SaveSnapshotSuccess.
 924          * </ol>
 925          * @param o
 926          */
 927         @Override
 928         public void saveSnapshot(Object o) {
 929             // Make saving Snapshot successful
 930             commitSnapshot(-1L);
 931         }
 932     }
 933
 934     @VisibleForTesting
 935     void setCurrentBehavior(AbstractRaftActorBehavior behavior) {
 936         currentBehavior = behavior;
 937     }
 938
 939     protected RaftActorBehavior getCurrentBehavior() {
 940         return currentBehavior;
 941     }
 942
 943 }