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 Stopwatch recoveryTimer;
 111
 112     private int currentRecoveryBatchCount;
 113
 114     public RaftActor(String id, Map<String, String> peerAddresses) {
 115         this(id, peerAddresses, Optional.<ConfigParams>absent());
 116     }
 117
 118     public RaftActor(String id, Map<String, String> peerAddresses,
 119          Optional<ConfigParams> configParams) {
 120
 121         context = new RaftActorContextImpl(this.getSelf(),
 122             this.getContext(), id, new ElectionTermImpl(),
 123             -1, -1, replicatedLog, peerAddresses,
 124             (configParams.isPresent() ? configParams.get(): new DefaultConfigParamsImpl()),
 125             LOG);
 126     }
 127
 128     private void initRecoveryTimer() {
 129         if(recoveryTimer == null) {
 130             recoveryTimer = new Stopwatch();
 131             recoveryTimer.start();
 132         }
 133     }
 134
 135     @Override
 136     public void preStart() throws Exception {
 137         LOG.info("Starting recovery for {} with journal batch size {}", persistenceId(),
 138                 context.getConfigParams().getJournalRecoveryLogBatchSize());
 139
 140         super.preStart();
 141     }
 142
 143     @Override
 144     public void handleRecover(Object message) {
 145         if(persistence().isRecoveryApplicable()) {
 146             if (message instanceof SnapshotOffer) {
 147                 onRecoveredSnapshot((SnapshotOffer) message);
 148             } else if (message instanceof ReplicatedLogEntry) {
 149                 onRecoveredJournalLogEntry((ReplicatedLogEntry) message);
 150             } else if (message instanceof ApplyLogEntries) {
 151                 onRecoveredApplyLogEntries((ApplyLogEntries) message);
 152             } else if (message instanceof DeleteEntries) {
 153                 replicatedLog.removeFrom(((DeleteEntries) message).getFromIndex());
 154             } else if (message instanceof UpdateElectionTerm) {
 155                 context.getTermInformation().update(((UpdateElectionTerm) message).getCurrentTerm(),
 156                         ((UpdateElectionTerm) message).getVotedFor());
 157             } else if (message instanceof RecoveryCompleted) {
 158                 onRecoveryCompletedMessage();
 159             }
 160         } else {
 161             if (message instanceof RecoveryCompleted) {
 162                 // Delete all the messages from the akka journal so that we do not end up with consistency issues
 163                 // Note I am not using the dataPersistenceProvider and directly using the akka api here
 164                 deleteMessages(lastSequenceNr());
 165
 166                 // Delete all the akka snapshots as they will not be needed
 167                 deleteSnapshots(new SnapshotSelectionCriteria(scala.Long.MaxValue(), scala.Long.MaxValue()));
 168
 169                 onRecoveryComplete();
 170
 171                 initializeBehavior();
 172             }
 173         }
 174     }
 175
 176     private void onRecoveredSnapshot(SnapshotOffer offer) {
 177         if(LOG.isDebugEnabled()) {
 178             LOG.debug("{}: SnapshotOffer called..", persistenceId());
 179         }
 180
 181         initRecoveryTimer();
 182
 183         Snapshot snapshot = (Snapshot) offer.snapshot();
 184
 185         // Create a replicated log with the snapshot information
 186         // The replicated log can be used later on to retrieve this snapshot
 187         // when we need to install it on a peer
 188         replicatedLog = new ReplicatedLogImpl(snapshot);
 189
 190         context.setReplicatedLog(replicatedLog);
 191         context.setLastApplied(snapshot.getLastAppliedIndex());
 192         context.setCommitIndex(snapshot.getLastAppliedIndex());
 193
 194         Stopwatch timer = new Stopwatch();
 195         timer.start();
 196
 197         // Apply the snapshot to the actors state
 198         applyRecoverySnapshot(snapshot.getState());
 199
 200         timer.stop();
 201         LOG.info("Recovery snapshot applied for {} in {}: snapshotIndex={}, snapshotTerm={}, journal-size=" +
 202                 replicatedLog.size(), persistenceId(), timer.toString(),
 203                 replicatedLog.getSnapshotIndex(), replicatedLog.getSnapshotTerm());
 204     }
 205
 206     private void onRecoveredJournalLogEntry(ReplicatedLogEntry logEntry) {
 207         if(LOG.isDebugEnabled()) {
 208             LOG.debug("{}: Received ReplicatedLogEntry for recovery: {}", persistenceId(), logEntry.getIndex());
 209         }
 210
 211         replicatedLog.append(logEntry);
 212     }
 213
 214     private void onRecoveredApplyLogEntries(ApplyLogEntries ale) {
 215         if(LOG.isDebugEnabled()) {
 216             LOG.debug("{}: Received ApplyLogEntries for recovery, applying to state: {} to {}",
 217                     persistenceId(), context.getLastApplied() + 1, ale.getToIndex());
 218         }
 219
 220         for (long i = context.getLastApplied() + 1; i <= ale.getToIndex(); i++) {
 221             batchRecoveredLogEntry(replicatedLog.get(i));
 222         }
 223
 224         context.setLastApplied(ale.getToIndex());
 225         context.setCommitIndex(ale.getToIndex());
 226     }
 227
 228     private void batchRecoveredLogEntry(ReplicatedLogEntry logEntry) {
 229         initRecoveryTimer();
 230
 231         int batchSize = context.getConfigParams().getJournalRecoveryLogBatchSize();
 232         if(currentRecoveryBatchCount == 0) {
 233             startLogRecoveryBatch(batchSize);
 234         }
 235
 236         appendRecoveredLogEntry(logEntry.getData());
 237
 238         if(++currentRecoveryBatchCount >= batchSize) {
 239             endCurrentLogRecoveryBatch();
 240         }
 241     }
 242
 243     private void endCurrentLogRecoveryBatch() {
 244         applyCurrentLogRecoveryBatch();
 245         currentRecoveryBatchCount = 0;
 246     }
 247
 248     private void onRecoveryCompletedMessage() {
 249         if(currentRecoveryBatchCount > 0) {
 250             endCurrentLogRecoveryBatch();
 251         }
 252
 253         onRecoveryComplete();
 254
 255         String recoveryTime = "";
 256         if(recoveryTimer != null) {
 257             recoveryTimer.stop();
 258             recoveryTime = " in " + recoveryTimer.toString();
 259             recoveryTimer = null;
 260         }
 261
 262         LOG.info(
 263             "Recovery completed" + recoveryTime + " - Switching actor to Follower - " +
 264                 "Persistence Id =  " + persistenceId() +
 265                 " Last index in log={}, snapshotIndex={}, snapshotTerm={}, " +
 266                 "journal-size={}",
 267             replicatedLog.lastIndex(), replicatedLog.getSnapshotIndex(),
 268             replicatedLog.getSnapshotTerm(), replicatedLog.size());
 269
 270         initializeBehavior();
 271     }
 272
 273     protected void initializeBehavior(){
 274         changeCurrentBehavior(new Follower(context));
 275     }
 276
 277     protected void changeCurrentBehavior(RaftActorBehavior newBehavior){
 278         RaftActorBehavior oldBehavior = currentBehavior;
 279         currentBehavior = newBehavior;
 280         handleBehaviorChange(oldBehavior, currentBehavior);
 281     }
 282
 283     @Override public void handleCommand(Object message) {
 284         if (message instanceof ApplyState){
 285             ApplyState applyState = (ApplyState) message;
 286
 287             if(LOG.isDebugEnabled()) {
 288                 LOG.debug("{}: Applying state for log index {} data {}",
 289                     persistenceId(), applyState.getReplicatedLogEntry().getIndex(),
 290                     applyState.getReplicatedLogEntry().getData());
 291             }
 292
 293             applyState(applyState.getClientActor(), applyState.getIdentifier(),
 294                 applyState.getReplicatedLogEntry().getData());
 295
 296         } else if (message instanceof ApplyLogEntries){
 297             ApplyLogEntries ale = (ApplyLogEntries) message;
 298             if(LOG.isDebugEnabled()) {
 299                 LOG.debug("{}: Persisting ApplyLogEntries with index={}", persistenceId(), ale.getToIndex());
 300             }
 301             persistence().persist(new ApplyLogEntries(ale.getToIndex()), new Procedure<ApplyLogEntries>() {
 302                 @Override
 303                 public void apply(ApplyLogEntries param) throws Exception {
 304                 }
 305             });
 306
 307         } else if(message instanceof ApplySnapshot ) {
 308             Snapshot snapshot = ((ApplySnapshot) message).getSnapshot();
 309
 310             if(LOG.isDebugEnabled()) {
 311                 LOG.debug("{}: ApplySnapshot called on Follower Actor " +
 312                         "snapshotIndex:{}, snapshotTerm:{}", persistenceId(), snapshot.getLastAppliedIndex(),
 313                     snapshot.getLastAppliedTerm()
 314                 );
 315             }
 316
 317             applySnapshot(snapshot.getState());
 318
 319             //clears the followers log, sets the snapshot index to ensure adjusted-index works
 320             replicatedLog = new ReplicatedLogImpl(snapshot);
 321             context.setReplicatedLog(replicatedLog);
 322             context.setLastApplied(snapshot.getLastAppliedIndex());
 323
 324         } else if (message instanceof FindLeader) {
 325             getSender().tell(
 326                 new FindLeaderReply(getLeaderAddress()),
 327                 getSelf()
 328             );
 329
 330         } else if (message instanceof SaveSnapshotSuccess) {
 331             SaveSnapshotSuccess success = (SaveSnapshotSuccess) message;
 332             LOG.info("{}: SaveSnapshotSuccess received for snapshot", persistenceId());
 333
 334             long sequenceNumber = success.metadata().sequenceNr();
 335
 336             commitSnapshot(sequenceNumber);
 337
 338         } else if (message instanceof SaveSnapshotFailure) {
 339             SaveSnapshotFailure saveSnapshotFailure = (SaveSnapshotFailure) message;
 340
 341             LOG.error(saveSnapshotFailure.cause(), "{}: SaveSnapshotFailure received for snapshot Cause:",
 342                     persistenceId());
 343
 344             context.getReplicatedLog().snapshotRollback();
 345
 346             LOG.info("{}: Replicated Log rollbacked. Snapshot will be attempted in the next cycle." +
 347                 "snapshotIndex:{}, snapshotTerm:{}, log-size:{}", persistenceId(),
 348                 context.getReplicatedLog().getSnapshotIndex(),
 349                 context.getReplicatedLog().getSnapshotTerm(),
 350                 context.getReplicatedLog().size());
 351
 352         } else if (message instanceof CaptureSnapshot) {
 353             LOG.info("{}: CaptureSnapshot received by actor", persistenceId());
 354
 355             if(captureSnapshot == null) {
 356                 captureSnapshot = (CaptureSnapshot)message;
 357                 createSnapshot();
 358             }
 359
 360         } else if (message instanceof CaptureSnapshotReply){
 361             handleCaptureSnapshotReply(((CaptureSnapshotReply) message).getSnapshot());
 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: {}", persistenceId(), 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
 383         String oldBehaviorLeaderId = oldBehavior == null? null : oldBehavior.getLeaderId();
 384         String oldBehaviorState = oldBehavior == null? null : oldBehavior.state().name();
 385
 386         // it can happen that the state has not changed but the leader has changed.
 387         onLeaderChanged(oldBehaviorLeaderId, currentBehavior.getLeaderId());
 388
 389         if (getRoleChangeNotifier().isPresent() &&
 390                 (oldBehavior == null || (oldBehavior.state() != currentBehavior.state()))) {
 391             getRoleChangeNotifier().get().tell(
 392                     new RoleChanged(getId(), oldBehaviorState , currentBehavior.state().name()),
 393                     getSelf());
 394         }
 395     }
 396
 397     /**
 398      * When a derived RaftActor needs to persist something it must call
 399      * persistData.
 400      *
 401      * @param clientActor
 402      * @param identifier
 403      * @param data
 404      */
 405     protected void persistData(final ActorRef clientActor, final String identifier,
 406         final Payload data) {
 407
 408         ReplicatedLogEntry replicatedLogEntry = new ReplicatedLogImplEntry(
 409             context.getReplicatedLog().lastIndex() + 1,
 410             context.getTermInformation().getCurrentTerm(), data);
 411
 412         if(LOG.isDebugEnabled()) {
 413             LOG.debug("{}: Persist data {}", persistenceId(), replicatedLogEntry);
 414         }
 415
 416         final RaftActorContext raftContext = getRaftActorContext();
 417
 418         replicatedLog
 419                 .appendAndPersist(replicatedLogEntry, new Procedure<ReplicatedLogEntry>() {
 420                     @Override
 421                     public void apply(ReplicatedLogEntry replicatedLogEntry) throws Exception {
 422                         if(!hasFollowers()){
 423                             // Increment the Commit Index and the Last Applied values
 424                             raftContext.setCommitIndex(replicatedLogEntry.getIndex());
 425                             raftContext.setLastApplied(replicatedLogEntry.getIndex());
 426
 427                             // Apply the state immediately
 428                             applyState(clientActor, identifier, data);
 429
 430                             // Send a ApplyLogEntries message so that we write the fact that we applied
 431                             // the state to durable storage
 432                             self().tell(new ApplyLogEntries((int) replicatedLogEntry.getIndex()), self());
 433
 434                             // Check if the "real" snapshot capture has been initiated. If no then do the fake snapshot
 435                             if(!context.isSnapshotCaptureInitiated()){
 436                                 raftContext.getReplicatedLog().snapshotPreCommit(raftContext.getLastApplied(),
 437                                         raftContext.getTermInformation().getCurrentTerm());
 438                                 raftContext.getReplicatedLog().snapshotCommit();
 439                             } else {
 440                                 LOG.debug("{}: Skipping fake snapshotting for {} because real snapshotting is in progress",
 441                                         persistenceId(), getId());
 442                             }
 443                         } else if (clientActor != null) {
 444                             // Send message for replication
 445                             currentBehavior.handleMessage(getSelf(),
 446                                     new Replicate(clientActor, identifier,
 447                                             replicatedLogEntry)
 448                             );
 449                         }
 450
 451                     }
 452                 });    }
 453
 454     protected String getId() {
 455         return context.getId();
 456     }
 457
 458     /**
 459      * Derived actors can call the isLeader method to check if the current
 460      * RaftActor is the Leader or not
 461      *
 462      * @return true it this RaftActor is a Leader false otherwise
 463      */
 464     protected boolean isLeader() {
 465         return context.getId().equals(currentBehavior.getLeaderId());
 466     }
 467
 468     /**
 469      * Derived actor can call getLeader if they need a reference to the Leader.
 470      * This would be useful for example in forwarding a request to an actor
 471      * which is the leader
 472      *
 473      * @return A reference to the leader if known, null otherwise
 474      */
 475     protected ActorSelection getLeader(){
 476         String leaderAddress = getLeaderAddress();
 477
 478         if(leaderAddress == null){
 479             return null;
 480         }
 481
 482         return context.actorSelection(leaderAddress);
 483     }
 484
 485     /**
 486      *
 487      * @return the current leader's id
 488      */
 489     protected String getLeaderId(){
 490         return currentBehavior.getLeaderId();
 491     }
 492
 493     protected RaftState getRaftState() {
 494         return currentBehavior.state();
 495     }
 496
 497     protected ReplicatedLogEntry getLastLogEntry() {
 498         return replicatedLog.last();
 499     }
 500
 501     protected Long getCurrentTerm(){
 502         return context.getTermInformation().getCurrentTerm();
 503     }
 504
 505     protected Long getCommitIndex(){
 506         return context.getCommitIndex();
 507     }
 508
 509     protected Long getLastApplied(){
 510         return context.getLastApplied();
 511     }
 512
 513     protected RaftActorContext getRaftActorContext() {
 514         return context;
 515     }
 516
 517     /**
 518      * setPeerAddress sets the address of a known peer at a later time.
 519      * <p>
 520      * This is to account for situations where a we know that a peer
 521      * exists but we do not know an address up-front. This may also be used in
 522      * situations where a known peer starts off in a different location and we
 523      * need to change it's address
 524      * <p>
 525      * Note that if the peerId does not match the list of peers passed to
 526      * this actor during construction an IllegalStateException will be thrown.
 527      *
 528      * @param peerId
 529      * @param peerAddress
 530      */
 531     protected void setPeerAddress(String peerId, String peerAddress){
 532         context.setPeerAddress(peerId, peerAddress);
 533     }
 534
 535     protected void commitSnapshot(long sequenceNumber) {
 536         context.getReplicatedLog().snapshotCommit();
 537
 538         // TODO: Not sure if we want to be this aggressive with trimming stuff
 539         trimPersistentData(sequenceNumber);
 540     }
 541
 542     /**
 543      * The applyState method will be called by the RaftActor when some data
 544      * needs to be applied to the actor's state
 545      *
 546      * @param clientActor A reference to the client who sent this message. This
 547      *                    is the same reference that was passed to persistData
 548      *                    by the derived actor. clientActor may be null when
 549      *                    the RaftActor is behaving as a follower or during
 550      *                    recovery.
 551      * @param identifier  The identifier of the persisted data. This is also
 552      *                    the same identifier that was passed to persistData by
 553      *                    the derived actor. identifier may be null when
 554      *                    the RaftActor is behaving as a follower or during
 555      *                    recovery
 556      * @param data        A piece of data that was persisted by the persistData call.
 557      *                    This should NEVER be null.
 558      */
 559     protected abstract void applyState(ActorRef clientActor, String identifier,
 560         Object data);
 561
 562     /**
 563      * This method is called during recovery at the start of a batch of state entries. Derived
 564      * classes should perform any initialization needed to start a batch.
 565      */
 566     protected abstract void startLogRecoveryBatch(int maxBatchSize);
 567
 568     /**
 569      * This method is called during recovery to append state data to the current batch. This method
 570      * is called 1 or more times after {@link #startLogRecoveryBatch}.
 571      *
 572      * @param data the state data
 573      */
 574     protected abstract void appendRecoveredLogEntry(Payload data);
 575
 576     /**
 577      * This method is called during recovery to reconstruct the state of the actor.
 578      *
 579      * @param snapshot A snapshot of the state of the actor
 580      */
 581     protected abstract void applyRecoverySnapshot(byte[] snapshotBytes);
 582
 583     /**
 584      * This method is called during recovery at the end of a batch to apply the current batched
 585      * log entries. This method is called after {@link #appendRecoveredLogEntry}.
 586      */
 587     protected abstract void applyCurrentLogRecoveryBatch();
 588
 589     /**
 590      * This method is called when recovery is complete.
 591      */
 592     protected abstract void onRecoveryComplete();
 593
 594     /**
 595      * This method will be called by the RaftActor when a snapshot needs to be
 596      * created. The derived actor should respond with its current state.
 597      * <p/>
 598      * During recovery the state that is returned by the derived actor will
 599      * be passed back to it by calling the applySnapshot  method
 600      *
 601      * @return The current state of the actor
 602      */
 603     protected abstract void createSnapshot();
 604
 605     /**
 606      * This method can be called at any other point during normal
 607      * operations when the derived actor is out of sync with it's peers
 608      * and the only way to bring it in sync is by applying a snapshot
 609      *
 610      * @param snapshotBytes A snapshot of the state of the actor
 611      */
 612     protected abstract void applySnapshot(byte[] snapshotBytes);
 613
 614     /**
 615      * This method will be called by the RaftActor when the state of the
 616      * RaftActor changes. The derived actor can then use methods like
 617      * isLeader or getLeader to do something useful
 618      */
 619     protected abstract void onStateChanged();
 620
 621     protected abstract DataPersistenceProvider persistence();
 622
 623     /**
 624      * Notifier Actor for this RaftActor to notify when a role change happens
 625      * @return ActorRef - ActorRef of the notifier or Optional.absent if none.
 626      */
 627     protected abstract Optional<ActorRef> getRoleChangeNotifier();
 628
 629     protected void onLeaderChanged(String oldLeader, String newLeader){};
 630
 631     private void trimPersistentData(long sequenceNumber) {
 632         // Trim akka snapshots
 633         // FIXME : Not sure how exactly the SnapshotSelectionCriteria is applied
 634         // For now guessing that it is ANDed.
 635         persistence().deleteSnapshots(new SnapshotSelectionCriteria(
 636             sequenceNumber - context.getConfigParams().getSnapshotBatchCount(), 43200000));
 637
 638         // Trim akka journal
 639         persistence().deleteMessages(sequenceNumber);
 640     }
 641
 642     private String getLeaderAddress(){
 643         if(isLeader()){
 644             return getSelf().path().toString();
 645         }
 646         String leaderId = currentBehavior.getLeaderId();
 647         if (leaderId == null) {
 648             return null;
 649         }
 650         String peerAddress = context.getPeerAddress(leaderId);
 651         if(LOG.isDebugEnabled()) {
 652             LOG.debug("{}: getLeaderAddress leaderId = {} peerAddress = {}",
 653                     persistenceId(), leaderId, peerAddress);
 654         }
 655
 656         return peerAddress;
 657     }
 658
 659     private void handleCaptureSnapshotReply(byte[] snapshotBytes) {
 660         LOG.info("{}: CaptureSnapshotReply received by actor: snapshot size {}", persistenceId(), snapshotBytes.length);
 661
 662         // create a snapshot object from the state provided and save it
 663         // when snapshot is saved async, SaveSnapshotSuccess is raised.
 664
 665         Snapshot sn = Snapshot.create(snapshotBytes,
 666             context.getReplicatedLog().getFrom(captureSnapshot.getLastAppliedIndex() + 1),
 667             captureSnapshot.getLastIndex(), captureSnapshot.getLastTerm(),
 668             captureSnapshot.getLastAppliedIndex(), captureSnapshot.getLastAppliedTerm());
 669
 670         persistence().saveSnapshot(sn);
 671
 672         LOG.info("{}: Persisting of snapshot done:{}", persistenceId(), sn.getLogMessage());
 673
 674         //be greedy and remove entries from in-mem journal which are in the snapshot
 675         // and update snapshotIndex and snapshotTerm without waiting for the success,
 676
 677         context.getReplicatedLog().snapshotPreCommit(
 678             captureSnapshot.getLastAppliedIndex(),
 679             captureSnapshot.getLastAppliedTerm());
 680
 681         LOG.info("{}: Removed in-memory snapshotted entries, adjusted snaphsotIndex:{} " +
 682             "and term:{}", persistenceId(), captureSnapshot.getLastAppliedIndex(),
 683             captureSnapshot.getLastAppliedTerm());
 684
 685         if (isLeader() && captureSnapshot.isInstallSnapshotInitiated()) {
 686             // this would be call straight to the leader and won't initiate in serialization
 687             currentBehavior.handleMessage(getSelf(), new SendInstallSnapshot(
 688                     ByteString.copyFrom(snapshotBytes)));
 689         }
 690
 691         captureSnapshot = null;
 692         context.setSnapshotCaptureInitiated(false);
 693     }
 694
 695     protected boolean hasFollowers(){
 696         return getRaftActorContext().getPeerAddresses().keySet().size() > 0;
 697     }
 698
 699     private class ReplicatedLogImpl extends AbstractReplicatedLogImpl {
 700
 701         private static final int DATA_SIZE_DIVIDER = 5;
 702         private long dataSizeSinceLastSnapshot = 0;
 703
 704         public ReplicatedLogImpl(Snapshot snapshot) {
 705             super(snapshot.getLastAppliedIndex(), snapshot.getLastAppliedTerm(),
 706                 snapshot.getUnAppliedEntries());
 707         }
 708
 709         public ReplicatedLogImpl() {
 710             super();
 711         }
 712
 713         @Override public void removeFromAndPersist(long logEntryIndex) {
 714             int adjustedIndex = adjustedIndex(logEntryIndex);
 715
 716             if (adjustedIndex < 0) {
 717                 return;
 718             }
 719
 720             // FIXME: Maybe this should be done after the command is saved
 721             journal.subList(adjustedIndex , journal.size()).clear();
 722
 723             persistence().persist(new DeleteEntries(adjustedIndex), new Procedure<DeleteEntries>(){
 724
 725                 @Override public void apply(DeleteEntries param)
 726                     throws Exception {
 727                     //FIXME : Doing nothing for now
 728                     dataSize = 0;
 729                     for(ReplicatedLogEntry entry : journal){
 730                         dataSize += entry.size();
 731                     }
 732                 }
 733             });
 734         }
 735
 736         @Override public void appendAndPersist(
 737             final ReplicatedLogEntry replicatedLogEntry) {
 738             appendAndPersist(replicatedLogEntry, null);
 739         }
 740
 741         @Override
 742         public int dataSize() {
 743             return dataSize;
 744         }
 745
 746         public void appendAndPersist(
 747             final ReplicatedLogEntry replicatedLogEntry,
 748             final Procedure<ReplicatedLogEntry> callback)  {
 749
 750             if(LOG.isDebugEnabled()) {
 751                 LOG.debug("{}: Append log entry and persist {} ", persistenceId(), replicatedLogEntry);
 752             }
 753
 754             // FIXME : By adding the replicated log entry to the in-memory journal we are not truly ensuring durability of the logs
 755             journal.add(replicatedLogEntry);
 756
 757             // When persisting events with persist it is guaranteed that the
 758             // persistent actor will not receive further commands between the
 759             // persist call and the execution(s) of the associated event
 760             // handler. This also holds for multiple persist calls in context
 761             // of a single command.
 762             persistence().persist(replicatedLogEntry,
 763                 new Procedure<ReplicatedLogEntry>() {
 764                     @Override
 765                     public void apply(ReplicatedLogEntry evt) throws Exception {
 766                         int logEntrySize = replicatedLogEntry.size();
 767
 768                         dataSize += logEntrySize;
 769                         long dataSizeForCheck = dataSize;
 770
 771                         dataSizeSinceLastSnapshot += logEntrySize;
 772                         long journalSize = lastIndex()+1;
 773
 774                         if(!hasFollowers()) {
 775                             // When we do not have followers we do not maintain an in-memory log
 776                             // due to this the journalSize will never become anything close to the
 777                             // snapshot batch count. In fact will mostly be 1.
 778                             // Similarly since the journal's dataSize depends on the entries in the
 779                             // journal the journal's dataSize will never reach a value close to the
 780                             // memory threshold.
 781                             // By maintaining the dataSize outside the journal we are tracking essentially
 782                             // what we have written to the disk however since we no longer are in
 783                             // need of doing a snapshot just for the sake of freeing up memory we adjust
 784                             // the real size of data by the DATA_SIZE_DIVIDER so that we do not snapshot as often
 785                             // as if we were maintaining a real snapshot
 786                             dataSizeForCheck = dataSizeSinceLastSnapshot / DATA_SIZE_DIVIDER;
 787                         }
 788
 789                         long dataThreshold = Runtime.getRuntime().totalMemory() *
 790                                 getRaftActorContext().getConfigParams().getSnapshotDataThresholdPercentage() / 100;
 791
 792                         // when a snaphsot is being taken, captureSnapshot != null
 793                         if (!context.isSnapshotCaptureInitiated() &&
 794                                 ( journalSize % context.getConfigParams().getSnapshotBatchCount() == 0 ||
 795                                         dataSizeForCheck > dataThreshold)) {
 796
 797                             dataSizeSinceLastSnapshot = 0;
 798
 799                             LOG.info("{}: Initiating Snapshot Capture..", persistenceId());
 800                             long lastAppliedIndex = -1;
 801                             long lastAppliedTerm = -1;
 802
 803                             ReplicatedLogEntry lastAppliedEntry = get(context.getLastApplied());
 804                             if (!hasFollowers()) {
 805                                 lastAppliedIndex = replicatedLogEntry.getIndex();
 806                                 lastAppliedTerm = replicatedLogEntry.getTerm();
 807                             } else if (lastAppliedEntry != null) {
 808                                 lastAppliedIndex = lastAppliedEntry.getIndex();
 809                                 lastAppliedTerm = lastAppliedEntry.getTerm();
 810                             }
 811
 812                             if(LOG.isDebugEnabled()) {
 813                                 LOG.debug("{}: Snapshot Capture logSize: {}", persistenceId(), journal.size());
 814                                 LOG.debug("{}: Snapshot Capture lastApplied:{} ",
 815                                         persistenceId(), context.getLastApplied());
 816                                 LOG.debug("{}: Snapshot Capture lastAppliedIndex:{}", persistenceId(),
 817                                         lastAppliedIndex);
 818                                 LOG.debug("{}: Snapshot Capture lastAppliedTerm:{}", persistenceId(),
 819                                         lastAppliedTerm);
 820                             }
 821
 822                             // send a CaptureSnapshot to self to make the expensive operation async.
 823                             getSelf().tell(new CaptureSnapshot(
 824                                 lastIndex(), lastTerm(), lastAppliedIndex, lastAppliedTerm),
 825                                 null);
 826                             context.setSnapshotCaptureInitiated(true);
 827                         }
 828                         if(callback != null){
 829                             callback.apply(replicatedLogEntry);
 830                         }
 831                     }
 832                 }
 833             );
 834         }
 835
 836     }
 837
 838     static class DeleteEntries implements Serializable {
 839         private static final long serialVersionUID = 1L;
 840         private final int fromIndex;
 841
 842         public DeleteEntries(int fromIndex) {
 843             this.fromIndex = fromIndex;
 844         }
 845
 846         public int getFromIndex() {
 847             return fromIndex;
 848         }
 849     }
 850
 851
 852     private class ElectionTermImpl implements ElectionTerm {
 853         /**
 854          * Identifier of the actor whose election term information this is
 855          */
 856         private long currentTerm = 0;
 857         private String votedFor = null;
 858
 859         @Override
 860         public long getCurrentTerm() {
 861             return currentTerm;
 862         }
 863
 864         @Override
 865         public String getVotedFor() {
 866             return votedFor;
 867         }
 868
 869         @Override public void update(long currentTerm, String votedFor) {
 870             if(LOG.isDebugEnabled()) {
 871                 LOG.debug("{}: Set currentTerm={}, votedFor={}", persistenceId(), currentTerm, votedFor);
 872             }
 873             this.currentTerm = currentTerm;
 874             this.votedFor = votedFor;
 875         }
 876
 877         @Override
 878         public void updateAndPersist(long currentTerm, String votedFor){
 879             update(currentTerm, votedFor);
 880             // FIXME : Maybe first persist then update the state
 881             persistence().persist(new UpdateElectionTerm(this.currentTerm, this.votedFor), new Procedure<UpdateElectionTerm>(){
 882
 883                 @Override public void apply(UpdateElectionTerm param)
 884                     throws Exception {
 885
 886                 }
 887             });
 888         }
 889     }
 890
 891     static class UpdateElectionTerm implements Serializable {
 892         private static final long serialVersionUID = 1L;
 893         private final long currentTerm;
 894         private final String votedFor;
 895
 896         public UpdateElectionTerm(long currentTerm, String votedFor) {
 897             this.currentTerm = currentTerm;
 898             this.votedFor = votedFor;
 899         }
 900
 901         public long getCurrentTerm() {
 902             return currentTerm;
 903         }
 904
 905         public String getVotedFor() {
 906             return votedFor;
 907         }
 908     }
 909
 910     protected class NonPersistentRaftDataProvider extends NonPersistentDataProvider {
 911
 912         public NonPersistentRaftDataProvider(){
 913
 914         }
 915
 916         /**
 917          * The way snapshotting works is,
 918          * <ol>
 919          * <li> RaftActor calls createSnapshot on the Shard
 920          * <li> Shard sends a CaptureSnapshotReply and RaftActor then calls saveSnapshot
 921          * <li> When saveSnapshot is invoked on the akka-persistence API it uses the SnapshotStore to save the snapshot.
 922          * The SnapshotStore sends SaveSnapshotSuccess or SaveSnapshotFailure. When the RaftActor gets SaveSnapshot
 923          * success it commits the snapshot to the in-memory journal. This commitSnapshot is mimicking what is done
 924          * in SaveSnapshotSuccess.
 925          * </ol>
 926          * @param o
 927          */
 928         @Override
 929         public void saveSnapshot(Object o) {
 930             // Make saving Snapshot successful
 931             commitSnapshot(-1L);
 932         }
 933     }
 934
 935     @VisibleForTesting
 936     void setCurrentBehavior(AbstractRaftActorBehavior behavior) {
 937         currentBehavior = behavior;
 938     }
 939
 940     protected RaftActorBehavior getCurrentBehavior() {
 941         return currentBehavior;
 942     }
 943
 944 }