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 akka.persistence.UntypedPersistentActor;
  22 import com.google.common.base.Optional;
  23 import com.google.common.base.Stopwatch;
  24 import com.google.protobuf.ByteString;
  25 import org.opendaylight.controller.cluster.raft.base.messages.ApplyLogEntries;
  26 import org.opendaylight.controller.cluster.raft.base.messages.ApplySnapshot;
  27 import org.opendaylight.controller.cluster.raft.base.messages.ApplyState;
  28 import org.opendaylight.controller.cluster.raft.base.messages.CaptureSnapshot;
  29 import org.opendaylight.controller.cluster.raft.base.messages.CaptureSnapshotReply;
  30 import org.opendaylight.controller.cluster.raft.base.messages.Replicate;
  31 import org.opendaylight.controller.cluster.raft.base.messages.SendHeartBeat;
  32 import org.opendaylight.controller.cluster.raft.behaviors.Follower;
  33 import org.opendaylight.controller.cluster.raft.behaviors.RaftActorBehavior;
  34 import org.opendaylight.controller.cluster.raft.client.messages.AddRaftPeer;
  35 import org.opendaylight.controller.cluster.raft.client.messages.FindLeader;
  36 import org.opendaylight.controller.cluster.raft.client.messages.FindLeaderReply;
  37 import org.opendaylight.controller.cluster.raft.client.messages.RemoveRaftPeer;
  38 import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
  39 import org.opendaylight.controller.cluster.raft.protobuff.client.messages.Payload;
  40 import org.opendaylight.controller.protobuff.messages.cluster.raft.AppendEntriesMessages;
  41 import java.io.Serializable;
  42 import java.util.Map;
  43
  44 /**
  45  * RaftActor encapsulates a state machine that needs to be kept synchronized
  46  * in a cluster. It implements the RAFT algorithm as described in the paper
  47  * <a href='https://ramcloud.stanford.edu/wiki/download/attachments/11370504/raft.pdf'>
  48  * In Search of an Understandable Consensus Algorithm</a>
  49  * <p/>
  50  * RaftActor has 3 states and each state has a certain behavior associated
  51  * with it. A Raft actor can behave as,
  52  * <ul>
  53  * <li> A Leader </li>
  54  * <li> A Follower (or) </li>
  55  * <li> A Candidate </li>
  56  * </ul>
  57  * <p/>
  58  * <p/>
  59  * A RaftActor MUST be a Leader in order to accept requests from clients to
  60  * change the state of it's encapsulated state machine. Once a RaftActor becomes
  61  * a Leader it is also responsible for ensuring that all followers ultimately
  62  * have the same log and therefore the same state machine as itself.
  63  * <p/>
  64  * <p/>
  65  * The current behavior of a RaftActor determines how election for leadership
  66  * is initiated and how peer RaftActors react to request for votes.
  67  * <p/>
  68  * <p/>
  69  * Each RaftActor also needs to know the current election term. It uses this
  70  * information for a couple of things. One is to simply figure out who it
  71  * voted for in the last election. Another is to figure out if the message
  72  * it received to update it's state is stale.
  73  * <p/>
  74  * <p/>
  75  * The RaftActor uses akka-persistence to store it's replicated log.
  76  * Furthermore through it's behaviors a Raft Actor determines
  77  * <p/>
  78  * <ul>
  79  * <li> when a log entry should be persisted </li>
  80  * <li> when a log entry should be applied to the state machine (and) </li>
  81  * <li> when a snapshot should be saved </li>
  82  * </ul>
  83  */
  84 public abstract class RaftActor extends UntypedPersistentActor {
  85     protected final LoggingAdapter LOG =
  86         Logging.getLogger(getContext().system(), this);
  87
  88     /**
  89      * The current state determines the current behavior of a RaftActor
  90      * A Raft Actor always starts off in the Follower State
  91      */
  92     private RaftActorBehavior currentBehavior;
  93
  94     /**
  95      * This context should NOT be passed directly to any other actor it is
  96      * only to be consumed by the RaftActorBehaviors
  97      */
  98     private final RaftActorContext context;
  99
 100     /**
 101      * The in-memory journal
 102      */
 103     private ReplicatedLogImpl replicatedLog = new ReplicatedLogImpl();
 104
 105     private CaptureSnapshot captureSnapshot = null;
 106
 107     private volatile boolean hasSnapshotCaptureInitiated = false;
 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         super.preStart();
 139     }
 140
 141     @Override
 142     public void onReceiveRecover(Object message) {
 143         if (message instanceof SnapshotOffer) {
 144             onRecoveredSnapshot((SnapshotOffer)message);
 145         } else if (message instanceof ReplicatedLogEntry) {
 146             onRecoveredJournalLogEntry((ReplicatedLogEntry)message);
 147         } else if (message instanceof ApplyLogEntries) {
 148             onRecoveredApplyLogEntries((ApplyLogEntries)message);
 149         } else if (message instanceof DeleteEntries) {
 150             replicatedLog.removeFrom(((DeleteEntries) message).getFromIndex());
 151         } else if (message instanceof UpdateElectionTerm) {
 152             context.getTermInformation().update(((UpdateElectionTerm) message).getCurrentTerm(),
 153                     ((UpdateElectionTerm) message).getVotedFor());
 154         } else if (message instanceof RecoveryCompleted) {
 155             onRecoveryCompletedMessage();
 156         }
 157     }
 158
 159     private void onRecoveredSnapshot(SnapshotOffer offer) {
 160         if(LOG.isDebugEnabled()) {
 161             LOG.debug("SnapshotOffer called..");
 162         }
 163
 164         initRecoveryTimer();
 165
 166         Snapshot snapshot = (Snapshot) offer.snapshot();
 167
 168         // Create a replicated log with the snapshot information
 169         // The replicated log can be used later on to retrieve this snapshot
 170         // when we need to install it on a peer
 171         replicatedLog = new ReplicatedLogImpl(snapshot);
 172
 173         context.setReplicatedLog(replicatedLog);
 174         context.setLastApplied(snapshot.getLastAppliedIndex());
 175         context.setCommitIndex(snapshot.getLastAppliedIndex());
 176
 177         Stopwatch timer = new Stopwatch();
 178         timer.start();
 179
 180         // Apply the snapshot to the actors state
 181         applyRecoverySnapshot(ByteString.copyFrom(snapshot.getState()));
 182
 183         timer.stop();
 184         LOG.info("Recovery snapshot applied for {} in {}: snapshotIndex={}, snapshotTerm={}, journal-size=" +
 185                 replicatedLog.size(), persistenceId(), timer.toString(),
 186                 replicatedLog.snapshotIndex, replicatedLog.snapshotTerm);
 187     }
 188
 189     private void onRecoveredJournalLogEntry(ReplicatedLogEntry logEntry) {
 190         if(LOG.isDebugEnabled()) {
 191             LOG.debug("Received ReplicatedLogEntry for recovery: {}", logEntry.getIndex());
 192         }
 193
 194         replicatedLog.append(logEntry);
 195     }
 196
 197     private void onRecoveredApplyLogEntries(ApplyLogEntries ale) {
 198         if(LOG.isDebugEnabled()) {
 199             LOG.debug("Received ApplyLogEntries for recovery, applying to state: {} to {}",
 200                     context.getLastApplied() + 1, ale.getToIndex());
 201         }
 202
 203         for (long i = context.getLastApplied() + 1; i <= ale.getToIndex(); i++) {
 204             batchRecoveredLogEntry(replicatedLog.get(i));
 205         }
 206
 207         context.setLastApplied(ale.getToIndex());
 208         context.setCommitIndex(ale.getToIndex());
 209     }
 210
 211     private void batchRecoveredLogEntry(ReplicatedLogEntry logEntry) {
 212         initRecoveryTimer();
 213
 214         int batchSize = context.getConfigParams().getJournalRecoveryLogBatchSize();
 215         if(currentRecoveryBatchCount == 0) {
 216             startLogRecoveryBatch(batchSize);
 217         }
 218
 219         appendRecoveredLogEntry(logEntry.getData());
 220
 221         if(++currentRecoveryBatchCount >= batchSize) {
 222             endCurrentLogRecoveryBatch();
 223         }
 224     }
 225
 226     private void endCurrentLogRecoveryBatch() {
 227         applyCurrentLogRecoveryBatch();
 228         currentRecoveryBatchCount = 0;
 229     }
 230
 231     private void onRecoveryCompletedMessage() {
 232         if(currentRecoveryBatchCount > 0) {
 233             endCurrentLogRecoveryBatch();
 234         }
 235
 236         onRecoveryComplete();
 237
 238         String recoveryTime = "";
 239         if(recoveryTimer != null) {
 240             recoveryTimer.stop();
 241             recoveryTime = " in " + recoveryTimer.toString();
 242             recoveryTimer = null;
 243         }
 244
 245         LOG.info(
 246             "Recovery completed" + recoveryTime + " - Switching actor to Follower - " +
 247                 "Persistence Id =  " + persistenceId() +
 248                 " Last index in log={}, snapshotIndex={}, snapshotTerm={}, " +
 249                 "journal-size={}",
 250             replicatedLog.lastIndex(), replicatedLog.snapshotIndex,
 251             replicatedLog.snapshotTerm, replicatedLog.size());
 252
 253         currentBehavior = new Follower(context);
 254         onStateChanged();
 255     }
 256
 257     @Override public void onReceiveCommand(Object message) {
 258         if (message instanceof ApplyState){
 259             ApplyState applyState = (ApplyState) message;
 260
 261             if(LOG.isDebugEnabled()) {
 262                 LOG.debug("Applying state for log index {} data {}",
 263                     applyState.getReplicatedLogEntry().getIndex(),
 264                     applyState.getReplicatedLogEntry().getData());
 265             }
 266
 267             applyState(applyState.getClientActor(), applyState.getIdentifier(),
 268                 applyState.getReplicatedLogEntry().getData());
 269
 270         } else if (message instanceof ApplyLogEntries){
 271             ApplyLogEntries ale = (ApplyLogEntries) message;
 272             if(LOG.isDebugEnabled()) {
 273                 LOG.debug("Persisting ApplyLogEntries with index={}", ale.getToIndex());
 274             }
 275             persist(new ApplyLogEntries(ale.getToIndex()), new Procedure<ApplyLogEntries>() {
 276                 @Override
 277                 public void apply(ApplyLogEntries param) throws Exception {
 278                 }
 279             });
 280
 281         } else if(message instanceof ApplySnapshot ) {
 282             Snapshot snapshot = ((ApplySnapshot) message).getSnapshot();
 283
 284             if(LOG.isDebugEnabled()) {
 285                 LOG.debug("ApplySnapshot called on Follower Actor " +
 286                         "snapshotIndex:{}, snapshotTerm:{}", snapshot.getLastAppliedIndex(),
 287                     snapshot.getLastAppliedTerm()
 288                 );
 289             }
 290             applySnapshot(ByteString.copyFrom(snapshot.getState()));
 291
 292             //clears the followers log, sets the snapshot index to ensure adjusted-index works
 293             replicatedLog = new ReplicatedLogImpl(snapshot);
 294             context.setReplicatedLog(replicatedLog);
 295             context.setLastApplied(snapshot.getLastAppliedIndex());
 296
 297         } else if (message instanceof FindLeader) {
 298             getSender().tell(
 299                 new FindLeaderReply(getLeaderAddress()),
 300                 getSelf()
 301             );
 302
 303         } else if (message instanceof SaveSnapshotSuccess) {
 304             SaveSnapshotSuccess success = (SaveSnapshotSuccess) message;
 305             LOG.info("SaveSnapshotSuccess received for snapshot");
 306
 307             context.getReplicatedLog().snapshotCommit();
 308
 309             // TODO: Not sure if we want to be this aggressive with trimming stuff
 310             trimPersistentData(success.metadata().sequenceNr());
 311
 312         } else if (message instanceof SaveSnapshotFailure) {
 313             SaveSnapshotFailure saveSnapshotFailure = (SaveSnapshotFailure) message;
 314
 315             LOG.info("saveSnapshotFailure.metadata():{}", saveSnapshotFailure.metadata().toString());
 316             LOG.error(saveSnapshotFailure.cause(), "SaveSnapshotFailure received for snapshot Cause:");
 317
 318             context.getReplicatedLog().snapshotRollback();
 319
 320             LOG.info("Replicated Log rollbacked. Snapshot will be attempted in the next cycle." +
 321                 "snapshotIndex:{}, snapshotTerm:{}, log-size:{}",
 322                 context.getReplicatedLog().getSnapshotIndex(),
 323                 context.getReplicatedLog().getSnapshotTerm(),
 324                 context.getReplicatedLog().size());
 325
 326         } else if (message instanceof AddRaftPeer){
 327
 328             // FIXME : Do not add raft peers like this.
 329             // When adding a new Peer we have to ensure that the a majority of
 330             // the peers know about the new Peer. Doing it this way may cause
 331             // a situation where multiple Leaders may emerge
 332             AddRaftPeer arp = (AddRaftPeer)message;
 333            context.addToPeers(arp.getName(), arp.getAddress());
 334
 335         } else if (message instanceof RemoveRaftPeer){
 336
 337             RemoveRaftPeer rrp = (RemoveRaftPeer)message;
 338             context.removePeer(rrp.getName());
 339
 340         } else if (message instanceof CaptureSnapshot) {
 341             LOG.info("CaptureSnapshot received by actor");
 342             CaptureSnapshot cs = (CaptureSnapshot)message;
 343             captureSnapshot = cs;
 344             createSnapshot();
 345
 346         } else if (message instanceof CaptureSnapshotReply){
 347             LOG.info("CaptureSnapshotReply received by actor");
 348             CaptureSnapshotReply csr = (CaptureSnapshotReply) message;
 349
 350             ByteString stateInBytes = csr.getSnapshot();
 351             LOG.info("CaptureSnapshotReply stateInBytes size:{}", stateInBytes.size());
 352             handleCaptureSnapshotReply(stateInBytes);
 353
 354         } else {
 355             if (!(message instanceof AppendEntriesMessages.AppendEntries)
 356                 && !(message instanceof AppendEntriesReply) && !(message instanceof SendHeartBeat)) {
 357                 if(LOG.isDebugEnabled()) {
 358                     LOG.debug("onReceiveCommand: message: {}", message.getClass());
 359                 }
 360             }
 361
 362             RaftActorBehavior oldBehavior = currentBehavior;
 363             currentBehavior = currentBehavior.handleMessage(getSender(), message);
 364
 365             if(oldBehavior != currentBehavior){
 366                 onStateChanged();
 367             }
 368
 369             onLeaderChanged(oldBehavior.getLeaderId(), currentBehavior.getLeaderId());
 370         }
 371     }
 372
 373     public java.util.Set<String> getPeers() {
 374         return context.getPeerAddresses().keySet();
 375     }
 376
 377     protected String getReplicatedLogState() {
 378         return "snapshotIndex=" + context.getReplicatedLog().getSnapshotIndex()
 379             + ", snapshotTerm=" + context.getReplicatedLog().getSnapshotTerm()
 380             + ", im-mem journal size=" + context.getReplicatedLog().size();
 381     }
 382
 383
 384     /**
 385      * When a derived RaftActor needs to persist something it must call
 386      * persistData.
 387      *
 388      * @param clientActor
 389      * @param identifier
 390      * @param data
 391      */
 392     protected void persistData(ActorRef clientActor, String identifier,
 393         Payload data) {
 394
 395         ReplicatedLogEntry replicatedLogEntry = new ReplicatedLogImplEntry(
 396             context.getReplicatedLog().lastIndex() + 1,
 397             context.getTermInformation().getCurrentTerm(), data);
 398
 399         if(LOG.isDebugEnabled()) {
 400             LOG.debug("Persist data {}", replicatedLogEntry);
 401         }
 402
 403         replicatedLog
 404             .appendAndPersist(clientActor, identifier, replicatedLogEntry);
 405     }
 406
 407     protected String getId() {
 408         return context.getId();
 409     }
 410
 411     /**
 412      * Derived actors can call the isLeader method to check if the current
 413      * RaftActor is the Leader or not
 414      *
 415      * @return true it this RaftActor is a Leader false otherwise
 416      */
 417     protected boolean isLeader() {
 418         return context.getId().equals(currentBehavior.getLeaderId());
 419     }
 420
 421     /**
 422      * Derived actor can call getLeader if they need a reference to the Leader.
 423      * This would be useful for example in forwarding a request to an actor
 424      * which is the leader
 425      *
 426      * @return A reference to the leader if known, null otherwise
 427      */
 428     protected ActorSelection getLeader(){
 429         String leaderAddress = getLeaderAddress();
 430
 431         if(leaderAddress == null){
 432             return null;
 433         }
 434
 435         return context.actorSelection(leaderAddress);
 436     }
 437
 438     /**
 439      *
 440      * @return the current leader's id
 441      */
 442     protected String getLeaderId(){
 443         return currentBehavior.getLeaderId();
 444     }
 445
 446     protected RaftState getRaftState() {
 447         return currentBehavior.state();
 448     }
 449
 450     protected ReplicatedLogEntry getLastLogEntry() {
 451         return replicatedLog.last();
 452     }
 453
 454     protected Long getCurrentTerm(){
 455         return context.getTermInformation().getCurrentTerm();
 456     }
 457
 458     protected Long getCommitIndex(){
 459         return context.getCommitIndex();
 460     }
 461
 462     protected Long getLastApplied(){
 463         return context.getLastApplied();
 464     }
 465
 466     protected RaftActorContext getRaftActorContext() {
 467         return context;
 468     }
 469
 470     /**
 471      * setPeerAddress sets the address of a known peer at a later time.
 472      * <p>
 473      * This is to account for situations where a we know that a peer
 474      * exists but we do not know an address up-front. This may also be used in
 475      * situations where a known peer starts off in a different location and we
 476      * need to change it's address
 477      * <p>
 478      * Note that if the peerId does not match the list of peers passed to
 479      * this actor during construction an IllegalStateException will be thrown.
 480      *
 481      * @param peerId
 482      * @param peerAddress
 483      */
 484     protected void setPeerAddress(String peerId, String peerAddress){
 485         context.setPeerAddress(peerId, peerAddress);
 486     }
 487
 488
 489
 490     /**
 491      * The applyState method will be called by the RaftActor when some data
 492      * needs to be applied to the actor's state
 493      *
 494      * @param clientActor A reference to the client who sent this message. This
 495      *                    is the same reference that was passed to persistData
 496      *                    by the derived actor. clientActor may be null when
 497      *                    the RaftActor is behaving as a follower or during
 498      *                    recovery.
 499      * @param identifier  The identifier of the persisted data. This is also
 500      *                    the same identifier that was passed to persistData by
 501      *                    the derived actor. identifier may be null when
 502      *                    the RaftActor is behaving as a follower or during
 503      *                    recovery
 504      * @param data        A piece of data that was persisted by the persistData call.
 505      *                    This should NEVER be null.
 506      */
 507     protected abstract void applyState(ActorRef clientActor, String identifier,
 508         Object data);
 509
 510     /**
 511      * This method is called during recovery at the start of a batch of state entries. Derived
 512      * classes should perform any initialization needed to start a batch.
 513      */
 514     protected abstract void startLogRecoveryBatch(int maxBatchSize);
 515
 516     /**
 517      * This method is called during recovery to append state data to the current batch. This method
 518      * is called 1 or more times after {@link #startRecoveryStateBatch}.
 519      *
 520      * @param data the state data
 521      */
 522     protected abstract void appendRecoveredLogEntry(Payload data);
 523
 524     /**
 525      * This method is called during recovery to reconstruct the state of the actor.
 526      *
 527      * @param snapshot A snapshot of the state of the actor
 528      */
 529     protected abstract void applyRecoverySnapshot(ByteString snapshot);
 530
 531     /**
 532      * This method is called during recovery at the end of a batch to apply the current batched
 533      * log entries. This method is called after {@link #appendRecoveryLogEntry}.
 534      */
 535     protected abstract void applyCurrentLogRecoveryBatch();
 536
 537     /**
 538      * This method is called when recovery is complete.
 539      */
 540     protected abstract void onRecoveryComplete();
 541
 542     /**
 543      * This method will be called by the RaftActor when a snapshot needs to be
 544      * created. The derived actor should respond with its current state.
 545      * <p/>
 546      * During recovery the state that is returned by the derived actor will
 547      * be passed back to it by calling the applySnapshot  method
 548      *
 549      * @return The current state of the actor
 550      */
 551     protected abstract void createSnapshot();
 552
 553     /**
 554      * This method can be called at any other point during normal
 555      * operations when the derived actor is out of sync with it's peers
 556      * and the only way to bring it in sync is by applying a snapshot
 557      *
 558      * @param snapshot A snapshot of the state of the actor
 559      */
 560     protected abstract void applySnapshot(ByteString snapshot);
 561
 562     /**
 563      * This method will be called by the RaftActor when the state of the
 564      * RaftActor changes. The derived actor can then use methods like
 565      * isLeader or getLeader to do something useful
 566      */
 567     protected abstract void onStateChanged();
 568
 569     protected void onLeaderChanged(String oldLeader, String newLeader){};
 570
 571     private void trimPersistentData(long sequenceNumber) {
 572         // Trim akka snapshots
 573         // FIXME : Not sure how exactly the SnapshotSelectionCriteria is applied
 574         // For now guessing that it is ANDed.
 575         deleteSnapshots(new SnapshotSelectionCriteria(
 576             sequenceNumber - context.getConfigParams().getSnapshotBatchCount(), 43200000));
 577
 578         // Trim akka journal
 579         deleteMessages(sequenceNumber);
 580     }
 581
 582     private String getLeaderAddress(){
 583         if(isLeader()){
 584             return getSelf().path().toString();
 585         }
 586         String leaderId = currentBehavior.getLeaderId();
 587         if (leaderId == null) {
 588             return null;
 589         }
 590         String peerAddress = context.getPeerAddress(leaderId);
 591         if(LOG.isDebugEnabled()) {
 592             LOG.debug("getLeaderAddress leaderId = {} peerAddress = {}",
 593                     leaderId, peerAddress);
 594         }
 595
 596         return peerAddress;
 597     }
 598
 599     private void handleCaptureSnapshotReply(ByteString stateInBytes) {
 600         // create a snapshot object from the state provided and save it
 601         // when snapshot is saved async, SaveSnapshotSuccess is raised.
 602
 603         Snapshot sn = Snapshot.create(stateInBytes.toByteArray(),
 604             context.getReplicatedLog().getFrom(captureSnapshot.getLastAppliedIndex() + 1),
 605             captureSnapshot.getLastIndex(), captureSnapshot.getLastTerm(),
 606             captureSnapshot.getLastAppliedIndex(), captureSnapshot.getLastAppliedTerm());
 607
 608         saveSnapshot(sn);
 609
 610         LOG.info("Persisting of snapshot done:{}", sn.getLogMessage());
 611
 612         //be greedy and remove entries from in-mem journal which are in the snapshot
 613         // and update snapshotIndex and snapshotTerm without waiting for the success,
 614
 615         context.getReplicatedLog().snapshotPreCommit(stateInBytes,
 616             captureSnapshot.getLastAppliedIndex(),
 617             captureSnapshot.getLastAppliedTerm());
 618
 619         LOG.info("Removed in-memory snapshotted entries, adjusted snaphsotIndex:{} " +
 620             "and term:{}", captureSnapshot.getLastAppliedIndex(),
 621             captureSnapshot.getLastAppliedTerm());
 622
 623         captureSnapshot = null;
 624         hasSnapshotCaptureInitiated = false;
 625     }
 626
 627
 628     private class ReplicatedLogImpl extends AbstractReplicatedLogImpl {
 629
 630         public ReplicatedLogImpl(Snapshot snapshot) {
 631             super(ByteString.copyFrom(snapshot.getState()),
 632                 snapshot.getLastAppliedIndex(), snapshot.getLastAppliedTerm(),
 633                 snapshot.getUnAppliedEntries());
 634         }
 635
 636         public ReplicatedLogImpl() {
 637             super();
 638         }
 639
 640         @Override public void removeFromAndPersist(long logEntryIndex) {
 641             int adjustedIndex = adjustedIndex(logEntryIndex);
 642
 643             if (adjustedIndex < 0) {
 644                 return;
 645             }
 646
 647             // FIXME: Maybe this should be done after the command is saved
 648             journal.subList(adjustedIndex , journal.size()).clear();
 649
 650             persist(new DeleteEntries(adjustedIndex), new Procedure<DeleteEntries>(){
 651
 652                 @Override public void apply(DeleteEntries param)
 653                     throws Exception {
 654                     //FIXME : Doing nothing for now
 655                 }
 656             });
 657         }
 658
 659         @Override public void appendAndPersist(
 660             final ReplicatedLogEntry replicatedLogEntry) {
 661             appendAndPersist(null, null, replicatedLogEntry);
 662         }
 663
 664         public void appendAndPersist(final ActorRef clientActor,
 665             final String identifier,
 666             final ReplicatedLogEntry replicatedLogEntry) {
 667
 668             if(LOG.isDebugEnabled()) {
 669                 LOG.debug("Append log entry and persist {} ", replicatedLogEntry);
 670             }
 671
 672             // FIXME : By adding the replicated log entry to the in-memory journal we are not truly ensuring durability of the logs
 673             journal.add(replicatedLogEntry);
 674
 675             // When persisting events with persist it is guaranteed that the
 676             // persistent actor will not receive further commands between the
 677             // persist call and the execution(s) of the associated event
 678             // handler. This also holds for multiple persist calls in context
 679             // of a single command.
 680             persist(replicatedLogEntry,
 681                 new Procedure<ReplicatedLogEntry>() {
 682                     @Override
 683                     public void apply(ReplicatedLogEntry evt) throws Exception {
 684                         // when a snaphsot is being taken, captureSnapshot != null
 685                         if (hasSnapshotCaptureInitiated == false &&
 686                             journal.size() % context.getConfigParams().getSnapshotBatchCount() == 0) {
 687
 688                             LOG.info("Initiating Snapshot Capture..");
 689                             long lastAppliedIndex = -1;
 690                             long lastAppliedTerm = -1;
 691
 692                             ReplicatedLogEntry lastAppliedEntry = get(context.getLastApplied());
 693                             if (lastAppliedEntry != null) {
 694                                 lastAppliedIndex = lastAppliedEntry.getIndex();
 695                                 lastAppliedTerm = lastAppliedEntry.getTerm();
 696                             }
 697
 698                             if(LOG.isDebugEnabled()) {
 699                                 LOG.debug("Snapshot Capture logSize: {}", journal.size());
 700                                 LOG.debug("Snapshot Capture lastApplied:{} ",
 701                                     context.getLastApplied());
 702                                 LOG.debug("Snapshot Capture lastAppliedIndex:{}", lastAppliedIndex);
 703                                 LOG.debug("Snapshot Capture lastAppliedTerm:{}", lastAppliedTerm);
 704                             }
 705
 706                             // send a CaptureSnapshot to self to make the expensive operation async.
 707                             getSelf().tell(new CaptureSnapshot(
 708                                 lastIndex(), lastTerm(), lastAppliedIndex, lastAppliedTerm),
 709                                 null);
 710                             hasSnapshotCaptureInitiated = true;
 711                         }
 712                         // Send message for replication
 713                         if (clientActor != null) {
 714                             currentBehavior.handleMessage(getSelf(),
 715                                 new Replicate(clientActor, identifier,
 716                                     replicatedLogEntry)
 717                             );
 718                         }
 719                     }
 720                 }
 721             );
 722         }
 723
 724     }
 725
 726     private static class DeleteEntries implements Serializable {
 727         private final int fromIndex;
 728
 729
 730         public DeleteEntries(int fromIndex) {
 731             this.fromIndex = fromIndex;
 732         }
 733
 734         public int getFromIndex() {
 735             return fromIndex;
 736         }
 737     }
 738
 739
 740     private class ElectionTermImpl implements ElectionTerm {
 741         /**
 742          * Identifier of the actor whose election term information this is
 743          */
 744         private long currentTerm = 0;
 745         private String votedFor = null;
 746
 747         @Override
 748         public long getCurrentTerm() {
 749             return currentTerm;
 750         }
 751
 752         @Override
 753         public String getVotedFor() {
 754             return votedFor;
 755         }
 756
 757         @Override public void update(long currentTerm, String votedFor) {
 758             if(LOG.isDebugEnabled()) {
 759                 LOG.debug("Set currentTerm={}, votedFor={}", currentTerm, votedFor);
 760             }
 761             this.currentTerm = currentTerm;
 762             this.votedFor = votedFor;
 763         }
 764
 765         @Override
 766         public void updateAndPersist(long currentTerm, String votedFor){
 767             update(currentTerm, votedFor);
 768             // FIXME : Maybe first persist then update the state
 769             persist(new UpdateElectionTerm(this.currentTerm, this.votedFor), new Procedure<UpdateElectionTerm>(){
 770
 771                 @Override public void apply(UpdateElectionTerm param)
 772                     throws Exception {
 773
 774                 }
 775             });
 776         }
 777     }
 778
 779     private static class UpdateElectionTerm implements Serializable {
 780         private final long currentTerm;
 781         private final String votedFor;
 782
 783         public UpdateElectionTerm(long currentTerm, String votedFor) {
 784             this.currentTerm = currentTerm;
 785             this.votedFor = votedFor;
 786         }
 787
 788         public long getCurrentTerm() {
 789             return currentTerm;
 790         }
 791
 792         public String getVotedFor() {
 793             return votedFor;
 794         }
 795     }
 796
 797 }