opendaylight/md-sal/sal-akka-raft/src/main/java/org/opendaylight/controller/cluster/raft/behaviors/AbstractLeader.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.behaviors;
  10
  11 import akka.actor.ActorRef;
  12 import akka.actor.ActorSelection;
  13 import akka.actor.Cancellable;
  14 import com.google.common.annotations.VisibleForTesting;
  15 import com.google.common.base.Optional;
  16 import com.google.common.base.Preconditions;
  17 import com.google.protobuf.ByteString;
  18 import java.io.IOException;
  19 import java.util.Collection;
  20 import java.util.Collections;
  21 import java.util.HashMap;
  22 import java.util.Iterator;
  23 import java.util.LinkedList;
  24 import java.util.List;
  25 import java.util.Map;
  26 import java.util.Map.Entry;
  27 import javax.annotation.Nullable;
  28 import org.opendaylight.controller.cluster.raft.ClientRequestTracker;
  29 import org.opendaylight.controller.cluster.raft.ClientRequestTrackerImpl;
  30 import org.opendaylight.controller.cluster.raft.FollowerLogInformation;
  31 import org.opendaylight.controller.cluster.raft.FollowerLogInformationImpl;
  32 import org.opendaylight.controller.cluster.raft.PeerInfo;
  33 import org.opendaylight.controller.cluster.raft.RaftActorContext;
  34 import org.opendaylight.controller.cluster.raft.RaftState;
  35 import org.opendaylight.controller.cluster.raft.ReplicatedLogEntry;
  36 import org.opendaylight.controller.cluster.raft.Snapshot;
  37 import org.opendaylight.controller.cluster.raft.VotingState;
  38 import org.opendaylight.controller.cluster.raft.base.messages.Replicate;
  39 import org.opendaylight.controller.cluster.raft.base.messages.SendHeartBeat;
  40 import org.opendaylight.controller.cluster.raft.base.messages.SendInstallSnapshot;
  41 import org.opendaylight.controller.cluster.raft.messages.AppendEntries;
  42 import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
  43 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshot;
  44 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshotReply;
  45 import org.opendaylight.controller.cluster.raft.messages.RaftRPC;
  46 import org.opendaylight.controller.cluster.raft.messages.RequestVoteReply;
  47 import org.opendaylight.controller.cluster.raft.messages.UnInitializedFollowerSnapshotReply;
  48 import scala.concurrent.duration.FiniteDuration;
  49
  50 /**
  51  * The behavior of a RaftActor when it is in the Leader state
  52  * <p/>
  53  * Leaders:
  54  * <ul>
  55  * <li> Upon election: send initial empty AppendEntries RPCs
  56  * (heartbeat) to each server; repeat during idle periods to
  57  * prevent election timeouts (§5.2)
  58  * <li> If command received from client: append entry to local log,
  59  * respond after entry applied to state machine (§5.3)
  60  * <li> If last log index ≥ nextIndex for a follower: send
  61  * AppendEntries RPC with log entries starting at nextIndex
  62  * <ul>
  63  * <li> If successful: update nextIndex and matchIndex for
  64  * follower (§5.3)
  65  * <li> If AppendEntries fails because of log inconsistency:
  66  * decrement nextIndex and retry (§5.3)
  67  * </ul>
  68  * <li> If there exists an N such that N > commitIndex, a majority
  69  * of matchIndex[i] ≥ N, and log[N].term == currentTerm:
  70  * set commitIndex = N (§5.3, §5.4).
  71  */
  72 public abstract class AbstractLeader extends AbstractRaftActorBehavior {
  73
  74     // The index of the first chunk that is sent when installing a snapshot
  75     public static final int FIRST_CHUNK_INDEX = 1;
  76
  77     // The index that the follower should respond with if it needs the install snapshot to be reset
  78     public static final int INVALID_CHUNK_INDEX = -1;
  79
  80     // This would be passed as the hash code of the last chunk when sending the first chunk
  81     public static final int INITIAL_LAST_CHUNK_HASH_CODE = -1;
  82
  83     private final Map<String, FollowerLogInformation> followerToLog = new HashMap<>();
  84     private final Map<String, FollowerToSnapshot> mapFollowerToSnapshot = new HashMap<>();
  85
  86     private Cancellable heartbeatSchedule = null;
  87
  88     private final Collection<ClientRequestTracker> trackerList = new LinkedList<>();
  89
  90     private int minReplicationCount;
  91
  92     private Optional<SnapshotHolder> snapshot;
  93
  94     public AbstractLeader(RaftActorContext context) {
  95         super(context, RaftState.Leader);
  96
  97         setLeaderPayloadVersion(context.getPayloadVersion());
  98
  99         for(PeerInfo peerInfo: context.getPeers()) {
 100             FollowerLogInformation followerLogInformation = new FollowerLogInformationImpl(peerInfo, -1, context);
 101             followerToLog.put(peerInfo.getId(), followerLogInformation);
 102         }
 103
 104         leaderId = context.getId();
 105
 106         LOG.debug("{}: Election: Leader has following peers: {}", logName(), getFollowerIds());
 107
 108         updateMinReplicaCount();
 109
 110         snapshot = Optional.absent();
 111
 112         // Immediately schedule a heartbeat
 113         // Upon election: send initial empty AppendEntries RPCs
 114         // (heartbeat) to each server; repeat during idle periods to
 115         // prevent election timeouts (§5.2)
 116         sendAppendEntries(0, false);
 117
 118         // It is important to schedule this heartbeat here
 119         scheduleHeartBeat(context.getConfigParams().getHeartBeatInterval());
 120     }
 121
 122     /**
 123      * Return an immutable collection of follower identifiers.
 124      *
 125      * @return Collection of follower IDs
 126      */
 127     public final Collection<String> getFollowerIds() {
 128         return followerToLog.keySet();
 129     }
 130
 131     public void addFollower(String followerId) {
 132         FollowerLogInformation followerLogInformation = new FollowerLogInformationImpl(
 133                 context.getPeerInfo(followerId), -1, context);
 134         followerToLog.put(followerId, followerLogInformation);
 135
 136         if(heartbeatSchedule == null) {
 137             scheduleHeartBeat(context.getConfigParams().getHeartBeatInterval());
 138         }
 139     }
 140
 141     public void removeFollower(String followerId) {
 142         followerToLog.remove(followerId);
 143         mapFollowerToSnapshot.remove(followerId);
 144     }
 145
 146     public void updateMinReplicaCount() {
 147         int numVoting = 0;
 148         for(PeerInfo peer: context.getPeers()) {
 149             if(peer.isVoting()) {
 150                 numVoting++;
 151             }
 152         }
 153
 154         minReplicationCount = getMajorityVoteCount(numVoting);
 155     }
 156
 157     protected int getMinIsolatedLeaderPeerCount(){
 158       //the isolated Leader peer count will be 1 less than the majority vote count.
 159         //this is because the vote count has the self vote counted in it
 160         //for e.g
 161         //0 peers = 1 votesRequired , minIsolatedLeaderPeerCount = 0
 162         //2 peers = 2 votesRequired , minIsolatedLeaderPeerCount = 1
 163         //4 peers = 3 votesRequired, minIsolatedLeaderPeerCount = 2
 164
 165         return minReplicationCount > 0 ? (minReplicationCount - 1) : 0;
 166     }
 167
 168     @VisibleForTesting
 169     void setSnapshot(@Nullable Snapshot snapshot) {
 170         if(snapshot != null) {
 171             this.snapshot = Optional.of(new SnapshotHolder(snapshot));
 172         } else {
 173             this.snapshot = Optional.absent();
 174         }
 175     }
 176
 177     @Override
 178     protected RaftActorBehavior handleAppendEntries(ActorRef sender,
 179         AppendEntries appendEntries) {
 180
 181         LOG.debug("{}: handleAppendEntries: {}", logName(), appendEntries);
 182
 183         return this;
 184     }
 185
 186     @Override
 187     protected RaftActorBehavior handleAppendEntriesReply(ActorRef sender,
 188         AppendEntriesReply appendEntriesReply) {
 189
 190         if(LOG.isTraceEnabled()) {
 191             LOG.trace("{}: handleAppendEntriesReply: {}", logName(), appendEntriesReply);
 192         }
 193
 194         // Update the FollowerLogInformation
 195         String followerId = appendEntriesReply.getFollowerId();
 196         FollowerLogInformation followerLogInformation =
 197             followerToLog.get(followerId);
 198
 199         if(followerLogInformation == null){
 200             LOG.error("{}: handleAppendEntriesReply - unknown follower {}", logName(), followerId);
 201             return this;
 202         }
 203
 204         if(followerLogInformation.timeSinceLastActivity() >
 205                 context.getConfigParams().getElectionTimeOutInterval().toMillis()) {
 206             LOG.warn("{} : handleAppendEntriesReply delayed beyond election timeout, " +
 207                             "appendEntriesReply : {}, timeSinceLastActivity : {}, lastApplied : {}, commitIndex : {}",
 208                     logName(), appendEntriesReply, followerLogInformation.timeSinceLastActivity(),
 209                     context.getLastApplied(), context.getCommitIndex());
 210         }
 211
 212         followerLogInformation.markFollowerActive();
 213         followerLogInformation.setPayloadVersion(appendEntriesReply.getPayloadVersion());
 214
 215         boolean updated = false;
 216         if (appendEntriesReply.isSuccess()) {
 217             updated = updateFollowerLogInformation(followerLogInformation, appendEntriesReply);
 218         } else {
 219             LOG.debug("{}: handleAppendEntriesReply: received unsuccessful reply: {}", logName(), appendEntriesReply);
 220
 221             long followerLastLogIndex = appendEntriesReply.getLogLastIndex();
 222             ReplicatedLogEntry followersLastLogEntry = context.getReplicatedLog().get(followerLastLogIndex);
 223             if(appendEntriesReply.isForceInstallSnapshot()) {
 224                 // Reset the followers match and next index. This is to signal that this follower has nothing
 225                 // in common with this Leader and so would require a snapshot to be installed
 226                 followerLogInformation.setMatchIndex(-1);
 227                 followerLogInformation.setNextIndex(-1);
 228
 229                 // Force initiate a snapshot capture
 230                 initiateCaptureSnapshot(followerId);
 231             } else if(followerLastLogIndex < 0 || (followersLastLogEntry != null &&
 232                     followersLastLogEntry.getTerm() == appendEntriesReply.getLogLastTerm())) {
 233                 // The follower's log is empty or the last entry is present in the leader's journal
 234                 // and the terms match so the follower is just behind the leader's journal from
 235                 // the last snapshot, if any. We'll catch up the follower quickly by starting at the
 236                 // follower's last log index.
 237
 238                 updated = updateFollowerLogInformation(followerLogInformation, appendEntriesReply);
 239             } else {
 240                 // TODO: When we find that the follower is out of sync with the
 241                 // Leader we simply decrement that followers next index by 1.
 242                 // Would it be possible to do better than this? The RAFT spec
 243                 // does not explicitly deal with it but may be something for us to
 244                 // think about.
 245
 246                 followerLogInformation.decrNextIndex();
 247             }
 248         }
 249
 250         // Now figure out if this reply warrants a change in the commitIndex
 251         // If there exists an N such that N > commitIndex, a majority
 252         // of matchIndex[i] ≥ N, and log[N].term == currentTerm:
 253         // set commitIndex = N (§5.3, §5.4).
 254         for (long N = context.getCommitIndex() + 1; ; N++) {
 255             int replicatedCount = 1;
 256
 257             for (FollowerLogInformation info : followerToLog.values()) {
 258                 final PeerInfo peerInfo = context.getPeerInfo(info.getId());
 259                 if(info.getMatchIndex() >= N && (peerInfo != null && peerInfo.isVoting())) {
 260                     replicatedCount++;
 261                 }
 262             }
 263
 264             if (replicatedCount >= minReplicationCount) {
 265                 ReplicatedLogEntry replicatedLogEntry = context.getReplicatedLog().get(N);
 266                 if (replicatedLogEntry == null) {
 267                     break;
 268                 }
 269
 270                 // Don't update the commit index if the log entry is from a previous term, as per §5.4.1:
 271                 // "Raft never commits log entries from previous terms by counting replicas".
 272                 // However we keep looping so we can make progress when new entries in the current term
 273                 // reach consensus, as per §5.4.1: "once an entry from the current term is committed by
 274                 // counting replicas, then all prior entries are committed indirectly".
 275                 if (replicatedLogEntry.getTerm() == currentTerm()) {
 276                     context.setCommitIndex(N);
 277                 }
 278             } else {
 279                 break;
 280             }
 281         }
 282
 283         // Apply the change to the state machine
 284         if (context.getCommitIndex() > context.getLastApplied()) {
 285             if(LOG.isDebugEnabled()) {
 286                 LOG.debug("{}: handleAppendEntriesReply from {}: applying to log - commitIndex: {}, lastAppliedIndex: {}",
 287                         logName(), followerId, context.getCommitIndex(), context.getLastApplied());
 288             }
 289
 290             applyLogToStateMachine(context.getCommitIndex());
 291         }
 292
 293         if (!context.getSnapshotManager().isCapturing()) {
 294             purgeInMemoryLog();
 295         }
 296
 297         //Send the next log entry immediately, if possible, no need to wait for heartbeat to trigger that event
 298         sendUpdatesToFollower(followerId, followerLogInformation, false, !updated);
 299         return this;
 300     }
 301
 302     private boolean updateFollowerLogInformation(FollowerLogInformation followerLogInformation,
 303             AppendEntriesReply appendEntriesReply) {
 304         boolean updated = followerLogInformation.setMatchIndex(appendEntriesReply.getLogLastIndex());
 305         updated = followerLogInformation.setNextIndex(appendEntriesReply.getLogLastIndex() + 1) || updated;
 306
 307         if(updated && LOG.isDebugEnabled()) {
 308             LOG.debug("{}: handleAppendEntriesReply - FollowerLogInformation for {} updated: matchIndex: {}, nextIndex: {}",
 309                     logName(), followerLogInformation.getId(), followerLogInformation.getMatchIndex(),
 310                     followerLogInformation.getNextIndex());
 311         }
 312         return updated;
 313     }
 314
 315     private void purgeInMemoryLog() {
 316         //find the lowest index across followers which has been replicated to all.
 317         // lastApplied if there are no followers, so that we keep clearing the log for single-node
 318         // we would delete the in-mem log from that index on, in-order to minimize mem usage
 319         // we would also share this info thru AE with the followers so that they can delete their log entries as well.
 320         long minReplicatedToAllIndex = followerToLog.isEmpty() ? context.getLastApplied() : Long.MAX_VALUE;
 321         for (FollowerLogInformation info : followerToLog.values()) {
 322             minReplicatedToAllIndex = Math.min(minReplicatedToAllIndex, info.getMatchIndex());
 323         }
 324
 325         super.performSnapshotWithoutCapture(minReplicatedToAllIndex);
 326     }
 327
 328     @Override
 329     protected ClientRequestTracker removeClientRequestTracker(long logIndex) {
 330         final Iterator<ClientRequestTracker> it = trackerList.iterator();
 331         while (it.hasNext()) {
 332             final ClientRequestTracker t = it.next();
 333             if (t.getIndex() == logIndex) {
 334                 it.remove();
 335                 return t;
 336             }
 337         }
 338
 339         return null;
 340     }
 341
 342     @Override
 343     protected ClientRequestTracker findClientRequestTracker(long logIndex) {
 344         for (ClientRequestTracker tracker : trackerList) {
 345             if (tracker.getIndex() == logIndex) {
 346                 return tracker;
 347             }
 348         }
 349         return null;
 350     }
 351
 352     @Override
 353     protected RaftActorBehavior handleRequestVoteReply(ActorRef sender,
 354         RequestVoteReply requestVoteReply) {
 355         return this;
 356     }
 357
 358     protected void beforeSendHeartbeat(){}
 359
 360     @Override
 361     public RaftActorBehavior handleMessage(ActorRef sender, Object originalMessage) {
 362         Preconditions.checkNotNull(sender, "sender should not be null");
 363
 364         Object message = fromSerializableMessage(originalMessage);
 365
 366         if (message instanceof RaftRPC) {
 367             RaftRPC rpc = (RaftRPC) message;
 368             // If RPC request or response contains term T > currentTerm:
 369             // set currentTerm = T, convert to follower (§5.1)
 370             // This applies to all RPC messages and responses
 371             if (rpc.getTerm() > context.getTermInformation().getCurrentTerm()) {
 372                 LOG.debug("{}: Term {} in \"{}\" message is greater than leader's term {} - switching to Follower",
 373                         logName(), rpc.getTerm(), rpc, context.getTermInformation().getCurrentTerm());
 374
 375                 context.getTermInformation().updateAndPersist(rpc.getTerm(), null);
 376
 377                 return internalSwitchBehavior(RaftState.Follower);
 378             }
 379         }
 380
 381         if (message instanceof SendHeartBeat) {
 382             beforeSendHeartbeat();
 383             sendHeartBeat();
 384             scheduleHeartBeat(context.getConfigParams().getHeartBeatInterval());
 385             return this;
 386
 387         } else if(message instanceof SendInstallSnapshot) {
 388             // received from RaftActor
 389             setSnapshot(((SendInstallSnapshot) message).getSnapshot());
 390             sendInstallSnapshot();
 391
 392         } else if (message instanceof Replicate) {
 393             replicate((Replicate) message);
 394
 395         } else if (message instanceof InstallSnapshotReply){
 396             handleInstallSnapshotReply((InstallSnapshotReply) message);
 397
 398         }
 399
 400
 401         return super.handleMessage(sender, message);
 402     }
 403
 404     private void handleInstallSnapshotReply(InstallSnapshotReply reply) {
 405         LOG.debug("{}: handleInstallSnapshotReply: {}", logName(), reply);
 406
 407         String followerId = reply.getFollowerId();
 408         FollowerToSnapshot followerToSnapshot = mapFollowerToSnapshot.get(followerId);
 409
 410         if (followerToSnapshot == null) {
 411             LOG.error("{}: FollowerToSnapshot not found for follower {} in InstallSnapshotReply",
 412                     logName(), followerId);
 413             return;
 414         }
 415
 416         FollowerLogInformation followerLogInformation = followerToLog.get(followerId);
 417         if(followerLogInformation == null) {
 418             // This can happen during AddServer if it times out.
 419             LOG.error("{}: FollowerLogInformation not found for follower {} in InstallSnapshotReply",
 420                     logName(), followerId);
 421             mapFollowerToSnapshot.remove(followerId);
 422             return;
 423         }
 424
 425         followerLogInformation.markFollowerActive();
 426
 427         if (followerToSnapshot.getChunkIndex() == reply.getChunkIndex()) {
 428             boolean wasLastChunk = false;
 429             if (reply.isSuccess()) {
 430                 if(followerToSnapshot.isLastChunk(reply.getChunkIndex())) {
 431                     //this was the last chunk reply
 432                     if(LOG.isDebugEnabled()) {
 433                         LOG.debug("{}: InstallSnapshotReply received, " +
 434                                 "last chunk received, Chunk: {}. Follower: {} Setting nextIndex: {}",
 435                                 logName(), reply.getChunkIndex(), followerId,
 436                             context.getReplicatedLog().getSnapshotIndex() + 1
 437                         );
 438                     }
 439
 440                     long followerMatchIndex = snapshot.get().getLastIncludedIndex();
 441                     followerLogInformation.setMatchIndex(followerMatchIndex);
 442                     followerLogInformation.setNextIndex(followerMatchIndex + 1);
 443                     mapFollowerToSnapshot.remove(followerId);
 444
 445                     LOG.debug("{}: follower: {}, matchIndex set to {}, nextIndex set to {}",
 446                         logName(), followerId, followerLogInformation.getMatchIndex(),
 447                         followerLogInformation.getNextIndex());
 448
 449                     if (mapFollowerToSnapshot.isEmpty()) {
 450                         // once there are no pending followers receiving snapshots
 451                         // we can remove snapshot from the memory
 452                         setSnapshot(null);
 453                     }
 454                     wasLastChunk = true;
 455                     if(context.getPeerInfo(followerId).getVotingState() == VotingState.VOTING_NOT_INITIALIZED){
 456                         UnInitializedFollowerSnapshotReply unInitFollowerSnapshotSuccess =
 457                                              new UnInitializedFollowerSnapshotReply(followerId);
 458                         context.getActor().tell(unInitFollowerSnapshotSuccess, context.getActor());
 459                         LOG.debug("Sent message UnInitializedFollowerSnapshotReply to self");
 460                     }
 461                 } else {
 462                     followerToSnapshot.markSendStatus(true);
 463                 }
 464             } else {
 465                 LOG.info("{}: InstallSnapshotReply received sending snapshot chunk failed, Will retry, Chunk: {}",
 466                         logName(), reply.getChunkIndex());
 467
 468                 followerToSnapshot.markSendStatus(false);
 469             }
 470
 471             if (wasLastChunk && !context.getSnapshotManager().isCapturing()) {
 472                 // Since the follower is now caught up try to purge the log.
 473                 purgeInMemoryLog();
 474             } else if (!wasLastChunk && followerToSnapshot.canSendNextChunk()) {
 475                 ActorSelection followerActor = context.getPeerActorSelection(followerId);
 476                 if(followerActor != null) {
 477                     sendSnapshotChunk(followerActor, followerId);
 478                 }
 479             }
 480
 481         } else {
 482             LOG.error("{}: Chunk index {} in InstallSnapshotReply from follower {} does not match expected index {}",
 483                     logName(), reply.getChunkIndex(), followerId,
 484                     followerToSnapshot.getChunkIndex());
 485
 486             if(reply.getChunkIndex() == INVALID_CHUNK_INDEX){
 487                 // Since the Follower did not find this index to be valid we should reset the follower snapshot
 488                 // so that Installing the snapshot can resume from the beginning
 489                 followerToSnapshot.reset();
 490             }
 491         }
 492     }
 493
 494     private void replicate(Replicate replicate) {
 495         long logIndex = replicate.getReplicatedLogEntry().getIndex();
 496
 497         LOG.debug("{}: Replicate message: identifier: {}, logIndex: {}", logName(),
 498                 replicate.getIdentifier(), logIndex);
 499
 500         // Create a tracker entry we will use this later to notify the
 501         // client actor
 502         trackerList.add(
 503             new ClientRequestTrackerImpl(replicate.getClientActor(),
 504                 replicate.getIdentifier(),
 505                 logIndex)
 506         );
 507
 508         boolean applyModificationToState = followerToLog.isEmpty()
 509                 || context.getRaftPolicy().applyModificationToStateBeforeConsensus();
 510
 511         if(applyModificationToState){
 512             context.setCommitIndex(logIndex);
 513             applyLogToStateMachine(logIndex);
 514         }
 515
 516         if (!followerToLog.isEmpty()) {
 517             sendAppendEntries(0, false);
 518         }
 519     }
 520
 521     protected void sendAppendEntries(long timeSinceLastActivityInterval, boolean isHeartbeat) {
 522         // Send an AppendEntries to all followers
 523         for (Entry<String, FollowerLogInformation> e : followerToLog.entrySet()) {
 524             final String followerId = e.getKey();
 525             final FollowerLogInformation followerLogInformation = e.getValue();
 526             // This checks helps not to send a repeat message to the follower
 527             if(!followerLogInformation.isFollowerActive() ||
 528                     followerLogInformation.timeSinceLastActivity() >= timeSinceLastActivityInterval) {
 529                 sendUpdatesToFollower(followerId, followerLogInformation, true, isHeartbeat);
 530             }
 531         }
 532     }
 533
 534     /**
 535      *
 536      * This method checks if any update needs to be sent to the given follower. This includes append log entries,
 537      * sending next snapshot chunk, and initiating a snapshot.
 538      * @return true if any update is sent, false otherwise
 539      */
 540
 541     private void sendUpdatesToFollower(String followerId, FollowerLogInformation followerLogInformation,
 542                                        boolean sendHeartbeat, boolean isHeartbeat) {
 543
 544         ActorSelection followerActor = context.getPeerActorSelection(followerId);
 545         if (followerActor != null) {
 546             long followerNextIndex = followerLogInformation.getNextIndex();
 547             boolean isFollowerActive = followerLogInformation.isFollowerActive();
 548             boolean sendAppendEntries = false;
 549             List<ReplicatedLogEntry> entries = Collections.emptyList();
 550
 551             if (mapFollowerToSnapshot.get(followerId) != null) {
 552                 // if install snapshot is in process , then sent next chunk if possible
 553                 if (isFollowerActive && mapFollowerToSnapshot.get(followerId).canSendNextChunk()) {
 554                     sendSnapshotChunk(followerActor, followerId);
 555                 } else if(sendHeartbeat) {
 556                     // we send a heartbeat even if we have not received a reply for the last chunk
 557                     sendAppendEntries = true;
 558                 }
 559             } else {
 560                 long leaderLastIndex = context.getReplicatedLog().lastIndex();
 561                 long leaderSnapShotIndex = context.getReplicatedLog().getSnapshotIndex();
 562
 563                 if((!isHeartbeat && LOG.isDebugEnabled()) || LOG.isTraceEnabled()) {
 564                     LOG.debug("{}: Checking sendAppendEntries for follower {}: active: {}, followerNextIndex: {}, leaderLastIndex: {}, leaderSnapShotIndex: {}",
 565                             logName(), followerId, isFollowerActive, followerNextIndex, leaderLastIndex, leaderSnapShotIndex);
 566                 }
 567
 568                 if (isFollowerActive && context.getReplicatedLog().isPresent(followerNextIndex)) {
 569
 570                     LOG.debug("{}: sendAppendEntries: {} is present for follower {}", logName(),
 571                             followerNextIndex, followerId);
 572
 573                     if(followerLogInformation.okToReplicate()) {
 574                         // Try to send all the entries in the journal but not exceeding the max data size
 575                         // for a single AppendEntries message.
 576                         int maxEntries = (int) context.getReplicatedLog().size();
 577                         entries = context.getReplicatedLog().getFrom(followerNextIndex, maxEntries,
 578                                 context.getConfigParams().getSnapshotChunkSize());
 579                         sendAppendEntries = true;
 580                     }
 581                 } else if (isFollowerActive && followerNextIndex >= 0 &&
 582                     leaderLastIndex > followerNextIndex && !context.getSnapshotManager().isCapturing()) {
 583                     // if the followers next index is not present in the leaders log, and
 584                     // if the follower is just not starting and if leader's index is more than followers index
 585                     // then snapshot should be sent
 586
 587                     if (LOG.isDebugEnabled()) {
 588                         LOG.debug(String.format("%s: InitiateInstallSnapshot to follower: %s," +
 589                                     "follower-nextIndex: %d, leader-snapshot-index: %d,  " +
 590                                     "leader-last-index: %d", logName(), followerId,
 591                                     followerNextIndex, leaderSnapShotIndex, leaderLastIndex));
 592                     }
 593
 594                     // Send heartbeat to follower whenever install snapshot is initiated.
 595                     sendAppendEntries = true;
 596                     if (canInstallSnapshot(followerNextIndex)) {
 597                         initiateCaptureSnapshot(followerId);
 598                     }
 599
 600                 } else if(sendHeartbeat) {
 601                     // we send an AppendEntries, even if the follower is inactive
 602                     // in-order to update the followers timestamp, in case it becomes active again
 603                     sendAppendEntries = true;
 604                 }
 605
 606             }
 607
 608             if(sendAppendEntries) {
 609                 sendAppendEntriesToFollower(followerActor, followerNextIndex,
 610                         entries, followerId);
 611             }
 612         }
 613     }
 614
 615     private void sendAppendEntriesToFollower(ActorSelection followerActor, long followerNextIndex,
 616         List<ReplicatedLogEntry> entries, String followerId) {
 617         AppendEntries appendEntries = new AppendEntries(currentTerm(), context.getId(),
 618             prevLogIndex(followerNextIndex),
 619             prevLogTerm(followerNextIndex), entries,
 620             context.getCommitIndex(), super.getReplicatedToAllIndex(), context.getPayloadVersion());
 621
 622         if(!entries.isEmpty() || LOG.isTraceEnabled()) {
 623             LOG.debug("{}: Sending AppendEntries to follower {}: {}", logName(), followerId,
 624                     appendEntries);
 625         }
 626
 627         followerActor.tell(appendEntries.toSerializable(), actor());
 628     }
 629
 630     /**
 631      * Install Snapshot works as follows
 632      * 1. Leader initiates the capture snapshot by sending a CaptureSnapshot message to actor
 633      * 2. RaftActor on receipt of the CaptureSnapshotReply (from Shard), stores the received snapshot in the replicated log
 634      * and makes a call to Leader's handleMessage , with SendInstallSnapshot message.
 635      * 3. Leader , picks the snapshot from im-mem ReplicatedLog and sends it in chunks to the Follower
 636      * 4. On complete, Follower sends back a InstallSnapshotReply.
 637      * 5. On receipt of the InstallSnapshotReply for the last chunk, Leader marks the install complete for that follower
 638      * and replenishes the memory by deleting the snapshot in Replicated log.
 639      * 6. If another follower requires a snapshot and a snapshot has been collected (via CaptureSnapshotReply)
 640      * then send the existing snapshot in chunks to the follower.
 641      * @param followerId
 642      */
 643     public boolean initiateCaptureSnapshot(String followerId) {
 644         if (snapshot.isPresent()) {
 645             // if a snapshot is present in the memory, most likely another install is in progress
 646             // no need to capture snapshot.
 647             // This could happen if another follower needs an install when one is going on.
 648             final ActorSelection followerActor = context.getPeerActorSelection(followerId);
 649             sendSnapshotChunk(followerActor, followerId);
 650             return true;
 651         } else {
 652             return context.getSnapshotManager().captureToInstall(context.getReplicatedLog().last(),
 653                     this.getReplicatedToAllIndex(), followerId);
 654         }
 655     }
 656
 657     private boolean canInstallSnapshot(long nextIndex){
 658         // If the follower's nextIndex is -1 then we might as well send it a snapshot
 659         // Otherwise send it a snapshot only if the nextIndex is not present in the log but is present
 660         // in the snapshot
 661         return (nextIndex == -1 ||
 662                 (!context.getReplicatedLog().isPresent(nextIndex)
 663                         && context.getReplicatedLog().isInSnapshot(nextIndex)));
 664
 665     }
 666
 667
 668     private void sendInstallSnapshot() {
 669         LOG.debug("{}: sendInstallSnapshot", logName());
 670         for (Entry<String, FollowerLogInformation> e : followerToLog.entrySet()) {
 671             String followerId = e.getKey();
 672             ActorSelection followerActor = context.getPeerActorSelection(followerId);
 673             FollowerLogInformation followerLogInfo = e.getValue();
 674
 675             if (followerActor != null) {
 676                 long nextIndex = followerLogInfo.getNextIndex();
 677                 if (context.getPeerInfo(followerId).getVotingState() == VotingState.VOTING_NOT_INITIALIZED ||
 678                         canInstallSnapshot(nextIndex)) {
 679                     sendSnapshotChunk(followerActor, followerId);
 680                 }
 681             }
 682         }
 683     }
 684
 685     /**
 686      *  Sends a snapshot chunk to a given follower
 687      *  InstallSnapshot should qualify as a heartbeat too.
 688      */
 689     private void sendSnapshotChunk(ActorSelection followerActor, String followerId) {
 690         try {
 691             if (snapshot.isPresent()) {
 692                 ByteString nextSnapshotChunk = getNextSnapshotChunk(followerId, snapshot.get().getSnapshotBytes());
 693
 694                 // Note: the previous call to getNextSnapshotChunk has the side-effect of adding
 695                 // followerId to the followerToSnapshot map.
 696                 FollowerToSnapshot followerToSnapshot = mapFollowerToSnapshot.get(followerId);
 697
 698                 followerActor.tell(
 699                     new InstallSnapshot(currentTerm(), context.getId(),
 700                         snapshot.get().getLastIncludedIndex(),
 701                         snapshot.get().getLastIncludedTerm(),
 702                         nextSnapshotChunk,
 703                         followerToSnapshot.incrementChunkIndex(),
 704                         followerToSnapshot.getTotalChunks(),
 705                         Optional.of(followerToSnapshot.getLastChunkHashCode())
 706                     ).toSerializable(),
 707                     actor()
 708                 );
 709
 710                 if(LOG.isDebugEnabled()) {
 711                     LOG.debug("{}: InstallSnapshot sent to follower {}, Chunk: {}/{}",
 712                             logName(), followerActor.path(), followerToSnapshot.getChunkIndex(),
 713                             followerToSnapshot.getTotalChunks());
 714                 }
 715             }
 716         } catch (IOException e) {
 717             LOG.error("{}: InstallSnapshot failed for Leader.", logName(), e);
 718         }
 719     }
 720
 721     /**
 722      * Acccepts snaphot as ByteString, enters into map for future chunks
 723      * creates and return a ByteString chunk
 724      */
 725     private ByteString getNextSnapshotChunk(String followerId, ByteString snapshotBytes) throws IOException {
 726         FollowerToSnapshot followerToSnapshot = mapFollowerToSnapshot.get(followerId);
 727         if (followerToSnapshot == null) {
 728             followerToSnapshot = new FollowerToSnapshot(snapshotBytes);
 729             mapFollowerToSnapshot.put(followerId, followerToSnapshot);
 730         }
 731         ByteString nextChunk = followerToSnapshot.getNextChunk();
 732
 733         LOG.debug("{}: next snapshot chunk size for follower {}: {}", logName(), followerId, nextChunk.size());
 734
 735         return nextChunk;
 736     }
 737
 738     private void sendHeartBeat() {
 739         if (!followerToLog.isEmpty()) {
 740             LOG.trace("{}: Sending heartbeat", logName());
 741             sendAppendEntries(context.getConfigParams().getHeartBeatInterval().toMillis(), true);
 742         }
 743     }
 744
 745     private void stopHeartBeat() {
 746         if (heartbeatSchedule != null && !heartbeatSchedule.isCancelled()) {
 747             heartbeatSchedule.cancel();
 748         }
 749     }
 750
 751     private void scheduleHeartBeat(FiniteDuration interval) {
 752         if (followerToLog.isEmpty()) {
 753             // Optimization - do not bother scheduling a heartbeat as there are
 754             // no followers
 755             return;
 756         }
 757
 758         stopHeartBeat();
 759
 760         // Schedule a heartbeat. When the scheduler triggers a SendHeartbeat
 761         // message is sent to itself.
 762         // Scheduling the heartbeat only once here because heartbeats do not
 763         // need to be sent if there are other messages being sent to the remote
 764         // actor.
 765         heartbeatSchedule = context.getActorSystem().scheduler().scheduleOnce(
 766             interval, context.getActor(), new SendHeartBeat(),
 767             context.getActorSystem().dispatcher(), context.getActor());
 768     }
 769
 770     @Override
 771     public void close() throws Exception {
 772         stopHeartBeat();
 773     }
 774
 775     @Override
 776     public String getLeaderId() {
 777         return context.getId();
 778     }
 779
 780     protected boolean isLeaderIsolated() {
 781         int minPresent = getMinIsolatedLeaderPeerCount();
 782         for (FollowerLogInformation followerLogInformation : followerToLog.values()) {
 783             if (followerLogInformation.isFollowerActive()) {
 784                 --minPresent;
 785                 if (minPresent == 0) {
 786                     break;
 787                 }
 788             }
 789         }
 790         return (minPresent != 0);
 791     }
 792
 793     /**
 794      * Encapsulates the snapshot bytestring and handles the logic of sending
 795      * snapshot chunks
 796      */
 797     protected class FollowerToSnapshot {
 798         private final ByteString snapshotBytes;
 799         private int offset = 0;
 800         // the next snapshot chunk is sent only if the replyReceivedForOffset matches offset
 801         private int replyReceivedForOffset;
 802         // if replyStatus is false, the previous chunk is attempted
 803         private boolean replyStatus = false;
 804         private int chunkIndex;
 805         private final int totalChunks;
 806         private int lastChunkHashCode = AbstractLeader.INITIAL_LAST_CHUNK_HASH_CODE;
 807         private int nextChunkHashCode = AbstractLeader.INITIAL_LAST_CHUNK_HASH_CODE;
 808
 809         public FollowerToSnapshot(ByteString snapshotBytes) {
 810             this.snapshotBytes = snapshotBytes;
 811             int size = snapshotBytes.size();
 812             totalChunks = ( size / context.getConfigParams().getSnapshotChunkSize()) +
 813                 ((size % context.getConfigParams().getSnapshotChunkSize()) > 0 ? 1 : 0);
 814             if(LOG.isDebugEnabled()) {
 815                 LOG.debug("{}: Snapshot {} bytes, total chunks to send:{}",
 816                         logName(), size, totalChunks);
 817             }
 818             replyReceivedForOffset = -1;
 819             chunkIndex = AbstractLeader.FIRST_CHUNK_INDEX;
 820         }
 821
 822         public ByteString getSnapshotBytes() {
 823             return snapshotBytes;
 824         }
 825
 826         public int incrementOffset() {
 827             if(replyStatus) {
 828                 // if prev chunk failed, we would want to sent the same chunk again
 829                 offset = offset + context.getConfigParams().getSnapshotChunkSize();
 830             }
 831             return offset;
 832         }
 833
 834         public int incrementChunkIndex() {
 835             if (replyStatus) {
 836                 // if prev chunk failed, we would want to sent the same chunk again
 837                 chunkIndex =  chunkIndex + 1;
 838             }
 839             return chunkIndex;
 840         }
 841
 842         public int getChunkIndex() {
 843             return chunkIndex;
 844         }
 845
 846         public int getTotalChunks() {
 847             return totalChunks;
 848         }
 849
 850         public boolean canSendNextChunk() {
 851             // we only send a false if a chunk is sent but we have not received a reply yet
 852             return replyReceivedForOffset == offset;
 853         }
 854
 855         public boolean isLastChunk(int chunkIndex) {
 856             return totalChunks == chunkIndex;
 857         }
 858
 859         public void markSendStatus(boolean success) {
 860             if (success) {
 861                 // if the chunk sent was successful
 862                 replyReceivedForOffset = offset;
 863                 replyStatus = true;
 864                 lastChunkHashCode = nextChunkHashCode;
 865             } else {
 866                 // if the chunk sent was failure
 867                 replyReceivedForOffset = offset;
 868                 replyStatus = false;
 869             }
 870         }
 871
 872         public ByteString getNextChunk() {
 873             int snapshotLength = getSnapshotBytes().size();
 874             int start = incrementOffset();
 875             int size = context.getConfigParams().getSnapshotChunkSize();
 876             if (context.getConfigParams().getSnapshotChunkSize() > snapshotLength) {
 877                 size = snapshotLength;
 878             } else {
 879                 if ((start + context.getConfigParams().getSnapshotChunkSize()) > snapshotLength) {
 880                     size = snapshotLength - start;
 881                 }
 882             }
 883
 884
 885             LOG.debug("{}: Next chunk: length={}, offset={},size={}", logName(),
 886                     snapshotLength, start, size);
 887
 888             ByteString substring = getSnapshotBytes().substring(start, start + size);
 889             nextChunkHashCode = substring.hashCode();
 890             return substring;
 891         }
 892
 893         /**
 894          * reset should be called when the Follower needs to be sent the snapshot from the beginning
 895          */
 896         public void reset(){
 897             offset = 0;
 898             replyStatus = false;
 899             replyReceivedForOffset = offset;
 900             chunkIndex = AbstractLeader.FIRST_CHUNK_INDEX;
 901             lastChunkHashCode = AbstractLeader.INITIAL_LAST_CHUNK_HASH_CODE;
 902         }
 903
 904         public int getLastChunkHashCode() {
 905             return lastChunkHashCode;
 906         }
 907     }
 908
 909     // called from example-actor for printing the follower-states
 910     public String printFollowerStates() {
 911         final StringBuilder sb = new StringBuilder();
 912
 913         sb.append('[');
 914         for (FollowerLogInformation followerLogInformation : followerToLog.values()) {
 915             sb.append('{');
 916             sb.append(followerLogInformation.getId());
 917             sb.append(" state:");
 918             sb.append(followerLogInformation.isFollowerActive());
 919             sb.append("},");
 920         }
 921         sb.append(']');
 922
 923         return sb.toString();
 924     }
 925
 926     @VisibleForTesting
 927     public FollowerLogInformation getFollower(String followerId) {
 928         return followerToLog.get(followerId);
 929     }
 930
 931     @VisibleForTesting
 932     protected void setFollowerSnapshot(String followerId, FollowerToSnapshot snapshot) {
 933         mapFollowerToSnapshot.put(followerId, snapshot);
 934     }
 935
 936     @VisibleForTesting
 937     public int followerSnapshotSize() {
 938         return mapFollowerToSnapshot.size();
 939     }
 940
 941     @VisibleForTesting
 942     public int followerLogSize() {
 943         return followerToLog.size();
 944     }
 945
 946     private static class SnapshotHolder {
 947         private final long lastIncludedTerm;
 948         private final long lastIncludedIndex;
 949         private final ByteString snapshotBytes;
 950
 951         SnapshotHolder(Snapshot snapshot) {
 952             this.lastIncludedTerm = snapshot.getLastAppliedTerm();
 953             this.lastIncludedIndex = snapshot.getLastAppliedIndex();
 954             this.snapshotBytes = ByteString.copyFrom(snapshot.getState());
 955         }
 956
 957         long getLastIncludedTerm() {
 958             return lastIncludedTerm;
 959         }
 960
 961         long getLastIncludedIndex() {
 962             return lastIncludedIndex;
 963         }
 964
 965         ByteString getSnapshotBytes() {
 966             return snapshotBytes;
 967         }
 968     }
 969 }