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