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         if(LOG.isTraceEnabled()) {
 255             LOG.trace("{}: handleAppendEntriesReply from {}: commitIndex: {}, lastAppliedIndex: {}, currentTerm: {}",
 256                     logName(), followerId, context.getCommitIndex(), context.getLastApplied(), currentTerm());
 257         }
 258
 259         for (long N = context.getCommitIndex() + 1; ; N++) {
 260             int replicatedCount = 1;
 261
 262             LOG.trace("{}: checking Nth index {}", logName(), N);
 263             for (FollowerLogInformation info : followerToLog.values()) {
 264                 final PeerInfo peerInfo = context.getPeerInfo(info.getId());
 265                 if(info.getMatchIndex() >= N && (peerInfo != null && peerInfo.isVoting())) {
 266                     replicatedCount++;
 267                 } else if(LOG.isDebugEnabled()) {
 268                     LOG.debug("{}: Not counting follower {} - matchIndex: {}, {}", logName(), info.getId(),
 269                             info.getMatchIndex(), peerInfo);
 270                 }
 271             }
 272
 273             if(LOG.isTraceEnabled()) {
 274                 LOG.trace("{}: replicatedCount {}, minReplicationCount: {}", logName(), replicatedCount, minReplicationCount);
 275             }
 276
 277             if (replicatedCount >= minReplicationCount) {
 278                 ReplicatedLogEntry replicatedLogEntry = context.getReplicatedLog().get(N);
 279                 if (replicatedLogEntry == null) {
 280                     LOG.debug("{}: ReplicatedLogEntry not found for index {} - snapshotIndex: {}, journal size: {}",
 281                             logName(), N, context.getReplicatedLog().getSnapshotIndex(), context.getReplicatedLog().size());
 282                     break;
 283                 }
 284
 285                 // Don't update the commit index if the log entry is from a previous term, as per §5.4.1:
 286                 // "Raft never commits log entries from previous terms by counting replicas".
 287                 // However we keep looping so we can make progress when new entries in the current term
 288                 // reach consensus, as per §5.4.1: "once an entry from the current term is committed by
 289                 // counting replicas, then all prior entries are committed indirectly".
 290                 if (replicatedLogEntry.getTerm() == currentTerm()) {
 291                     LOG.trace("{}: Setting commit index to {}", logName(), N);
 292                     context.setCommitIndex(N);
 293                 } else {
 294                     LOG.debug("{}: Not updating commit index to {} - retrieved log entry with index {}, term {} does not match the current term {}",
 295                             logName(), N, replicatedLogEntry.getIndex(), replicatedLogEntry.getTerm(), currentTerm());
 296                 }
 297             } else {
 298                 LOG.trace("{}: minReplicationCount not reached - breaking", logName());
 299                 break;
 300             }
 301         }
 302
 303         // Apply the change to the state machine
 304         if (context.getCommitIndex() > context.getLastApplied()) {
 305             if(LOG.isDebugEnabled()) {
 306                 LOG.debug("{}: handleAppendEntriesReply from {}: applying to log - commitIndex: {}, lastAppliedIndex: {}",
 307                         logName(), followerId, context.getCommitIndex(), context.getLastApplied());
 308             }
 309
 310             applyLogToStateMachine(context.getCommitIndex());
 311         }
 312
 313         if (!context.getSnapshotManager().isCapturing()) {
 314             purgeInMemoryLog();
 315         }
 316
 317         //Send the next log entry immediately, if possible, no need to wait for heartbeat to trigger that event
 318         sendUpdatesToFollower(followerId, followerLogInformation, false, !updated);
 319         return this;
 320     }
 321
 322     private boolean updateFollowerLogInformation(FollowerLogInformation followerLogInformation,
 323             AppendEntriesReply appendEntriesReply) {
 324         boolean updated = followerLogInformation.setMatchIndex(appendEntriesReply.getLogLastIndex());
 325         updated = followerLogInformation.setNextIndex(appendEntriesReply.getLogLastIndex() + 1) || updated;
 326
 327         if(updated && LOG.isDebugEnabled()) {
 328             LOG.debug("{}: handleAppendEntriesReply - FollowerLogInformation for {} updated: matchIndex: {}, nextIndex: {}",
 329                     logName(), followerLogInformation.getId(), followerLogInformation.getMatchIndex(),
 330                     followerLogInformation.getNextIndex());
 331         }
 332         return updated;
 333     }
 334
 335     private void purgeInMemoryLog() {
 336         //find the lowest index across followers which has been replicated to all.
 337         // lastApplied if there are no followers, so that we keep clearing the log for single-node
 338         // we would delete the in-mem log from that index on, in-order to minimize mem usage
 339         // we would also share this info thru AE with the followers so that they can delete their log entries as well.
 340         long minReplicatedToAllIndex = followerToLog.isEmpty() ? context.getLastApplied() : Long.MAX_VALUE;
 341         for (FollowerLogInformation info : followerToLog.values()) {
 342             minReplicatedToAllIndex = Math.min(minReplicatedToAllIndex, info.getMatchIndex());
 343         }
 344
 345         super.performSnapshotWithoutCapture(minReplicatedToAllIndex);
 346     }
 347
 348     @Override
 349     protected ClientRequestTracker removeClientRequestTracker(long logIndex) {
 350         final Iterator<ClientRequestTracker> it = trackerList.iterator();
 351         while (it.hasNext()) {
 352             final ClientRequestTracker t = it.next();
 353             if (t.getIndex() == logIndex) {
 354                 it.remove();
 355                 return t;
 356             }
 357         }
 358
 359         return null;
 360     }
 361
 362     @Override
 363     protected ClientRequestTracker findClientRequestTracker(long logIndex) {
 364         for (ClientRequestTracker tracker : trackerList) {
 365             if (tracker.getIndex() == logIndex) {
 366                 return tracker;
 367             }
 368         }
 369         return null;
 370     }
 371
 372     @Override
 373     protected RaftActorBehavior handleRequestVoteReply(ActorRef sender,
 374         RequestVoteReply requestVoteReply) {
 375         return this;
 376     }
 377
 378     protected void beforeSendHeartbeat(){}
 379
 380     @Override
 381     public RaftActorBehavior handleMessage(ActorRef sender, Object originalMessage) {
 382         Preconditions.checkNotNull(sender, "sender should not be null");
 383
 384         Object message = fromSerializableMessage(originalMessage);
 385
 386         if (message instanceof RaftRPC) {
 387             RaftRPC rpc = (RaftRPC) message;
 388             // If RPC request or response contains term T > currentTerm:
 389             // set currentTerm = T, convert to follower (§5.1)
 390             // This applies to all RPC messages and responses
 391             if (rpc.getTerm() > context.getTermInformation().getCurrentTerm()) {
 392                 LOG.debug("{}: Term {} in \"{}\" message is greater than leader's term {} - switching to Follower",
 393                         logName(), rpc.getTerm(), rpc, context.getTermInformation().getCurrentTerm());
 394
 395                 context.getTermInformation().updateAndPersist(rpc.getTerm(), null);
 396
 397                 return internalSwitchBehavior(RaftState.Follower);
 398             }
 399         }
 400
 401         if (message instanceof SendHeartBeat) {
 402             beforeSendHeartbeat();
 403             sendHeartBeat();
 404             scheduleHeartBeat(context.getConfigParams().getHeartBeatInterval());
 405             return this;
 406
 407         } else if(message instanceof SendInstallSnapshot) {
 408             // received from RaftActor
 409             setSnapshot(((SendInstallSnapshot) message).getSnapshot());
 410             sendInstallSnapshot();
 411
 412         } else if (message instanceof Replicate) {
 413             replicate((Replicate) message);
 414
 415         } else if (message instanceof InstallSnapshotReply){
 416             handleInstallSnapshotReply((InstallSnapshotReply) message);
 417
 418         }
 419
 420
 421         return super.handleMessage(sender, message);
 422     }
 423
 424     private void handleInstallSnapshotReply(InstallSnapshotReply reply) {
 425         LOG.debug("{}: handleInstallSnapshotReply: {}", logName(), reply);
 426
 427         String followerId = reply.getFollowerId();
 428         FollowerToSnapshot followerToSnapshot = mapFollowerToSnapshot.get(followerId);
 429
 430         if (followerToSnapshot == null) {
 431             LOG.error("{}: FollowerToSnapshot not found for follower {} in InstallSnapshotReply",
 432                     logName(), followerId);
 433             return;
 434         }
 435
 436         FollowerLogInformation followerLogInformation = followerToLog.get(followerId);
 437         if(followerLogInformation == null) {
 438             // This can happen during AddServer if it times out.
 439             LOG.error("{}: FollowerLogInformation not found for follower {} in InstallSnapshotReply",
 440                     logName(), followerId);
 441             mapFollowerToSnapshot.remove(followerId);
 442             return;
 443         }
 444
 445         followerLogInformation.markFollowerActive();
 446
 447         if (followerToSnapshot.getChunkIndex() == reply.getChunkIndex()) {
 448             boolean wasLastChunk = false;
 449             if (reply.isSuccess()) {
 450                 if(followerToSnapshot.isLastChunk(reply.getChunkIndex())) {
 451                     //this was the last chunk reply
 452                     if(LOG.isDebugEnabled()) {
 453                         LOG.debug("{}: InstallSnapshotReply received, " +
 454                                 "last chunk received, Chunk: {}. Follower: {} Setting nextIndex: {}",
 455                                 logName(), reply.getChunkIndex(), followerId,
 456                             context.getReplicatedLog().getSnapshotIndex() + 1
 457                         );
 458                     }
 459
 460                     long followerMatchIndex = snapshot.get().getLastIncludedIndex();
 461                     followerLogInformation.setMatchIndex(followerMatchIndex);
 462                     followerLogInformation.setNextIndex(followerMatchIndex + 1);
 463                     mapFollowerToSnapshot.remove(followerId);
 464
 465                     LOG.debug("{}: follower: {}, matchIndex set to {}, nextIndex set to {}",
 466                         logName(), followerId, followerLogInformation.getMatchIndex(),
 467                         followerLogInformation.getNextIndex());
 468
 469                     if (mapFollowerToSnapshot.isEmpty()) {
 470                         // once there are no pending followers receiving snapshots
 471                         // we can remove snapshot from the memory
 472                         setSnapshot(null);
 473                     }
 474                     wasLastChunk = true;
 475                     if(context.getPeerInfo(followerId).getVotingState() == VotingState.VOTING_NOT_INITIALIZED){
 476                         UnInitializedFollowerSnapshotReply unInitFollowerSnapshotSuccess =
 477                                              new UnInitializedFollowerSnapshotReply(followerId);
 478                         context.getActor().tell(unInitFollowerSnapshotSuccess, context.getActor());
 479                         LOG.debug("Sent message UnInitializedFollowerSnapshotReply to self");
 480                     }
 481                 } else {
 482                     followerToSnapshot.markSendStatus(true);
 483                 }
 484             } else {
 485                 LOG.info("{}: InstallSnapshotReply received sending snapshot chunk failed, Will retry, Chunk: {}",
 486                         logName(), reply.getChunkIndex());
 487
 488                 followerToSnapshot.markSendStatus(false);
 489             }
 490
 491             if (wasLastChunk && !context.getSnapshotManager().isCapturing()) {
 492                 // Since the follower is now caught up try to purge the log.
 493                 purgeInMemoryLog();
 494             } else if (!wasLastChunk && followerToSnapshot.canSendNextChunk()) {
 495                 ActorSelection followerActor = context.getPeerActorSelection(followerId);
 496                 if(followerActor != null) {
 497                     sendSnapshotChunk(followerActor, followerId);
 498                 }
 499             }
 500
 501         } else {
 502             LOG.error("{}: Chunk index {} in InstallSnapshotReply from follower {} does not match expected index {}",
 503                     logName(), reply.getChunkIndex(), followerId,
 504                     followerToSnapshot.getChunkIndex());
 505
 506             if(reply.getChunkIndex() == INVALID_CHUNK_INDEX){
 507                 // Since the Follower did not find this index to be valid we should reset the follower snapshot
 508                 // so that Installing the snapshot can resume from the beginning
 509                 followerToSnapshot.reset();
 510             }
 511         }
 512     }
 513
 514     private void replicate(Replicate replicate) {
 515         long logIndex = replicate.getReplicatedLogEntry().getIndex();
 516
 517         LOG.debug("{}: Replicate message: identifier: {}, logIndex: {}", logName(),
 518                 replicate.getIdentifier(), logIndex);
 519
 520         // Create a tracker entry we will use this later to notify the
 521         // client actor
 522         trackerList.add(
 523             new ClientRequestTrackerImpl(replicate.getClientActor(),
 524                 replicate.getIdentifier(),
 525                 logIndex)
 526         );
 527
 528         boolean applyModificationToState = followerToLog.isEmpty()
 529                 || context.getRaftPolicy().applyModificationToStateBeforeConsensus();
 530
 531         if(applyModificationToState){
 532             context.setCommitIndex(logIndex);
 533             applyLogToStateMachine(logIndex);
 534         }
 535
 536         if (!followerToLog.isEmpty()) {
 537             sendAppendEntries(0, false);
 538         }
 539     }
 540
 541     protected void sendAppendEntries(long timeSinceLastActivityInterval, boolean isHeartbeat) {
 542         // Send an AppendEntries to all followers
 543         for (Entry<String, FollowerLogInformation> e : followerToLog.entrySet()) {
 544             final String followerId = e.getKey();
 545             final FollowerLogInformation followerLogInformation = e.getValue();
 546             // This checks helps not to send a repeat message to the follower
 547             if(!followerLogInformation.isFollowerActive() ||
 548                     followerLogInformation.timeSinceLastActivity() >= timeSinceLastActivityInterval) {
 549                 sendUpdatesToFollower(followerId, followerLogInformation, true, isHeartbeat);
 550             }
 551         }
 552     }
 553
 554     /**
 555      *
 556      * This method checks if any update needs to be sent to the given follower. This includes append log entries,
 557      * sending next snapshot chunk, and initiating a snapshot.
 558      * @return true if any update is sent, false otherwise
 559      */
 560
 561     private void sendUpdatesToFollower(String followerId, FollowerLogInformation followerLogInformation,
 562                                        boolean sendHeartbeat, boolean isHeartbeat) {
 563
 564         ActorSelection followerActor = context.getPeerActorSelection(followerId);
 565         if (followerActor != null) {
 566             long followerNextIndex = followerLogInformation.getNextIndex();
 567             boolean isFollowerActive = followerLogInformation.isFollowerActive();
 568             boolean sendAppendEntries = false;
 569             List<ReplicatedLogEntry> entries = Collections.emptyList();
 570
 571             if (mapFollowerToSnapshot.get(followerId) != null) {
 572                 // if install snapshot is in process , then sent next chunk if possible
 573                 if (isFollowerActive && mapFollowerToSnapshot.get(followerId).canSendNextChunk()) {
 574                     sendSnapshotChunk(followerActor, followerId);
 575                 } else if(sendHeartbeat) {
 576                     // we send a heartbeat even if we have not received a reply for the last chunk
 577                     sendAppendEntries = true;
 578                 }
 579             } else {
 580                 long leaderLastIndex = context.getReplicatedLog().lastIndex();
 581                 long leaderSnapShotIndex = context.getReplicatedLog().getSnapshotIndex();
 582
 583                 if((!isHeartbeat && LOG.isDebugEnabled()) || LOG.isTraceEnabled()) {
 584                     LOG.debug("{}: Checking sendAppendEntries for follower {}: active: {}, followerNextIndex: {}, leaderLastIndex: {}, leaderSnapShotIndex: {}",
 585                             logName(), followerId, isFollowerActive, followerNextIndex, leaderLastIndex, leaderSnapShotIndex);
 586                 }
 587
 588                 if (isFollowerActive && context.getReplicatedLog().isPresent(followerNextIndex)) {
 589
 590                     LOG.debug("{}: sendAppendEntries: {} is present for follower {}", logName(),
 591                             followerNextIndex, followerId);
 592
 593                     if(followerLogInformation.okToReplicate()) {
 594                         // Try to send all the entries in the journal but not exceeding the max data size
 595                         // for a single AppendEntries message.
 596                         int maxEntries = (int) context.getReplicatedLog().size();
 597                         entries = context.getReplicatedLog().getFrom(followerNextIndex, maxEntries,
 598                                 context.getConfigParams().getSnapshotChunkSize());
 599                         sendAppendEntries = true;
 600                     }
 601                 } else if (isFollowerActive && followerNextIndex >= 0 &&
 602                     leaderLastIndex > followerNextIndex && !context.getSnapshotManager().isCapturing()) {
 603                     // if the followers next index is not present in the leaders log, and
 604                     // if the follower is just not starting and if leader's index is more than followers index
 605                     // then snapshot should be sent
 606
 607                     if (LOG.isDebugEnabled()) {
 608                         LOG.debug(String.format("%s: InitiateInstallSnapshot to follower: %s," +
 609                                     "follower-nextIndex: %d, leader-snapshot-index: %d,  " +
 610                                     "leader-last-index: %d", logName(), followerId,
 611                                     followerNextIndex, leaderSnapShotIndex, leaderLastIndex));
 612                     }
 613
 614                     // Send heartbeat to follower whenever install snapshot is initiated.
 615                     sendAppendEntries = true;
 616                     if (canInstallSnapshot(followerNextIndex)) {
 617                         initiateCaptureSnapshot(followerId);
 618                     }
 619
 620                 } else if(sendHeartbeat) {
 621                     // we send an AppendEntries, even if the follower is inactive
 622                     // in-order to update the followers timestamp, in case it becomes active again
 623                     sendAppendEntries = true;
 624                 }
 625
 626             }
 627
 628             if(sendAppendEntries) {
 629                 sendAppendEntriesToFollower(followerActor, followerNextIndex,
 630                         entries, followerId);
 631             }
 632         }
 633     }
 634
 635     private void sendAppendEntriesToFollower(ActorSelection followerActor, long followerNextIndex,
 636         List<ReplicatedLogEntry> entries, String followerId) {
 637         AppendEntries appendEntries = new AppendEntries(currentTerm(), context.getId(),
 638             prevLogIndex(followerNextIndex),
 639             prevLogTerm(followerNextIndex), entries,
 640             context.getCommitIndex(), super.getReplicatedToAllIndex(), context.getPayloadVersion());
 641
 642         if(!entries.isEmpty() || LOG.isTraceEnabled()) {
 643             LOG.debug("{}: Sending AppendEntries to follower {}: {}", logName(), followerId,
 644                     appendEntries);
 645         }
 646
 647         followerActor.tell(appendEntries.toSerializable(), actor());
 648     }
 649
 650     /**
 651      * Install Snapshot works as follows
 652      * 1. Leader initiates the capture snapshot by sending a CaptureSnapshot message to actor
 653      * 2. RaftActor on receipt of the CaptureSnapshotReply (from Shard), stores the received snapshot in the replicated log
 654      * and makes a call to Leader's handleMessage , with SendInstallSnapshot message.
 655      * 3. Leader , picks the snapshot from im-mem ReplicatedLog and sends it in chunks to the Follower
 656      * 4. On complete, Follower sends back a InstallSnapshotReply.
 657      * 5. On receipt of the InstallSnapshotReply for the last chunk, Leader marks the install complete for that follower
 658      * and replenishes the memory by deleting the snapshot in Replicated log.
 659      * 6. If another follower requires a snapshot and a snapshot has been collected (via CaptureSnapshotReply)
 660      * then send the existing snapshot in chunks to the follower.
 661      * @param followerId
 662      */
 663     public boolean initiateCaptureSnapshot(String followerId) {
 664         if (snapshot.isPresent()) {
 665             // if a snapshot is present in the memory, most likely another install is in progress
 666             // no need to capture snapshot.
 667             // This could happen if another follower needs an install when one is going on.
 668             final ActorSelection followerActor = context.getPeerActorSelection(followerId);
 669             sendSnapshotChunk(followerActor, followerId);
 670             return true;
 671         } else {
 672             return context.getSnapshotManager().captureToInstall(context.getReplicatedLog().last(),
 673                     this.getReplicatedToAllIndex(), followerId);
 674         }
 675     }
 676
 677     private boolean canInstallSnapshot(long nextIndex){
 678         // If the follower's nextIndex is -1 then we might as well send it a snapshot
 679         // Otherwise send it a snapshot only if the nextIndex is not present in the log but is present
 680         // in the snapshot
 681         return (nextIndex == -1 ||
 682                 (!context.getReplicatedLog().isPresent(nextIndex)
 683                         && context.getReplicatedLog().isInSnapshot(nextIndex)));
 684
 685     }
 686
 687
 688     private void sendInstallSnapshot() {
 689         LOG.debug("{}: sendInstallSnapshot", logName());
 690         for (Entry<String, FollowerLogInformation> e : followerToLog.entrySet()) {
 691             String followerId = e.getKey();
 692             ActorSelection followerActor = context.getPeerActorSelection(followerId);
 693             FollowerLogInformation followerLogInfo = e.getValue();
 694
 695             if (followerActor != null) {
 696                 long nextIndex = followerLogInfo.getNextIndex();
 697                 if (context.getPeerInfo(followerId).getVotingState() == VotingState.VOTING_NOT_INITIALIZED ||
 698                         canInstallSnapshot(nextIndex)) {
 699                     sendSnapshotChunk(followerActor, followerId);
 700                 }
 701             }
 702         }
 703     }
 704
 705     /**
 706      *  Sends a snapshot chunk to a given follower
 707      *  InstallSnapshot should qualify as a heartbeat too.
 708      */
 709     private void sendSnapshotChunk(ActorSelection followerActor, String followerId) {
 710         try {
 711             if (snapshot.isPresent()) {
 712                 ByteString nextSnapshotChunk = getNextSnapshotChunk(followerId, snapshot.get().getSnapshotBytes());
 713
 714                 // Note: the previous call to getNextSnapshotChunk has the side-effect of adding
 715                 // followerId to the followerToSnapshot map.
 716                 FollowerToSnapshot followerToSnapshot = mapFollowerToSnapshot.get(followerId);
 717
 718                 followerActor.tell(
 719                     new InstallSnapshot(currentTerm(), context.getId(),
 720                         snapshot.get().getLastIncludedIndex(),
 721                         snapshot.get().getLastIncludedTerm(),
 722                         nextSnapshotChunk,
 723                         followerToSnapshot.incrementChunkIndex(),
 724                         followerToSnapshot.getTotalChunks(),
 725                         Optional.of(followerToSnapshot.getLastChunkHashCode())
 726                     ).toSerializable(),
 727                     actor()
 728                 );
 729
 730                 if(LOG.isDebugEnabled()) {
 731                     LOG.debug("{}: InstallSnapshot sent to follower {}, Chunk: {}/{}",
 732                             logName(), followerActor.path(), followerToSnapshot.getChunkIndex(),
 733                             followerToSnapshot.getTotalChunks());
 734                 }
 735             }
 736         } catch (IOException e) {
 737             LOG.error("{}: InstallSnapshot failed for Leader.", logName(), e);
 738         }
 739     }
 740
 741     /**
 742      * Acccepts snaphot as ByteString, enters into map for future chunks
 743      * creates and return a ByteString chunk
 744      */
 745     private ByteString getNextSnapshotChunk(String followerId, ByteString snapshotBytes) throws IOException {
 746         FollowerToSnapshot followerToSnapshot = mapFollowerToSnapshot.get(followerId);
 747         if (followerToSnapshot == null) {
 748             followerToSnapshot = new FollowerToSnapshot(snapshotBytes);
 749             mapFollowerToSnapshot.put(followerId, followerToSnapshot);
 750         }
 751         ByteString nextChunk = followerToSnapshot.getNextChunk();
 752
 753         LOG.debug("{}: next snapshot chunk size for follower {}: {}", logName(), followerId, nextChunk.size());
 754
 755         return nextChunk;
 756     }
 757
 758     private void sendHeartBeat() {
 759         if (!followerToLog.isEmpty()) {
 760             LOG.trace("{}: Sending heartbeat", logName());
 761             sendAppendEntries(context.getConfigParams().getHeartBeatInterval().toMillis(), true);
 762         }
 763     }
 764
 765     private void stopHeartBeat() {
 766         if (heartbeatSchedule != null && !heartbeatSchedule.isCancelled()) {
 767             heartbeatSchedule.cancel();
 768         }
 769     }
 770
 771     private void scheduleHeartBeat(FiniteDuration interval) {
 772         if (followerToLog.isEmpty()) {
 773             // Optimization - do not bother scheduling a heartbeat as there are
 774             // no followers
 775             return;
 776         }
 777
 778         stopHeartBeat();
 779
 780         // Schedule a heartbeat. When the scheduler triggers a SendHeartbeat
 781         // message is sent to itself.
 782         // Scheduling the heartbeat only once here because heartbeats do not
 783         // need to be sent if there are other messages being sent to the remote
 784         // actor.
 785         heartbeatSchedule = context.getActorSystem().scheduler().scheduleOnce(
 786             interval, context.getActor(), new SendHeartBeat(),
 787             context.getActorSystem().dispatcher(), context.getActor());
 788     }
 789
 790     @Override
 791     public void close() throws Exception {
 792         stopHeartBeat();
 793     }
 794
 795     @Override
 796     public String getLeaderId() {
 797         return context.getId();
 798     }
 799
 800     protected boolean isLeaderIsolated() {
 801         int minPresent = getMinIsolatedLeaderPeerCount();
 802         for (FollowerLogInformation followerLogInformation : followerToLog.values()) {
 803             if (followerLogInformation.isFollowerActive()) {
 804                 --minPresent;
 805                 if (minPresent == 0) {
 806                     break;
 807                 }
 808             }
 809         }
 810         return (minPresent != 0);
 811     }
 812
 813     /**
 814      * Encapsulates the snapshot bytestring and handles the logic of sending
 815      * snapshot chunks
 816      */
 817     protected class FollowerToSnapshot {
 818         private final ByteString snapshotBytes;
 819         private int offset = 0;
 820         // the next snapshot chunk is sent only if the replyReceivedForOffset matches offset
 821         private int replyReceivedForOffset;
 822         // if replyStatus is false, the previous chunk is attempted
 823         private boolean replyStatus = false;
 824         private int chunkIndex;
 825         private final int totalChunks;
 826         private int lastChunkHashCode = AbstractLeader.INITIAL_LAST_CHUNK_HASH_CODE;
 827         private int nextChunkHashCode = AbstractLeader.INITIAL_LAST_CHUNK_HASH_CODE;
 828
 829         public FollowerToSnapshot(ByteString snapshotBytes) {
 830             this.snapshotBytes = snapshotBytes;
 831             int size = snapshotBytes.size();
 832             totalChunks = ( size / context.getConfigParams().getSnapshotChunkSize()) +
 833                 ((size % context.getConfigParams().getSnapshotChunkSize()) > 0 ? 1 : 0);
 834             if(LOG.isDebugEnabled()) {
 835                 LOG.debug("{}: Snapshot {} bytes, total chunks to send:{}",
 836                         logName(), size, totalChunks);
 837             }
 838             replyReceivedForOffset = -1;
 839             chunkIndex = AbstractLeader.FIRST_CHUNK_INDEX;
 840         }
 841
 842         public ByteString getSnapshotBytes() {
 843             return snapshotBytes;
 844         }
 845
 846         public int incrementOffset() {
 847             if(replyStatus) {
 848                 // if prev chunk failed, we would want to sent the same chunk again
 849                 offset = offset + context.getConfigParams().getSnapshotChunkSize();
 850             }
 851             return offset;
 852         }
 853
 854         public int incrementChunkIndex() {
 855             if (replyStatus) {
 856                 // if prev chunk failed, we would want to sent the same chunk again
 857                 chunkIndex =  chunkIndex + 1;
 858             }
 859             return chunkIndex;
 860         }
 861
 862         public int getChunkIndex() {
 863             return chunkIndex;
 864         }
 865
 866         public int getTotalChunks() {
 867             return totalChunks;
 868         }
 869
 870         public boolean canSendNextChunk() {
 871             // we only send a false if a chunk is sent but we have not received a reply yet
 872             return replyReceivedForOffset == offset;
 873         }
 874
 875         public boolean isLastChunk(int chunkIndex) {
 876             return totalChunks == chunkIndex;
 877         }
 878
 879         public void markSendStatus(boolean success) {
 880             if (success) {
 881                 // if the chunk sent was successful
 882                 replyReceivedForOffset = offset;
 883                 replyStatus = true;
 884                 lastChunkHashCode = nextChunkHashCode;
 885             } else {
 886                 // if the chunk sent was failure
 887                 replyReceivedForOffset = offset;
 888                 replyStatus = false;
 889             }
 890         }
 891
 892         public ByteString getNextChunk() {
 893             int snapshotLength = getSnapshotBytes().size();
 894             int start = incrementOffset();
 895             int size = context.getConfigParams().getSnapshotChunkSize();
 896             if (context.getConfigParams().getSnapshotChunkSize() > snapshotLength) {
 897                 size = snapshotLength;
 898             } else {
 899                 if ((start + context.getConfigParams().getSnapshotChunkSize()) > snapshotLength) {
 900                     size = snapshotLength - start;
 901                 }
 902             }
 903
 904
 905             LOG.debug("{}: Next chunk: length={}, offset={},size={}", logName(),
 906                     snapshotLength, start, size);
 907
 908             ByteString substring = getSnapshotBytes().substring(start, start + size);
 909             nextChunkHashCode = substring.hashCode();
 910             return substring;
 911         }
 912
 913         /**
 914          * reset should be called when the Follower needs to be sent the snapshot from the beginning
 915          */
 916         public void reset(){
 917             offset = 0;
 918             replyStatus = false;
 919             replyReceivedForOffset = offset;
 920             chunkIndex = AbstractLeader.FIRST_CHUNK_INDEX;
 921             lastChunkHashCode = AbstractLeader.INITIAL_LAST_CHUNK_HASH_CODE;
 922         }
 923
 924         public int getLastChunkHashCode() {
 925             return lastChunkHashCode;
 926         }
 927     }
 928
 929     // called from example-actor for printing the follower-states
 930     public String printFollowerStates() {
 931         final StringBuilder sb = new StringBuilder();
 932
 933         sb.append('[');
 934         for (FollowerLogInformation followerLogInformation : followerToLog.values()) {
 935             sb.append('{');
 936             sb.append(followerLogInformation.getId());
 937             sb.append(" state:");
 938             sb.append(followerLogInformation.isFollowerActive());
 939             sb.append("},");
 940         }
 941         sb.append(']');
 942
 943         return sb.toString();
 944     }
 945
 946     @VisibleForTesting
 947     public FollowerLogInformation getFollower(String followerId) {
 948         return followerToLog.get(followerId);
 949     }
 950
 951     @VisibleForTesting
 952     protected void setFollowerSnapshot(String followerId, FollowerToSnapshot snapshot) {
 953         mapFollowerToSnapshot.put(followerId, snapshot);
 954     }
 955
 956     @VisibleForTesting
 957     public int followerSnapshotSize() {
 958         return mapFollowerToSnapshot.size();
 959     }
 960
 961     @VisibleForTesting
 962     public int followerLogSize() {
 963         return followerToLog.size();
 964     }
 965
 966     private static class SnapshotHolder {
 967         private final long lastIncludedTerm;
 968         private final long lastIncludedIndex;
 969         private final ByteString snapshotBytes;
 970
 971         SnapshotHolder(Snapshot snapshot) {
 972             this.lastIncludedTerm = snapshot.getLastAppliedTerm();
 973             this.lastIncludedIndex = snapshot.getLastAppliedIndex();
 974             this.snapshotBytes = ByteString.copyFrom(snapshot.getState());
 975         }
 976
 977         long getLastIncludedTerm() {
 978             return lastIncludedTerm;
 979         }
 980
 981         long getLastIncludedIndex() {
 982             return lastIncludedIndex;
 983         }
 984
 985         ByteString getSnapshotBytes() {
 986             return snapshotBytes;
 987         }
 988     }
 989 }