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