opendaylight/md-sal/sal-akka-raft/src/main/java/org/opendaylight/controller/cluster/raft/behaviors/Follower.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.Address;
  14 import akka.cluster.Cluster;
  15 import akka.cluster.ClusterEvent.CurrentClusterState;
  16 import akka.cluster.Member;
  17 import akka.cluster.MemberStatus;
  18 import akka.japi.Procedure;
  19 import com.google.common.annotations.VisibleForTesting;
  20 import com.google.common.base.Stopwatch;
  21 import java.io.IOException;
  22 import java.util.ArrayList;
  23 import java.util.List;
  24 import java.util.Optional;
  25 import java.util.Set;
  26 import java.util.concurrent.TimeUnit;
  27 import java.util.concurrent.atomic.AtomicBoolean;
  28 import javax.annotation.Nullable;
  29 import org.opendaylight.controller.cluster.raft.RaftActorContext;
  30 import org.opendaylight.controller.cluster.raft.RaftState;
  31 import org.opendaylight.controller.cluster.raft.ReplicatedLogEntry;
  32 import org.opendaylight.controller.cluster.raft.base.messages.ApplySnapshot;
  33 import org.opendaylight.controller.cluster.raft.base.messages.ElectionTimeout;
  34 import org.opendaylight.controller.cluster.raft.base.messages.TimeoutNow;
  35 import org.opendaylight.controller.cluster.raft.messages.AppendEntries;
  36 import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
  37 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshot;
  38 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshotReply;
  39 import org.opendaylight.controller.cluster.raft.messages.RaftRPC;
  40 import org.opendaylight.controller.cluster.raft.messages.RequestVote;
  41 import org.opendaylight.controller.cluster.raft.messages.RequestVoteReply;
  42 import org.opendaylight.controller.cluster.raft.persisted.ServerConfigurationPayload;
  43 import org.opendaylight.controller.cluster.raft.persisted.Snapshot;
  44
  45 /**
  46  * The behavior of a RaftActor in the Follower raft state.
  47  * <ul>
  48  * <li> Respond to RPCs from candidates and leaders
  49  * <li> If election timeout elapses without receiving AppendEntries
  50  * RPC from current leader or granting vote to candidate:
  51  * convert to candidate
  52  * </ul>
  53  */
  54 public class Follower extends AbstractRaftActorBehavior {
  55     private static final long MAX_ELECTION_TIMEOUT_FACTOR = 18;
  56
  57     private final SyncStatusTracker initialSyncStatusTracker;
  58
  59     private final Stopwatch lastLeaderMessageTimer = Stopwatch.createStarted();
  60     private SnapshotTracker snapshotTracker = null;
  61     private String leaderId;
  62     private short leaderPayloadVersion;
  63
  64     public Follower(final RaftActorContext context) {
  65         this(context, null, (short)-1);
  66     }
  67
  68     public Follower(final RaftActorContext context, final String initialLeaderId,
  69             final short initialLeaderPayloadVersion) {
  70         super(context, RaftState.Follower);
  71         this.leaderId = initialLeaderId;
  72         this.leaderPayloadVersion = initialLeaderPayloadVersion;
  73
  74         initialSyncStatusTracker = new SyncStatusTracker(context.getActor(), getId(), context.getConfigParams()
  75             .getSyncIndexThreshold());
  76
  77         if (context.getPeerIds().isEmpty() && getLeaderId() == null) {
  78             actor().tell(TimeoutNow.INSTANCE, actor());
  79         } else {
  80             scheduleElection(electionDuration());
  81         }
  82     }
  83
  84     @Override
  85     public final String getLeaderId() {
  86         return leaderId;
  87     }
  88
  89     @VisibleForTesting
  90     protected final void setLeaderId(@Nullable final String leaderId) {
  91         this.leaderId = leaderId;
  92     }
  93
  94     @Override
  95     public short getLeaderPayloadVersion() {
  96         return leaderPayloadVersion;
  97     }
  98
  99     @VisibleForTesting
 100     protected final void setLeaderPayloadVersion(final short leaderPayloadVersion) {
 101         this.leaderPayloadVersion = leaderPayloadVersion;
 102     }
 103
 104     private void restartLastLeaderMessageTimer() {
 105         if (lastLeaderMessageTimer.isRunning()) {
 106             lastLeaderMessageTimer.reset();
 107         }
 108
 109         lastLeaderMessageTimer.start();
 110     }
 111
 112     private boolean isLogEntryPresent(final long index) {
 113         if (context.getReplicatedLog().isInSnapshot(index)) {
 114             return true;
 115         }
 116
 117         ReplicatedLogEntry entry = context.getReplicatedLog().get(index);
 118         return entry != null;
 119
 120     }
 121
 122     private void updateInitialSyncStatus(final long currentLeaderCommit, final String newLeaderId) {
 123         initialSyncStatusTracker.update(newLeaderId, currentLeaderCommit, context.getCommitIndex());
 124     }
 125
 126     @Override
 127     protected RaftActorBehavior handleAppendEntries(final ActorRef sender, final AppendEntries appendEntries) {
 128
 129         int numLogEntries = appendEntries.getEntries() != null ? appendEntries.getEntries().size() : 0;
 130         if (log.isTraceEnabled()) {
 131             log.trace("{}: handleAppendEntries: {}", logName(), appendEntries);
 132         } else if (log.isDebugEnabled() && numLogEntries > 0) {
 133             log.debug("{}: handleAppendEntries: {}", logName(), appendEntries);
 134         }
 135
 136         // TODO : Refactor this method into a bunch of smaller methods
 137         // to make it easier to read. Before refactoring ensure tests
 138         // cover the code properly
 139
 140         if (snapshotTracker != null && !snapshotTracker.getLeaderId().equals(appendEntries.getLeaderId())) {
 141             log.debug("{}: snapshot install is in progress but the prior snapshot leaderId {} does not match the "
 142                 + "AppendEntries leaderId {}", logName(), snapshotTracker.getLeaderId(), appendEntries.getLeaderId());
 143             closeSnapshotTracker();
 144         }
 145
 146         if (snapshotTracker != null || context.getSnapshotManager().isApplying()) {
 147             // if snapshot install is in progress, follower should just acknowledge append entries with a reply.
 148             AppendEntriesReply reply = new AppendEntriesReply(context.getId(), currentTerm(), true,
 149                     lastIndex(), lastTerm(), context.getPayloadVersion());
 150
 151             log.debug("{}: snapshot install is in progress, replying immediately with {}", logName(), reply);
 152             sender.tell(reply, actor());
 153
 154             return this;
 155         }
 156
 157         // If we got here then we do appear to be talking to the leader
 158         leaderId = appendEntries.getLeaderId();
 159         leaderPayloadVersion = appendEntries.getPayloadVersion();
 160
 161         // First check if the logs are in sync or not
 162         long lastIndex = lastIndex();
 163
 164         if (isOutOfSync(appendEntries)) {
 165             // We found that the log was out of sync so just send a negative
 166             // reply and return
 167
 168             final AppendEntriesReply reply = new AppendEntriesReply(context.getId(), currentTerm(), false, lastIndex,
 169                     lastTerm(), context.getPayloadVersion());
 170
 171             log.info("{}: Follower is out-of-sync so sending negative reply: {}", logName(), reply);
 172             updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
 173             sender.tell(reply, actor());
 174             return this;
 175         }
 176
 177         if (appendEntries.getEntries() != null && appendEntries.getEntries().size() > 0) {
 178
 179             log.debug("{}: Number of entries to be appended = {}", logName(),
 180                         appendEntries.getEntries().size());
 181
 182             // 3. If an existing entry conflicts with a new one (same index
 183             // but different terms), delete the existing entry and all that
 184             // follow it (§5.3)
 185             int addEntriesFrom = 0;
 186             if (context.getReplicatedLog().size() > 0) {
 187
 188                 // Find the entry up until the one that is not in the follower's log
 189                 for (int i = 0;i < appendEntries.getEntries().size(); i++, addEntriesFrom++) {
 190                     ReplicatedLogEntry matchEntry = appendEntries.getEntries().get(i);
 191
 192                     if (!isLogEntryPresent(matchEntry.getIndex())) {
 193                         // newEntry not found in the log
 194                         break;
 195                     }
 196
 197                     long existingEntryTerm = getLogEntryTerm(matchEntry.getIndex());
 198
 199                     log.debug("{}: matchEntry {} is present: existingEntryTerm: {}", logName(), matchEntry,
 200                             existingEntryTerm);
 201
 202                     // existingEntryTerm == -1 means it's in the snapshot and not in the log. We don't know
 203                     // what the term was so we'll assume it matches.
 204                     if (existingEntryTerm == -1 || existingEntryTerm == matchEntry.getTerm()) {
 205                         continue;
 206                     }
 207
 208                     if (!context.getRaftPolicy().applyModificationToStateBeforeConsensus()) {
 209
 210                         log.info("{}: Removing entries from log starting at {}", logName(), matchEntry.getIndex());
 211
 212                         // Entries do not match so remove all subsequent entries
 213                         if (!context.getReplicatedLog().removeFromAndPersist(matchEntry.getIndex())) {
 214                             // Could not remove the entries - this means the matchEntry index must be in the
 215                             // snapshot and not the log. In this case the prior entries are part of the state
 216                             // so we must send back a reply to force a snapshot to completely re-sync the
 217                             // follower's log and state.
 218
 219                             log.info("{}: Could not remove entries - sending reply to force snapshot", logName());
 220                             updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
 221                             sender.tell(new AppendEntriesReply(context.getId(), currentTerm(), false, lastIndex,
 222                                     lastTerm(), context.getPayloadVersion(), true), actor());
 223                             return this;
 224                         }
 225
 226                         break;
 227                     } else {
 228                         updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
 229                         sender.tell(new AppendEntriesReply(context.getId(), currentTerm(), false, lastIndex,
 230                                 lastTerm(), context.getPayloadVersion(), true), actor());
 231                         return this;
 232                     }
 233                 }
 234             }
 235
 236             lastIndex = lastIndex();
 237             log.debug("{}: After cleanup, lastIndex: {}, entries to be added from: {}", logName(),
 238                     lastIndex, addEntriesFrom);
 239
 240             // When persistence successfully completes for each new log entry appended, we need to determine if we
 241             // should capture a snapshot to compact the persisted log. shouldCaptureSnapshot tracks whether or not
 242             // one of the log entries has exceeded the log size threshold whereby a snapshot should be taken. However
 243             // we don't initiate the snapshot at that log entry but rather after the last log entry has been persisted.
 244             // This is done because subsequent log entries after the one that tripped the threshold may have been
 245             // applied to the state already, as the persistence callback occurs async, and we want those entries
 246             // purged from the persisted log as well.
 247             final AtomicBoolean shouldCaptureSnapshot = new AtomicBoolean(false);
 248             final Procedure<ReplicatedLogEntry> appendAndPersistCallback = logEntry -> {
 249                 final ReplicatedLogEntry lastEntryToAppend = appendEntries.getEntries().get(
 250                         appendEntries.getEntries().size() - 1);
 251                 if (shouldCaptureSnapshot.get() && logEntry == lastEntryToAppend) {
 252                     context.getSnapshotManager().capture(context.getReplicatedLog().last(), getReplicatedToAllIndex());
 253                 }
 254             };
 255
 256             // 4. Append any new entries not already in the log
 257             for (int i = addEntriesFrom; i < appendEntries.getEntries().size(); i++) {
 258                 ReplicatedLogEntry entry = appendEntries.getEntries().get(i);
 259
 260                 log.debug("{}: Append entry to log {}", logName(), entry.getData());
 261
 262                 context.getReplicatedLog().appendAndPersist(entry, appendAndPersistCallback, false);
 263
 264                 shouldCaptureSnapshot.compareAndSet(false,
 265                         context.getReplicatedLog().shouldCaptureSnapshot(entry.getIndex()));
 266
 267                 if (entry.getData() instanceof ServerConfigurationPayload) {
 268                     context.updatePeerIds((ServerConfigurationPayload)entry.getData());
 269                 }
 270             }
 271
 272             log.debug("{}: Log size is now {}", logName(), context.getReplicatedLog().size());
 273         }
 274
 275         // 5. If leaderCommit > commitIndex, set commitIndex =
 276         // min(leaderCommit, index of last new entry)
 277
 278         lastIndex = lastIndex();
 279         long prevCommitIndex = context.getCommitIndex();
 280
 281         if (appendEntries.getLeaderCommit() > prevCommitIndex) {
 282             context.setCommitIndex(Math.min(appendEntries.getLeaderCommit(), lastIndex));
 283         }
 284
 285         if (prevCommitIndex != context.getCommitIndex()) {
 286             log.debug("{}: Commit index set to {}", logName(), context.getCommitIndex());
 287         }
 288
 289         AppendEntriesReply reply = new AppendEntriesReply(context.getId(), currentTerm(), true,
 290                 lastIndex, lastTerm(), context.getPayloadVersion());
 291
 292         if (log.isTraceEnabled()) {
 293             log.trace("{}: handleAppendEntries returning : {}", logName(), reply);
 294         } else if (log.isDebugEnabled() && numLogEntries > 0) {
 295             log.debug("{}: handleAppendEntries returning : {}", logName(), reply);
 296         }
 297
 298         // Reply to the leader before applying any previous state so as not to hold up leader consensus.
 299         updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
 300         sender.tell(reply, actor());
 301
 302         // If commitIndex > lastApplied: increment lastApplied, apply
 303         // log[lastApplied] to state machine (§5.3)
 304         // check if there are any entries to be applied. last-applied can be equal to last-index
 305         if (appendEntries.getLeaderCommit() > context.getLastApplied()
 306                 && context.getLastApplied() < lastIndex) {
 307             if (log.isDebugEnabled()) {
 308                 log.debug("{}: applyLogToStateMachine, appendEntries.getLeaderCommit(): {}, "
 309                         + "context.getLastApplied(): {}, lastIndex(): {}", logName(),
 310                     appendEntries.getLeaderCommit(), context.getLastApplied(), lastIndex);
 311             }
 312
 313             applyLogToStateMachine(appendEntries.getLeaderCommit());
 314         }
 315
 316         if (!context.getSnapshotManager().isCapturing()) {
 317             super.performSnapshotWithoutCapture(appendEntries.getReplicatedToAllIndex());
 318         }
 319
 320         return this;
 321     }
 322
 323     private boolean isOutOfSync(final AppendEntries appendEntries) {
 324
 325         final long lastIndex = lastIndex();
 326         if (lastIndex == -1 && appendEntries.getPrevLogIndex() != -1) {
 327
 328             // The follower's log is out of sync because the leader does have an entry at prevLogIndex and this
 329             // follower has no entries in it's log.
 330
 331             log.info("{}: The followers log is empty and the senders prevLogIndex is {}", logName(),
 332                 appendEntries.getPrevLogIndex());
 333             return true;
 334         }
 335
 336         if (lastIndex > -1) {
 337             if (isLogEntryPresent(appendEntries.getPrevLogIndex())) {
 338                 final long prevLogTerm = getLogEntryTerm(appendEntries.getPrevLogIndex());
 339                 if (prevLogTerm != appendEntries.getPrevLogTerm()) {
 340
 341                     // The follower's log is out of sync because the Leader's prevLogIndex entry does exist
 342                     // in the follower's log but it has a different term in it
 343
 344                     log.info("{}: The prevLogIndex {} was found in the log but the term {} is not equal to the append "
 345                             + "entries prevLogTerm {} - lastIndex: {}, snapshotIndex: {}", logName(),
 346                             appendEntries.getPrevLogIndex(), prevLogTerm, appendEntries.getPrevLogTerm(), lastIndex,
 347                             context.getReplicatedLog().getSnapshotIndex());
 348                     return true;
 349                 }
 350             } else if (appendEntries.getPrevLogIndex() != -1) {
 351
 352                 // The follower's log is out of sync because the Leader's prevLogIndex entry was not found in it's log
 353
 354                 log.info("{}: The log is not empty but the prevLogIndex {} was not found in it - lastIndex: {}, "
 355                         + "snapshotIndex: {}", logName(), appendEntries.getPrevLogIndex(), lastIndex,
 356                         context.getReplicatedLog().getSnapshotIndex());
 357                 return true;
 358             }
 359         }
 360
 361         if (appendEntries.getPrevLogIndex() == -1 && appendEntries.getPrevLogTerm() == -1
 362                 && appendEntries.getReplicatedToAllIndex() != -1) {
 363             if (!isLogEntryPresent(appendEntries.getReplicatedToAllIndex())) {
 364                 // This append entry comes from a leader who has it's log aggressively trimmed and so does not have
 365                 // the previous entry in it's in-memory journal
 366
 367                 log.info("{}: Cannot append entries because the replicatedToAllIndex {} does not appear to be in the "
 368                         + "in-memory journal", logName(), appendEntries.getReplicatedToAllIndex());
 369                 return true;
 370             }
 371
 372             final List<ReplicatedLogEntry> entries = appendEntries.getEntries();
 373             if (entries != null && entries.size() > 0 && !isLogEntryPresent(entries.get(0).getIndex() - 1)) {
 374                 log.info("{}: Cannot append entries because the calculated previousIndex {} was not found in the "
 375                         + "in-memory journal", logName(), entries.get(0).getIndex() - 1);
 376                 return true;
 377             }
 378         }
 379
 380         return false;
 381     }
 382
 383     @Override
 384     protected RaftActorBehavior handleAppendEntriesReply(final ActorRef sender,
 385         final AppendEntriesReply appendEntriesReply) {
 386         return this;
 387     }
 388
 389     @Override
 390     protected RaftActorBehavior handleRequestVoteReply(final ActorRef sender,
 391         final RequestVoteReply requestVoteReply) {
 392         return this;
 393     }
 394
 395     @Override
 396     public RaftActorBehavior handleMessage(final ActorRef sender, final Object message) {
 397         if (message instanceof ElectionTimeout || message instanceof TimeoutNow) {
 398             return handleElectionTimeout(message);
 399         }
 400
 401         if (!(message instanceof RaftRPC)) {
 402             // The rest of the processing requires the message to be a RaftRPC
 403             return null;
 404         }
 405
 406         final RaftRPC rpc = (RaftRPC) message;
 407         // If RPC request or response contains term T > currentTerm:
 408         // set currentTerm = T, convert to follower (§5.1)
 409         // This applies to all RPC messages and responses
 410         if (rpc.getTerm() > context.getTermInformation().getCurrentTerm()) {
 411             log.info("{}: Term {} in \"{}\" message is greater than follower's term {} - updating term",
 412                 logName(), rpc.getTerm(), rpc, context.getTermInformation().getCurrentTerm());
 413
 414             context.getTermInformation().updateAndPersist(rpc.getTerm(), null);
 415         }
 416
 417         if (rpc instanceof InstallSnapshot) {
 418             handleInstallSnapshot(sender, (InstallSnapshot) rpc);
 419             restartLastLeaderMessageTimer();
 420             scheduleElection(electionDuration());
 421             return this;
 422         }
 423
 424         if (!(rpc instanceof RequestVote) || canGrantVote((RequestVote) rpc)) {
 425             restartLastLeaderMessageTimer();
 426             scheduleElection(electionDuration());
 427         }
 428
 429         return super.handleMessage(sender, rpc);
 430     }
 431
 432     private RaftActorBehavior handleElectionTimeout(final Object message) {
 433         // If the message is ElectionTimeout, verify we haven't actually seen a message from the leader
 434         // during the election timeout interval. It may that the election timer expired b/c this actor
 435         // was busy and messages got delayed, in which case leader messages would be backed up in the
 436         // queue but would be processed before the ElectionTimeout message and thus would restart the
 437         // lastLeaderMessageTimer.
 438         long lastLeaderMessageInterval = lastLeaderMessageTimer.elapsed(TimeUnit.MILLISECONDS);
 439         long electionTimeoutInMillis = context.getConfigParams().getElectionTimeOutInterval().toMillis();
 440         boolean noLeaderMessageReceived = !lastLeaderMessageTimer.isRunning()
 441                 || lastLeaderMessageInterval >= electionTimeoutInMillis;
 442
 443         if (canStartElection()) {
 444             if (message instanceof TimeoutNow) {
 445                 log.debug("{}: Received TimeoutNow - switching to Candidate", logName());
 446                 return internalSwitchBehavior(RaftState.Candidate);
 447             } else if (noLeaderMessageReceived) {
 448                 // Check the cluster state to see if the leader is known to be up before we go to Candidate.
 449                 // However if we haven't heard from the leader in a long time even though the cluster state
 450                 // indicates it's up then something is wrong - leader might be stuck indefinitely - so switch
 451                 // to Candidate,
 452                 long maxElectionTimeout = electionTimeoutInMillis * MAX_ELECTION_TIMEOUT_FACTOR;
 453                 if (isLeaderAvailabilityKnown() && lastLeaderMessageInterval < maxElectionTimeout) {
 454                     log.debug("{}: Received ElectionTimeout but leader appears to be available", logName());
 455                     scheduleElection(electionDuration());
 456                 } else {
 457                     log.debug("{}: Received ElectionTimeout - switching to Candidate", logName());
 458                     return internalSwitchBehavior(RaftState.Candidate);
 459                 }
 460             } else {
 461                 log.debug("{}: Received ElectionTimeout but lastLeaderMessageInterval {} < election timeout {}",
 462                         logName(), lastLeaderMessageInterval, context.getConfigParams().getElectionTimeOutInterval());
 463                 scheduleElection(electionDuration());
 464             }
 465         } else if (message instanceof ElectionTimeout) {
 466             if (noLeaderMessageReceived) {
 467                 setLeaderId(null);
 468             }
 469
 470             scheduleElection(electionDuration());
 471         }
 472
 473         return this;
 474     }
 475
 476     private boolean isLeaderAvailabilityKnown() {
 477         if (leaderId == null) {
 478             return false;
 479         }
 480
 481         Optional<Cluster> cluster = context.getCluster();
 482         if (!cluster.isPresent()) {
 483             return false;
 484         }
 485
 486         ActorSelection leaderActor = context.getPeerActorSelection(leaderId);
 487         if (leaderActor == null) {
 488             return false;
 489         }
 490
 491         Address leaderAddress = leaderActor.anchorPath().address();
 492
 493         CurrentClusterState state = cluster.get().state();
 494         Set<Member> unreachable = state.getUnreachable();
 495
 496         log.debug("{}: Checking for leader {} in the cluster unreachable set {}", logName(), leaderAddress,
 497                 unreachable);
 498
 499         for (Member m: unreachable) {
 500             if (leaderAddress.equals(m.address())) {
 501                 log.info("{}: Leader {} is unreachable", logName(), leaderAddress);
 502                 return false;
 503             }
 504         }
 505
 506         for (Member m: state.getMembers()) {
 507             if (leaderAddress.equals(m.address())) {
 508                 if (m.status() == MemberStatus.up() || m.status() == MemberStatus.weaklyUp()) {
 509                     log.debug("{}: Leader {} cluster status is {} - leader is available", logName(),
 510                             leaderAddress, m.status());
 511                     return true;
 512                 } else {
 513                     log.debug("{}: Leader {} cluster status is {} - leader is unavailable", logName(),
 514                             leaderAddress, m.status());
 515                     return false;
 516                 }
 517             }
 518         }
 519
 520         log.debug("{}: Leader {} not found in the cluster member set", logName(), leaderAddress);
 521
 522         return false;
 523     }
 524
 525     private void handleInstallSnapshot(final ActorRef sender, final InstallSnapshot installSnapshot) {
 526
 527         log.debug("{}: handleInstallSnapshot: {}", logName(), installSnapshot);
 528
 529         leaderId = installSnapshot.getLeaderId();
 530
 531         if (snapshotTracker == null) {
 532             snapshotTracker = new SnapshotTracker(log, installSnapshot.getTotalChunks(), installSnapshot.getLeaderId(),
 533                     context);
 534         }
 535
 536         updateInitialSyncStatus(installSnapshot.getLastIncludedIndex(), installSnapshot.getLeaderId());
 537
 538         try {
 539             final InstallSnapshotReply reply = new InstallSnapshotReply(
 540                     currentTerm(), context.getId(), installSnapshot.getChunkIndex(), true);
 541
 542             if (snapshotTracker.addChunk(installSnapshot.getChunkIndex(), installSnapshot.getData(),
 543                     installSnapshot.getLastChunkHashCode())) {
 544
 545                 log.info("{}: Snapshot installed from leader: {}", logName(), installSnapshot.getLeaderId());
 546
 547                 Snapshot snapshot = Snapshot.create(
 548                         context.getSnapshotManager().convertSnapshot(snapshotTracker.getSnapshotBytes()),
 549                         new ArrayList<>(),
 550                         installSnapshot.getLastIncludedIndex(),
 551                         installSnapshot.getLastIncludedTerm(),
 552                         installSnapshot.getLastIncludedIndex(),
 553                         installSnapshot.getLastIncludedTerm(),
 554                         context.getTermInformation().getCurrentTerm(),
 555                         context.getTermInformation().getVotedFor(),
 556                         installSnapshot.getServerConfig().orNull());
 557
 558                 ApplySnapshot.Callback applySnapshotCallback = new ApplySnapshot.Callback() {
 559                     @Override
 560                     public void onSuccess() {
 561                         log.debug("{}: handleInstallSnapshot returning: {}", logName(), reply);
 562
 563                         sender.tell(reply, actor());
 564                     }
 565
 566                     @Override
 567                     public void onFailure() {
 568                         sender.tell(new InstallSnapshotReply(currentTerm(), context.getId(), -1, false), actor());
 569                     }
 570                 };
 571
 572                 actor().tell(new ApplySnapshot(snapshot, applySnapshotCallback), actor());
 573
 574                 closeSnapshotTracker();
 575             } else {
 576                 log.debug("{}: handleInstallSnapshot returning: {}", logName(), reply);
 577
 578                 sender.tell(reply, actor());
 579             }
 580         } catch (IOException e) {
 581             log.debug("{}: Exception in InstallSnapshot of follower", logName(), e);
 582
 583             sender.tell(new InstallSnapshotReply(currentTerm(), context.getId(),
 584                     -1, false), actor());
 585
 586             closeSnapshotTracker();
 587         }
 588     }
 589
 590     private void closeSnapshotTracker() {
 591         if (snapshotTracker != null) {
 592             snapshotTracker.close();
 593             snapshotTracker = null;
 594         }
 595     }
 596
 597     @Override
 598     public void close() {
 599         closeSnapshotTracker();
 600         stopElection();
 601     }
 602
 603     @VisibleForTesting
 604     SnapshotTracker getSnapshotTracker() {
 605         return snapshotTracker;
 606     }
 607 }