opendaylight/md-sal/sal-akka-raft/src/main/java/org/opendaylight/controller/cluster/raft/behaviors/Leader.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.base.Preconditions;
  15 import org.opendaylight.controller.cluster.raft.ClientRequestTracker;
  16 import org.opendaylight.controller.cluster.raft.ClientRequestTrackerImpl;
  17 import org.opendaylight.controller.cluster.raft.FollowerLogInformation;
  18 import org.opendaylight.controller.cluster.raft.FollowerLogInformationImpl;
  19 import org.opendaylight.controller.cluster.raft.RaftActorContext;
  20 import org.opendaylight.controller.cluster.raft.RaftState;
  21 import org.opendaylight.controller.cluster.raft.ReplicatedLogEntry;
  22 import org.opendaylight.controller.cluster.raft.internal.messages.ApplyState;
  23 import org.opendaylight.controller.cluster.raft.internal.messages.Replicate;
  24 import org.opendaylight.controller.cluster.raft.internal.messages.SendHeartBeat;
  25 import org.opendaylight.controller.cluster.raft.internal.messages.SendInstallSnapshot;
  26 import org.opendaylight.controller.cluster.raft.messages.AppendEntries;
  27 import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
  28 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshot;
  29 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshotReply;
  30 import org.opendaylight.controller.cluster.raft.messages.RequestVoteReply;
  31 import scala.concurrent.duration.FiniteDuration;
  32
  33 import java.util.ArrayList;
  34 import java.util.Collections;
  35 import java.util.HashMap;
  36 import java.util.List;
  37 import java.util.Map;
  38 import java.util.concurrent.TimeUnit;
  39 import java.util.concurrent.atomic.AtomicLong;
  40
  41 /**
  42  * The behavior of a RaftActor when it is in the Leader state
  43  * <p/>
  44  * Leaders:
  45  * <ul>
  46  * <li> Upon election: send initial empty AppendEntries RPCs
  47  * (heartbeat) to each server; repeat during idle periods to
  48  * prevent election timeouts (§5.2)
  49  * <li> If command received from client: append entry to local log,
  50  * respond after entry applied to state machine (§5.3)
  51  * <li> If last log index ≥ nextIndex for a follower: send
  52  * AppendEntries RPC with log entries starting at nextIndex
  53  * <ul>
  54  * <li> If successful: update nextIndex and matchIndex for
  55  * follower (§5.3)
  56  * <li> If AppendEntries fails because of log inconsistency:
  57  * decrement nextIndex and retry (§5.3)
  58  * </ul>
  59  * <li> If there exists an N such that N > commitIndex, a majority
  60  * of matchIndex[i] ≥ N, and log[N].term == currentTerm:
  61  * set commitIndex = N (§5.3, §5.4).
  62  */
  63 public class Leader extends AbstractRaftActorBehavior {
  64
  65
  66     private final Map<String, FollowerLogInformation> followerToLog =
  67         new HashMap();
  68
  69     private final Map<String, ActorSelection> followerToActor = new HashMap<>();
  70
  71     private Cancellable heartbeatSchedule = null;
  72     private Cancellable appendEntriesSchedule = null;
  73     private Cancellable installSnapshotSchedule = null;
  74
  75     private List<ClientRequestTracker> trackerList = new ArrayList<>();
  76
  77     private final int minReplicationCount;
  78
  79     public Leader(RaftActorContext context) {
  80         super(context);
  81
  82         if (lastIndex() >= 0) {
  83             context.setCommitIndex(lastIndex());
  84         }
  85
  86         for (String followerId : context.getPeerAddresses().keySet()) {
  87             FollowerLogInformation followerLogInformation =
  88                 new FollowerLogInformationImpl(followerId,
  89                     new AtomicLong(lastIndex()),
  90                     new AtomicLong(-1));
  91
  92             followerToActor.put(followerId,
  93                 context.actorSelection(context.getPeerAddress(followerId)));
  94
  95             followerToLog.put(followerId, followerLogInformation);
  96
  97         }
  98
  99         if (followerToActor.size() > 0) {
 100             minReplicationCount = (followerToActor.size() + 1) / 2 + 1;
 101         } else {
 102             minReplicationCount = 0;
 103         }
 104
 105
 106         // Immediately schedule a heartbeat
 107         // Upon election: send initial empty AppendEntries RPCs
 108         // (heartbeat) to each server; repeat during idle periods to
 109         // prevent election timeouts (§5.2)
 110         scheduleHeartBeat(new FiniteDuration(0, TimeUnit.SECONDS));
 111
 112         scheduleInstallSnapshotCheck(
 113             new FiniteDuration(HEART_BEAT_INTERVAL.length() * 1000,
 114                 HEART_BEAT_INTERVAL.unit())
 115         );
 116
 117     }
 118
 119     @Override protected RaftState handleAppendEntries(ActorRef sender,
 120         AppendEntries appendEntries, RaftState suggestedState) {
 121
 122         context.getLogger()
 123             .error("An unexpected AppendEntries received in state " + state());
 124
 125         return suggestedState;
 126     }
 127
 128     @Override protected RaftState handleAppendEntriesReply(ActorRef sender,
 129         AppendEntriesReply appendEntriesReply, RaftState suggestedState) {
 130
 131         // Do not take any other action since a behavior change is coming
 132         if (suggestedState != state())
 133             return suggestedState;
 134
 135         // Update the FollowerLogInformation
 136         String followerId = appendEntriesReply.getFollowerId();
 137         FollowerLogInformation followerLogInformation =
 138             followerToLog.get(followerId);
 139         if (appendEntriesReply.isSuccess()) {
 140             followerLogInformation
 141                 .setMatchIndex(appendEntriesReply.getLogLastIndex());
 142             followerLogInformation
 143                 .setNextIndex(appendEntriesReply.getLogLastIndex() + 1);
 144         } else {
 145             followerLogInformation.decrNextIndex();
 146         }
 147
 148         // Now figure out if this reply warrants a change in the commitIndex
 149         // If there exists an N such that N > commitIndex, a majority
 150         // of matchIndex[i] ≥ N, and log[N].term == currentTerm:
 151         // set commitIndex = N (§5.3, §5.4).
 152         for (long N = context.getCommitIndex() + 1; ; N++) {
 153             int replicatedCount = 1;
 154
 155             for (FollowerLogInformation info : followerToLog.values()) {
 156                 if (info.getMatchIndex().get() >= N) {
 157                     replicatedCount++;
 158                 }
 159             }
 160
 161             if (replicatedCount >= minReplicationCount) {
 162                 ReplicatedLogEntry replicatedLogEntry =
 163                     context.getReplicatedLog().get(N);
 164                 if (replicatedLogEntry != null
 165                     && replicatedLogEntry.getTerm()
 166                     == currentTerm()) {
 167                     context.setCommitIndex(N);
 168                 }
 169             } else {
 170                 break;
 171             }
 172         }
 173
 174         // Apply the change to the state machine
 175         if (context.getCommitIndex() > context.getLastApplied()) {
 176             applyLogToStateMachine(context.getCommitIndex());
 177         }
 178
 179         return suggestedState;
 180     }
 181
 182     protected ClientRequestTracker findClientRequestTracker(long logIndex) {
 183         for (ClientRequestTracker tracker : trackerList) {
 184             if (tracker.getIndex() == logIndex) {
 185                 return tracker;
 186             }
 187         }
 188
 189         return null;
 190     }
 191
 192     @Override protected RaftState handleRequestVoteReply(ActorRef sender,
 193         RequestVoteReply requestVoteReply, RaftState suggestedState) {
 194         return suggestedState;
 195     }
 196
 197     @Override public RaftState state() {
 198         return RaftState.Leader;
 199     }
 200
 201     @Override public RaftState handleMessage(ActorRef sender, Object message) {
 202         Preconditions.checkNotNull(sender, "sender should not be null");
 203
 204         try {
 205             if (message instanceof SendHeartBeat) {
 206                 return sendHeartBeat();
 207             } else if(message instanceof SendInstallSnapshot) {
 208                 installSnapshotIfNeeded();
 209             } else if (message instanceof Replicate) {
 210                 replicate((Replicate) message);
 211             } else if (message instanceof InstallSnapshotReply){
 212                 // FIXME : Should I be checking the term here too?
 213                 handleInstallSnapshotReply(
 214                     (InstallSnapshotReply) message);
 215             }
 216         } finally {
 217             scheduleHeartBeat(HEART_BEAT_INTERVAL);
 218         }
 219
 220         return super.handleMessage(sender, message);
 221     }
 222
 223     private void handleInstallSnapshotReply(InstallSnapshotReply message) {
 224         InstallSnapshotReply reply = message;
 225         String followerId = reply.getFollowerId();
 226         FollowerLogInformation followerLogInformation =
 227             followerToLog.get(followerId);
 228
 229         followerLogInformation
 230             .setMatchIndex(context.getReplicatedLog().getSnapshotIndex());
 231         followerLogInformation
 232             .setNextIndex(context.getReplicatedLog().getSnapshotIndex() + 1);
 233     }
 234
 235     private void replicate(Replicate replicate) {
 236         long logIndex = replicate.getReplicatedLogEntry().getIndex();
 237
 238         context.getLogger().debug("Replicate message " + logIndex);
 239
 240         if (followerToActor.size() == 0) {
 241             context.setCommitIndex(
 242                 replicate.getReplicatedLogEntry().getIndex());
 243
 244             context.getActor()
 245                 .tell(new ApplyState(replicate.getClientActor(),
 246                         replicate.getIdentifier(),
 247                         replicate.getReplicatedLogEntry()),
 248                     context.getActor()
 249                 );
 250         } else {
 251
 252             // Create a tracker entry we will use this later to notify the
 253             // client actor
 254             trackerList.add(
 255                 new ClientRequestTrackerImpl(replicate.getClientActor(),
 256                     replicate.getIdentifier(),
 257                     logIndex)
 258             );
 259
 260             sendAppendEntries();
 261         }
 262     }
 263
 264     private void sendAppendEntries() {
 265         // Send an AppendEntries to all followers
 266         for (String followerId : followerToActor.keySet()) {
 267             ActorSelection followerActor =
 268                 followerToActor.get(followerId);
 269
 270             FollowerLogInformation followerLogInformation =
 271                 followerToLog.get(followerId);
 272
 273             long nextIndex = followerLogInformation.getNextIndex().get();
 274
 275             List<ReplicatedLogEntry> entries = Collections.emptyList();
 276
 277             if(context.getReplicatedLog().isPresent(nextIndex)){
 278                 entries =
 279                     context.getReplicatedLog().getFrom(nextIndex);
 280             }
 281
 282             followerActor.tell(
 283                 new AppendEntries(currentTerm(), context.getId(),
 284                     prevLogIndex(nextIndex), prevLogTerm(nextIndex),
 285                     entries, context.getCommitIndex()
 286                 ),
 287                 actor()
 288             );
 289         }
 290     }
 291
 292     private void installSnapshotIfNeeded(){
 293         for (String followerId : followerToActor.keySet()) {
 294             ActorSelection followerActor =
 295                 followerToActor.get(followerId);
 296
 297             FollowerLogInformation followerLogInformation =
 298                 followerToLog.get(followerId);
 299
 300             long nextIndex = followerLogInformation.getNextIndex().get();
 301
 302             if(!context.getReplicatedLog().isPresent(nextIndex) && context.getReplicatedLog().isInSnapshot(nextIndex)){
 303                 followerActor.tell(
 304                     new InstallSnapshot(currentTerm(), context.getId(),
 305                         context.getReplicatedLog().getSnapshotIndex(),
 306                         context.getReplicatedLog().getSnapshotTerm(),
 307                         context.getReplicatedLog().getSnapshot()
 308                     ),
 309                     actor()
 310                 );
 311             }
 312         }
 313     }
 314
 315     private RaftState sendHeartBeat() {
 316         if (followerToActor.size() > 0) {
 317             sendAppendEntries();
 318         }
 319         return state();
 320     }
 321
 322     private void stopHeartBeat() {
 323         if (heartbeatSchedule != null && !heartbeatSchedule.isCancelled()) {
 324             heartbeatSchedule.cancel();
 325         }
 326     }
 327
 328     private void stopInstallSnapshotSchedule() {
 329         if (installSnapshotSchedule != null && !installSnapshotSchedule.isCancelled()) {
 330             installSnapshotSchedule.cancel();
 331         }
 332     }
 333
 334     private void scheduleHeartBeat(FiniteDuration interval) {
 335         if(followerToActor.keySet().size() == 0){
 336             // Optimization - do not bother scheduling a heartbeat as there are
 337             // no followers
 338             return;
 339         }
 340
 341         stopHeartBeat();
 342
 343         // Schedule a heartbeat. When the scheduler triggers a SendHeartbeat
 344         // message is sent to itself.
 345         // Scheduling the heartbeat only once here because heartbeats do not
 346         // need to be sent if there are other messages being sent to the remote
 347         // actor.
 348         heartbeatSchedule =
 349             context.getActorSystem().scheduler().scheduleOnce(
 350                 interval,
 351                 context.getActor(), new SendHeartBeat(),
 352                 context.getActorSystem().dispatcher(), context.getActor());
 353     }
 354
 355
 356     private void scheduleInstallSnapshotCheck(FiniteDuration interval) {
 357         if(followerToActor.keySet().size() == 0){
 358             // Optimization - do not bother scheduling a heartbeat as there are
 359             // no followers
 360             return;
 361         }
 362
 363         stopInstallSnapshotSchedule();
 364
 365         // Schedule a message to send append entries to followers that can
 366         // accept an append entries with some data in it
 367         installSnapshotSchedule =
 368             context.getActorSystem().scheduler().scheduleOnce(
 369                 interval,
 370                 context.getActor(), new SendInstallSnapshot(),
 371                 context.getActorSystem().dispatcher(), context.getActor());
 372     }
 373
 374
 375
 376     @Override public void close() throws Exception {
 377         stopHeartBeat();
 378     }
 379
 380     @Override public String getLeaderId() {
 381         return context.getId();
 382     }
 383
 384 }