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         context.getLogger().debug("Election:Leader has following peers:"+followerToActor.keySet());
  99
 100         if (followerToActor.size() > 0) {
 101             minReplicationCount = (followerToActor.size() + 1) / 2 + 1;
 102         } else {
 103             minReplicationCount = 0;
 104         }
 105
 106
 107         // Immediately schedule a heartbeat
 108         // Upon election: send initial empty AppendEntries RPCs
 109         // (heartbeat) to each server; repeat during idle periods to
 110         // prevent election timeouts (§5.2)
 111         scheduleHeartBeat(new FiniteDuration(0, TimeUnit.SECONDS));
 112
 113         scheduleInstallSnapshotCheck(
 114             new FiniteDuration(HEART_BEAT_INTERVAL.length() * 1000,
 115                 HEART_BEAT_INTERVAL.unit())
 116         );
 117
 118     }
 119
 120     @Override protected RaftState handleAppendEntries(ActorRef sender,
 121         AppendEntries appendEntries, RaftState suggestedState) {
 122
 123         context.getLogger()
 124             .error("An unexpected AppendEntries received in state " + state());
 125
 126         return suggestedState;
 127     }
 128
 129     @Override protected RaftState handleAppendEntriesReply(ActorRef sender,
 130         AppendEntriesReply appendEntriesReply, RaftState suggestedState) {
 131
 132         // Do not take any other action since a behavior change is coming
 133         if (suggestedState != state())
 134             return suggestedState;
 135
 136         // Update the FollowerLogInformation
 137         String followerId = appendEntriesReply.getFollowerId();
 138         FollowerLogInformation followerLogInformation =
 139             followerToLog.get(followerId);
 140         if (appendEntriesReply.isSuccess()) {
 141             followerLogInformation
 142                 .setMatchIndex(appendEntriesReply.getLogLastIndex());
 143             followerLogInformation
 144                 .setNextIndex(appendEntriesReply.getLogLastIndex() + 1);
 145         } else {
 146             followerLogInformation.decrNextIndex();
 147         }
 148
 149         // Now figure out if this reply warrants a change in the commitIndex
 150         // If there exists an N such that N > commitIndex, a majority
 151         // of matchIndex[i] ≥ N, and log[N].term == currentTerm:
 152         // set commitIndex = N (§5.3, §5.4).
 153         for (long N = context.getCommitIndex() + 1; ; N++) {
 154             int replicatedCount = 1;
 155
 156             for (FollowerLogInformation info : followerToLog.values()) {
 157                 if (info.getMatchIndex().get() >= N) {
 158                     replicatedCount++;
 159                 }
 160             }
 161
 162             if (replicatedCount >= minReplicationCount) {
 163                 ReplicatedLogEntry replicatedLogEntry =
 164                     context.getReplicatedLog().get(N);
 165                 if (replicatedLogEntry != null
 166                     && replicatedLogEntry.getTerm()
 167                     == currentTerm()) {
 168                     context.setCommitIndex(N);
 169                 }
 170             } else {
 171                 break;
 172             }
 173         }
 174
 175         // Apply the change to the state machine
 176         if (context.getCommitIndex() > context.getLastApplied()) {
 177             applyLogToStateMachine(context.getCommitIndex());
 178         }
 179
 180         return suggestedState;
 181     }
 182
 183     protected ClientRequestTracker findClientRequestTracker(long logIndex) {
 184         for (ClientRequestTracker tracker : trackerList) {
 185             if (tracker.getIndex() == logIndex) {
 186                 return tracker;
 187             }
 188         }
 189
 190         return null;
 191     }
 192
 193     @Override protected RaftState handleRequestVoteReply(ActorRef sender,
 194         RequestVoteReply requestVoteReply, RaftState suggestedState) {
 195         return suggestedState;
 196     }
 197
 198     @Override public RaftState state() {
 199         return RaftState.Leader;
 200     }
 201
 202     @Override public RaftState handleMessage(ActorRef sender, Object message) {
 203         Preconditions.checkNotNull(sender, "sender should not be null");
 204
 205         try {
 206             if (message instanceof SendHeartBeat) {
 207                 return sendHeartBeat();
 208             } else if(message instanceof SendInstallSnapshot) {
 209                 installSnapshotIfNeeded();
 210             } else if (message instanceof Replicate) {
 211                 replicate((Replicate) message);
 212             } else if (message instanceof InstallSnapshotReply){
 213                 // FIXME : Should I be checking the term here too?
 214                 handleInstallSnapshotReply(
 215                     (InstallSnapshotReply) message);
 216             }
 217         } finally {
 218             scheduleHeartBeat(HEART_BEAT_INTERVAL);
 219         }
 220
 221         return super.handleMessage(sender, message);
 222     }
 223
 224     private void handleInstallSnapshotReply(InstallSnapshotReply message) {
 225         InstallSnapshotReply reply = message;
 226         String followerId = reply.getFollowerId();
 227         FollowerLogInformation followerLogInformation =
 228             followerToLog.get(followerId);
 229
 230         followerLogInformation
 231             .setMatchIndex(context.getReplicatedLog().getSnapshotIndex());
 232         followerLogInformation
 233             .setNextIndex(context.getReplicatedLog().getSnapshotIndex() + 1);
 234     }
 235
 236     private void replicate(Replicate replicate) {
 237         long logIndex = replicate.getReplicatedLogEntry().getIndex();
 238
 239         context.getLogger().debug("Replicate message " + logIndex);
 240
 241         if (followerToActor.size() == 0) {
 242             context.setCommitIndex(
 243                 replicate.getReplicatedLogEntry().getIndex());
 244
 245             context.getActor()
 246                 .tell(new ApplyState(replicate.getClientActor(),
 247                         replicate.getIdentifier(),
 248                         replicate.getReplicatedLogEntry()),
 249                     context.getActor()
 250                 );
 251         } else {
 252
 253             // Create a tracker entry we will use this later to notify the
 254             // client actor
 255             trackerList.add(
 256                 new ClientRequestTrackerImpl(replicate.getClientActor(),
 257                     replicate.getIdentifier(),
 258                     logIndex)
 259             );
 260
 261             sendAppendEntries();
 262         }
 263     }
 264
 265     private void sendAppendEntries() {
 266         // Send an AppendEntries to all followers
 267         for (String followerId : followerToActor.keySet()) {
 268             ActorSelection followerActor =
 269                 followerToActor.get(followerId);
 270
 271             FollowerLogInformation followerLogInformation =
 272                 followerToLog.get(followerId);
 273
 274             long nextIndex = followerLogInformation.getNextIndex().get();
 275
 276             List<ReplicatedLogEntry> entries = Collections.emptyList();
 277
 278             if(context.getReplicatedLog().isPresent(nextIndex)){
 279                 entries =
 280                     context.getReplicatedLog().getFrom(nextIndex);
 281             }
 282
 283             followerActor.tell(
 284                 new AppendEntries(currentTerm(), context.getId(),
 285                     prevLogIndex(nextIndex), prevLogTerm(nextIndex),
 286                     entries, context.getCommitIndex()
 287                 ),
 288                 actor()
 289             );
 290         }
 291     }
 292
 293     private void installSnapshotIfNeeded(){
 294         for (String followerId : followerToActor.keySet()) {
 295             ActorSelection followerActor =
 296                 followerToActor.get(followerId);
 297
 298             FollowerLogInformation followerLogInformation =
 299                 followerToLog.get(followerId);
 300
 301             long nextIndex = followerLogInformation.getNextIndex().get();
 302
 303             if(!context.getReplicatedLog().isPresent(nextIndex) && context.getReplicatedLog().isInSnapshot(nextIndex)){
 304                 followerActor.tell(
 305                     new InstallSnapshot(currentTerm(), context.getId(),
 306                         context.getReplicatedLog().getSnapshotIndex(),
 307                         context.getReplicatedLog().getSnapshotTerm(),
 308                         context.getReplicatedLog().getSnapshot()
 309                     ),
 310                     actor()
 311                 );
 312             }
 313         }
 314     }
 315
 316     private RaftState sendHeartBeat() {
 317         if (followerToActor.size() > 0) {
 318             sendAppendEntries();
 319         }
 320         return state();
 321     }
 322
 323     private void stopHeartBeat() {
 324         if (heartbeatSchedule != null && !heartbeatSchedule.isCancelled()) {
 325             heartbeatSchedule.cancel();
 326         }
 327     }
 328
 329     private void stopInstallSnapshotSchedule() {
 330         if (installSnapshotSchedule != null && !installSnapshotSchedule.isCancelled()) {
 331             installSnapshotSchedule.cancel();
 332         }
 333     }
 334
 335     private void scheduleHeartBeat(FiniteDuration interval) {
 336         if(followerToActor.keySet().size() == 0){
 337             // Optimization - do not bother scheduling a heartbeat as there are
 338             // no followers
 339             return;
 340         }
 341
 342         stopHeartBeat();
 343
 344         // Schedule a heartbeat. When the scheduler triggers a SendHeartbeat
 345         // message is sent to itself.
 346         // Scheduling the heartbeat only once here because heartbeats do not
 347         // need to be sent if there are other messages being sent to the remote
 348         // actor.
 349         heartbeatSchedule =
 350             context.getActorSystem().scheduler().scheduleOnce(
 351                 interval,
 352                 context.getActor(), new SendHeartBeat(),
 353                 context.getActorSystem().dispatcher(), context.getActor());
 354     }
 355
 356
 357     private void scheduleInstallSnapshotCheck(FiniteDuration interval) {
 358         if(followerToActor.keySet().size() == 0){
 359             // Optimization - do not bother scheduling a heartbeat as there are
 360             // no followers
 361             return;
 362         }
 363
 364         stopInstallSnapshotSchedule();
 365
 366         // Schedule a message to send append entries to followers that can
 367         // accept an append entries with some data in it
 368         installSnapshotSchedule =
 369             context.getActorSystem().scheduler().scheduleOnce(
 370                 interval,
 371                 context.getActor(), new SendInstallSnapshot(),
 372                 context.getActorSystem().dispatcher(), context.getActor());
 373     }
 374
 375
 376
 377     @Override public void close() throws Exception {
 378         stopHeartBeat();
 379     }
 380
 381     @Override public String getLeaderId() {
 382         return context.getId();
 383     }
 384
 385 }