2 * Copyright (c) 2014 Cisco Systems, Inc. and others. All rights reserved.
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
9 package org.opendaylight.controller.cluster.raft.behaviors;
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;
26 import java.util.concurrent.TimeUnit;
27 import java.util.concurrent.atomic.AtomicBoolean;
28 import javax.annotation.Nullable;
29 import org.opendaylight.controller.cluster.messaging.MessageAssembler;
30 import org.opendaylight.controller.cluster.raft.RaftActorContext;
31 import org.opendaylight.controller.cluster.raft.RaftState;
32 import org.opendaylight.controller.cluster.raft.ReplicatedLogEntry;
33 import org.opendaylight.controller.cluster.raft.base.messages.ApplySnapshot;
34 import org.opendaylight.controller.cluster.raft.base.messages.ElectionTimeout;
35 import org.opendaylight.controller.cluster.raft.base.messages.TimeoutNow;
36 import org.opendaylight.controller.cluster.raft.messages.AppendEntries;
37 import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
38 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshot;
39 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshotReply;
40 import org.opendaylight.controller.cluster.raft.messages.RaftRPC;
41 import org.opendaylight.controller.cluster.raft.messages.RequestVote;
42 import org.opendaylight.controller.cluster.raft.messages.RequestVoteReply;
43 import org.opendaylight.controller.cluster.raft.persisted.ServerConfigurationPayload;
44 import org.opendaylight.controller.cluster.raft.persisted.Snapshot;
47 * The behavior of a RaftActor in the Follower raft state.
49 * <li> Respond to RPCs from candidates and leaders
50 * <li> If election timeout elapses without receiving AppendEntries
51 * RPC from current leader or granting vote to candidate:
52 * convert to candidate
55 public class Follower extends AbstractRaftActorBehavior {
56 private static final long MAX_ELECTION_TIMEOUT_FACTOR = 18;
58 private final SyncStatusTracker initialSyncStatusTracker;
60 private final MessageAssembler appendEntriesMessageAssembler;
62 private final Stopwatch lastLeaderMessageTimer = Stopwatch.createStarted();
63 private SnapshotTracker snapshotTracker = null;
64 private String leaderId;
65 private short leaderPayloadVersion;
67 public Follower(final RaftActorContext context) {
68 this(context, null, (short)-1);
71 public Follower(final RaftActorContext context, final String initialLeaderId,
72 final short initialLeaderPayloadVersion) {
73 super(context, RaftState.Follower);
74 this.leaderId = initialLeaderId;
75 this.leaderPayloadVersion = initialLeaderPayloadVersion;
77 initialSyncStatusTracker = new SyncStatusTracker(context.getActor(), getId(), context.getConfigParams()
78 .getSyncIndexThreshold());
80 appendEntriesMessageAssembler = MessageAssembler.builder().logContext(logName())
81 .fileBackedStreamFactory(context.getFileBackedOutputStreamFactory())
82 .assembledMessageCallback((message, sender) -> handleMessage(sender, message)).build();
84 if (context.getPeerIds().isEmpty() && getLeaderId() == null) {
85 actor().tell(TimeoutNow.INSTANCE, actor());
87 scheduleElection(electionDuration());
92 public final String getLeaderId() {
97 protected final void setLeaderId(@Nullable final String leaderId) {
98 this.leaderId = leaderId;
102 public short getLeaderPayloadVersion() {
103 return leaderPayloadVersion;
107 protected final void setLeaderPayloadVersion(final short leaderPayloadVersion) {
108 this.leaderPayloadVersion = leaderPayloadVersion;
111 private void restartLastLeaderMessageTimer() {
112 if (lastLeaderMessageTimer.isRunning()) {
113 lastLeaderMessageTimer.reset();
116 lastLeaderMessageTimer.start();
119 private boolean isLogEntryPresent(final long index) {
120 if (context.getReplicatedLog().isInSnapshot(index)) {
124 ReplicatedLogEntry entry = context.getReplicatedLog().get(index);
125 return entry != null;
129 private void updateInitialSyncStatus(final long currentLeaderCommit, final String newLeaderId) {
130 initialSyncStatusTracker.update(newLeaderId, currentLeaderCommit, context.getCommitIndex());
134 protected RaftActorBehavior handleAppendEntries(final ActorRef sender, final AppendEntries appendEntries) {
136 int numLogEntries = appendEntries.getEntries().size();
137 if (log.isTraceEnabled()) {
138 log.trace("{}: handleAppendEntries: {}", logName(), appendEntries);
139 } else if (log.isDebugEnabled() && numLogEntries > 0) {
140 log.debug("{}: handleAppendEntries: {}", logName(), appendEntries);
143 // TODO : Refactor this method into a bunch of smaller methods
144 // to make it easier to read. Before refactoring ensure tests
145 // cover the code properly
147 if (snapshotTracker != null && !snapshotTracker.getLeaderId().equals(appendEntries.getLeaderId())) {
148 log.debug("{}: snapshot install is in progress but the prior snapshot leaderId {} does not match the "
149 + "AppendEntries leaderId {}", logName(), snapshotTracker.getLeaderId(), appendEntries.getLeaderId());
150 closeSnapshotTracker();
153 if (snapshotTracker != null || context.getSnapshotManager().isApplying()) {
154 // if snapshot install is in progress, follower should just acknowledge append entries with a reply.
155 AppendEntriesReply reply = new AppendEntriesReply(context.getId(), currentTerm(), true,
156 lastIndex(), lastTerm(), context.getPayloadVersion());
158 log.debug("{}: snapshot install is in progress, replying immediately with {}", logName(), reply);
159 sender.tell(reply, actor());
164 // If we got here then we do appear to be talking to the leader
165 leaderId = appendEntries.getLeaderId();
166 leaderPayloadVersion = appendEntries.getPayloadVersion();
168 // First check if the logs are in sync or not
169 long lastIndex = lastIndex();
171 if (isOutOfSync(appendEntries)) {
172 // We found that the log was out of sync so just send a negative
175 final AppendEntriesReply reply = new AppendEntriesReply(context.getId(), currentTerm(), false, lastIndex,
176 lastTerm(), context.getPayloadVersion());
178 log.info("{}: Follower is out-of-sync so sending negative reply: {}", logName(), reply);
179 updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
180 sender.tell(reply, actor());
184 if (numLogEntries > 0) {
185 log.debug("{}: Number of entries to be appended = {}", logName(), numLogEntries);
187 // 3. If an existing entry conflicts with a new one (same index
188 // but different terms), delete the existing entry and all that
190 int addEntriesFrom = 0;
191 if (context.getReplicatedLog().size() > 0) {
193 // Find the entry up until the one that is not in the follower's log
194 for (int i = 0;i < numLogEntries; i++, addEntriesFrom++) {
195 ReplicatedLogEntry matchEntry = appendEntries.getEntries().get(i);
197 if (!isLogEntryPresent(matchEntry.getIndex())) {
198 // newEntry not found in the log
202 long existingEntryTerm = getLogEntryTerm(matchEntry.getIndex());
204 log.debug("{}: matchEntry {} is present: existingEntryTerm: {}", logName(), matchEntry,
207 // existingEntryTerm == -1 means it's in the snapshot and not in the log. We don't know
208 // what the term was so we'll assume it matches.
209 if (existingEntryTerm == -1 || existingEntryTerm == matchEntry.getTerm()) {
213 if (!context.getRaftPolicy().applyModificationToStateBeforeConsensus()) {
215 log.info("{}: Removing entries from log starting at {}", logName(), matchEntry.getIndex());
217 // Entries do not match so remove all subsequent entries
218 if (!context.getReplicatedLog().removeFromAndPersist(matchEntry.getIndex())) {
219 // Could not remove the entries - this means the matchEntry index must be in the
220 // snapshot and not the log. In this case the prior entries are part of the state
221 // so we must send back a reply to force a snapshot to completely re-sync the
222 // follower's log and state.
224 log.info("{}: Could not remove entries - sending reply to force snapshot", logName());
225 updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
226 sender.tell(new AppendEntriesReply(context.getId(), currentTerm(), false, lastIndex,
227 lastTerm(), context.getPayloadVersion(), true), actor());
233 updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
234 sender.tell(new AppendEntriesReply(context.getId(), currentTerm(), false, lastIndex,
235 lastTerm(), context.getPayloadVersion(), true), actor());
241 lastIndex = lastIndex();
242 log.debug("{}: After cleanup, lastIndex: {}, entries to be added from: {}", logName(),
243 lastIndex, addEntriesFrom);
245 // When persistence successfully completes for each new log entry appended, we need to determine if we
246 // should capture a snapshot to compact the persisted log. shouldCaptureSnapshot tracks whether or not
247 // one of the log entries has exceeded the log size threshold whereby a snapshot should be taken. However
248 // we don't initiate the snapshot at that log entry but rather after the last log entry has been persisted.
249 // This is done because subsequent log entries after the one that tripped the threshold may have been
250 // applied to the state already, as the persistence callback occurs async, and we want those entries
251 // purged from the persisted log as well.
252 final AtomicBoolean shouldCaptureSnapshot = new AtomicBoolean(false);
253 final Procedure<ReplicatedLogEntry> appendAndPersistCallback = logEntry -> {
254 final List<ReplicatedLogEntry> entries = appendEntries.getEntries();
255 final ReplicatedLogEntry lastEntryToAppend = entries.get(entries.size() - 1);
256 if (shouldCaptureSnapshot.get() && logEntry == lastEntryToAppend) {
257 context.getSnapshotManager().capture(context.getReplicatedLog().last(), getReplicatedToAllIndex());
261 // 4. Append any new entries not already in the log
262 for (int i = addEntriesFrom; i < numLogEntries; i++) {
263 ReplicatedLogEntry entry = appendEntries.getEntries().get(i);
265 log.debug("{}: Append entry to log {}", logName(), entry.getData());
267 context.getReplicatedLog().appendAndPersist(entry, appendAndPersistCallback, false);
269 shouldCaptureSnapshot.compareAndSet(false,
270 context.getReplicatedLog().shouldCaptureSnapshot(entry.getIndex()));
272 if (entry.getData() instanceof ServerConfigurationPayload) {
273 context.updatePeerIds((ServerConfigurationPayload)entry.getData());
277 log.debug("{}: Log size is now {}", logName(), context.getReplicatedLog().size());
280 // 5. If leaderCommit > commitIndex, set commitIndex =
281 // min(leaderCommit, index of last new entry)
283 lastIndex = lastIndex();
284 long prevCommitIndex = context.getCommitIndex();
286 if (appendEntries.getLeaderCommit() > prevCommitIndex) {
287 context.setCommitIndex(Math.min(appendEntries.getLeaderCommit(), lastIndex));
290 if (prevCommitIndex != context.getCommitIndex()) {
291 log.debug("{}: Commit index set to {}", logName(), context.getCommitIndex());
294 AppendEntriesReply reply = new AppendEntriesReply(context.getId(), currentTerm(), true,
295 lastIndex, lastTerm(), context.getPayloadVersion());
297 if (log.isTraceEnabled()) {
298 log.trace("{}: handleAppendEntries returning : {}", logName(), reply);
299 } else if (log.isDebugEnabled() && numLogEntries > 0) {
300 log.debug("{}: handleAppendEntries returning : {}", logName(), reply);
303 // Reply to the leader before applying any previous state so as not to hold up leader consensus.
304 updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
305 sender.tell(reply, actor());
307 // If commitIndex > lastApplied: increment lastApplied, apply
308 // log[lastApplied] to state machine (§5.3)
309 // check if there are any entries to be applied. last-applied can be equal to last-index
310 if (appendEntries.getLeaderCommit() > context.getLastApplied()
311 && context.getLastApplied() < lastIndex) {
312 if (log.isDebugEnabled()) {
313 log.debug("{}: applyLogToStateMachine, appendEntries.getLeaderCommit(): {}, "
314 + "context.getLastApplied(): {}, lastIndex(): {}", logName(),
315 appendEntries.getLeaderCommit(), context.getLastApplied(), lastIndex);
318 applyLogToStateMachine(appendEntries.getLeaderCommit());
321 if (!context.getSnapshotManager().isCapturing()) {
322 super.performSnapshotWithoutCapture(appendEntries.getReplicatedToAllIndex());
325 appendEntriesMessageAssembler.checkExpiredAssembledMessageState();
330 private boolean isOutOfSync(final AppendEntries appendEntries) {
332 final long lastIndex = lastIndex();
333 if (lastIndex == -1 && appendEntries.getPrevLogIndex() != -1) {
335 // The follower's log is out of sync because the leader does have an entry at prevLogIndex and this
336 // follower has no entries in it's log.
338 log.info("{}: The followers log is empty and the senders prevLogIndex is {}", logName(),
339 appendEntries.getPrevLogIndex());
343 if (lastIndex > -1) {
344 if (isLogEntryPresent(appendEntries.getPrevLogIndex())) {
345 final long prevLogTerm = getLogEntryTerm(appendEntries.getPrevLogIndex());
346 if (prevLogTerm != appendEntries.getPrevLogTerm()) {
348 // The follower's log is out of sync because the Leader's prevLogIndex entry does exist
349 // in the follower's log but it has a different term in it
351 log.info("{}: The prevLogIndex {} was found in the log but the term {} is not equal to the append "
352 + "entries prevLogTerm {} - lastIndex: {}, snapshotIndex: {}", logName(),
353 appendEntries.getPrevLogIndex(), prevLogTerm, appendEntries.getPrevLogTerm(), lastIndex,
354 context.getReplicatedLog().getSnapshotIndex());
357 } else if (appendEntries.getPrevLogIndex() != -1) {
359 // The follower's log is out of sync because the Leader's prevLogIndex entry was not found in it's log
361 log.info("{}: The log is not empty but the prevLogIndex {} was not found in it - lastIndex: {}, "
362 + "snapshotIndex: {}", logName(), appendEntries.getPrevLogIndex(), lastIndex,
363 context.getReplicatedLog().getSnapshotIndex());
368 if (appendEntries.getPrevLogIndex() == -1 && appendEntries.getPrevLogTerm() == -1
369 && appendEntries.getReplicatedToAllIndex() != -1) {
370 if (!isLogEntryPresent(appendEntries.getReplicatedToAllIndex())) {
371 // This append entry comes from a leader who has it's log aggressively trimmed and so does not have
372 // the previous entry in it's in-memory journal
374 log.info("{}: Cannot append entries because the replicatedToAllIndex {} does not appear to be in the "
375 + "in-memory journal", logName(), appendEntries.getReplicatedToAllIndex());
379 final List<ReplicatedLogEntry> entries = appendEntries.getEntries();
380 if (entries.size() > 0 && !isLogEntryPresent(entries.get(0).getIndex() - 1)) {
381 log.info("{}: Cannot append entries because the calculated previousIndex {} was not found in the "
382 + "in-memory journal", logName(), entries.get(0).getIndex() - 1);
391 protected RaftActorBehavior handleAppendEntriesReply(final ActorRef sender,
392 final AppendEntriesReply appendEntriesReply) {
397 protected RaftActorBehavior handleRequestVoteReply(final ActorRef sender,
398 final RequestVoteReply requestVoteReply) {
403 public RaftActorBehavior handleMessage(final ActorRef sender, final Object message) {
404 if (message instanceof ElectionTimeout || message instanceof TimeoutNow) {
405 return handleElectionTimeout(message);
408 if (appendEntriesMessageAssembler.handleMessage(message, actor())) {
412 if (!(message instanceof RaftRPC)) {
413 // The rest of the processing requires the message to be a RaftRPC
417 final RaftRPC rpc = (RaftRPC) message;
418 // If RPC request or response contains term T > currentTerm:
419 // set currentTerm = T, convert to follower (§5.1)
420 // This applies to all RPC messages and responses
421 if (rpc.getTerm() > context.getTermInformation().getCurrentTerm()) {
422 log.info("{}: Term {} in \"{}\" message is greater than follower's term {} - updating term",
423 logName(), rpc.getTerm(), rpc, context.getTermInformation().getCurrentTerm());
425 context.getTermInformation().updateAndPersist(rpc.getTerm(), null);
428 if (rpc instanceof InstallSnapshot) {
429 handleInstallSnapshot(sender, (InstallSnapshot) rpc);
430 restartLastLeaderMessageTimer();
431 scheduleElection(electionDuration());
435 if (!(rpc instanceof RequestVote) || canGrantVote((RequestVote) rpc)) {
436 restartLastLeaderMessageTimer();
437 scheduleElection(electionDuration());
440 return super.handleMessage(sender, rpc);
443 private RaftActorBehavior handleElectionTimeout(final Object message) {
444 // If the message is ElectionTimeout, verify we haven't actually seen a message from the leader
445 // during the election timeout interval. It may that the election timer expired b/c this actor
446 // was busy and messages got delayed, in which case leader messages would be backed up in the
447 // queue but would be processed before the ElectionTimeout message and thus would restart the
448 // lastLeaderMessageTimer.
449 long lastLeaderMessageInterval = lastLeaderMessageTimer.elapsed(TimeUnit.MILLISECONDS);
450 long electionTimeoutInMillis = context.getConfigParams().getElectionTimeOutInterval().toMillis();
451 boolean noLeaderMessageReceived = !lastLeaderMessageTimer.isRunning()
452 || lastLeaderMessageInterval >= electionTimeoutInMillis;
454 if (canStartElection()) {
455 if (message instanceof TimeoutNow) {
456 log.debug("{}: Received TimeoutNow - switching to Candidate", logName());
457 return internalSwitchBehavior(RaftState.Candidate);
458 } else if (noLeaderMessageReceived) {
459 // Check the cluster state to see if the leader is known to be up before we go to Candidate.
460 // However if we haven't heard from the leader in a long time even though the cluster state
461 // indicates it's up then something is wrong - leader might be stuck indefinitely - so switch
463 long maxElectionTimeout = electionTimeoutInMillis * MAX_ELECTION_TIMEOUT_FACTOR;
464 if (isLeaderAvailabilityKnown() && lastLeaderMessageInterval < maxElectionTimeout) {
465 log.debug("{}: Received ElectionTimeout but leader appears to be available", logName());
466 scheduleElection(electionDuration());
468 log.debug("{}: Received ElectionTimeout - switching to Candidate", logName());
469 return internalSwitchBehavior(RaftState.Candidate);
472 log.debug("{}: Received ElectionTimeout but lastLeaderMessageInterval {} < election timeout {}",
473 logName(), lastLeaderMessageInterval, context.getConfigParams().getElectionTimeOutInterval());
474 scheduleElection(electionDuration());
476 } else if (message instanceof ElectionTimeout) {
477 if (noLeaderMessageReceived) {
481 scheduleElection(electionDuration());
487 private boolean isLeaderAvailabilityKnown() {
488 if (leaderId == null) {
492 Optional<Cluster> cluster = context.getCluster();
493 if (!cluster.isPresent()) {
497 ActorSelection leaderActor = context.getPeerActorSelection(leaderId);
498 if (leaderActor == null) {
502 Address leaderAddress = leaderActor.anchorPath().address();
504 CurrentClusterState state = cluster.get().state();
505 Set<Member> unreachable = state.getUnreachable();
507 log.debug("{}: Checking for leader {} in the cluster unreachable set {}", logName(), leaderAddress,
510 for (Member m: unreachable) {
511 if (leaderAddress.equals(m.address())) {
512 log.info("{}: Leader {} is unreachable", logName(), leaderAddress);
517 for (Member m: state.getMembers()) {
518 if (leaderAddress.equals(m.address())) {
519 if (m.status() == MemberStatus.up() || m.status() == MemberStatus.weaklyUp()) {
520 log.debug("{}: Leader {} cluster status is {} - leader is available", logName(),
521 leaderAddress, m.status());
524 log.debug("{}: Leader {} cluster status is {} - leader is unavailable", logName(),
525 leaderAddress, m.status());
531 log.debug("{}: Leader {} not found in the cluster member set", logName(), leaderAddress);
536 private void handleInstallSnapshot(final ActorRef sender, final InstallSnapshot installSnapshot) {
538 log.debug("{}: handleInstallSnapshot: {}", logName(), installSnapshot);
540 leaderId = installSnapshot.getLeaderId();
542 if (snapshotTracker == null) {
543 snapshotTracker = new SnapshotTracker(log, installSnapshot.getTotalChunks(), installSnapshot.getLeaderId(),
547 updateInitialSyncStatus(installSnapshot.getLastIncludedIndex(), installSnapshot.getLeaderId());
550 final InstallSnapshotReply reply = new InstallSnapshotReply(
551 currentTerm(), context.getId(), installSnapshot.getChunkIndex(), true);
553 if (snapshotTracker.addChunk(installSnapshot.getChunkIndex(), installSnapshot.getData(),
554 installSnapshot.getLastChunkHashCode())) {
556 log.info("{}: Snapshot installed from leader: {}", logName(), installSnapshot.getLeaderId());
558 Snapshot snapshot = Snapshot.create(
559 context.getSnapshotManager().convertSnapshot(snapshotTracker.getSnapshotBytes()),
561 installSnapshot.getLastIncludedIndex(),
562 installSnapshot.getLastIncludedTerm(),
563 installSnapshot.getLastIncludedIndex(),
564 installSnapshot.getLastIncludedTerm(),
565 context.getTermInformation().getCurrentTerm(),
566 context.getTermInformation().getVotedFor(),
567 installSnapshot.getServerConfig().orNull());
569 ApplySnapshot.Callback applySnapshotCallback = new ApplySnapshot.Callback() {
571 public void onSuccess() {
572 log.debug("{}: handleInstallSnapshot returning: {}", logName(), reply);
574 sender.tell(reply, actor());
578 public void onFailure() {
579 sender.tell(new InstallSnapshotReply(currentTerm(), context.getId(), -1, false), actor());
583 actor().tell(new ApplySnapshot(snapshot, applySnapshotCallback), actor());
585 closeSnapshotTracker();
587 log.debug("{}: handleInstallSnapshot returning: {}", logName(), reply);
589 sender.tell(reply, actor());
591 } catch (IOException e) {
592 log.debug("{}: Exception in InstallSnapshot of follower", logName(), e);
594 sender.tell(new InstallSnapshotReply(currentTerm(), context.getId(),
595 -1, false), actor());
597 closeSnapshotTracker();
601 private void closeSnapshotTracker() {
602 if (snapshotTracker != null) {
603 snapshotTracker.close();
604 snapshotTracker = null;
609 public void close() {
610 closeSnapshotTracker();
612 appendEntriesMessageAssembler.close();
616 SnapshotTracker getSnapshotTracker() {
617 return snapshotTracker;