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.Optional;
25 import java.util.concurrent.TimeUnit;
26 import java.util.concurrent.atomic.AtomicBoolean;
27 import javax.annotation.Nullable;
28 import org.opendaylight.controller.cluster.raft.RaftActorContext;
29 import org.opendaylight.controller.cluster.raft.RaftState;
30 import org.opendaylight.controller.cluster.raft.ReplicatedLogEntry;
31 import org.opendaylight.controller.cluster.raft.base.messages.ApplySnapshot;
32 import org.opendaylight.controller.cluster.raft.base.messages.ElectionTimeout;
33 import org.opendaylight.controller.cluster.raft.base.messages.TimeoutNow;
34 import org.opendaylight.controller.cluster.raft.messages.AppendEntries;
35 import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
36 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshot;
37 import org.opendaylight.controller.cluster.raft.messages.InstallSnapshotReply;
38 import org.opendaylight.controller.cluster.raft.messages.RaftRPC;
39 import org.opendaylight.controller.cluster.raft.messages.RequestVote;
40 import org.opendaylight.controller.cluster.raft.messages.RequestVoteReply;
41 import org.opendaylight.controller.cluster.raft.persisted.ServerConfigurationPayload;
42 import org.opendaylight.controller.cluster.raft.persisted.Snapshot;
45 * The behavior of a RaftActor in the Follower raft state.
47 * <li> Respond to RPCs from candidates and leaders
48 * <li> If election timeout elapses without receiving AppendEntries
49 * RPC from current leader or granting vote to candidate:
50 * convert to candidate
53 public class Follower extends AbstractRaftActorBehavior {
54 private static final long MAX_ELECTION_TIMEOUT_FACTOR = 18;
56 private final SyncStatusTracker initialSyncStatusTracker;
58 private final Stopwatch lastLeaderMessageTimer = Stopwatch.createStarted();
59 private SnapshotTracker snapshotTracker = null;
60 private String leaderId;
61 private short leaderPayloadVersion;
63 public Follower(final RaftActorContext context) {
64 this(context, null, (short)-1);
67 public Follower(final RaftActorContext context, final String initialLeaderId,
68 final short initialLeaderPayloadVersion) {
69 super(context, RaftState.Follower);
70 this.leaderId = initialLeaderId;
71 this.leaderPayloadVersion = initialLeaderPayloadVersion;
73 initialSyncStatusTracker = new SyncStatusTracker(context.getActor(), getId(), context.getConfigParams()
74 .getSyncIndexThreshold());
76 if (context.getPeerIds().isEmpty() && getLeaderId() == null) {
77 actor().tell(TimeoutNow.INSTANCE, actor());
79 scheduleElection(electionDuration());
84 public final String getLeaderId() {
89 protected final void setLeaderId(@Nullable final String leaderId) {
90 this.leaderId = leaderId;
94 public short getLeaderPayloadVersion() {
95 return leaderPayloadVersion;
99 protected final void setLeaderPayloadVersion(final short leaderPayloadVersion) {
100 this.leaderPayloadVersion = leaderPayloadVersion;
103 private void restartLastLeaderMessageTimer() {
104 if (lastLeaderMessageTimer.isRunning()) {
105 lastLeaderMessageTimer.reset();
108 lastLeaderMessageTimer.start();
111 private boolean isLogEntryPresent(final long index) {
112 if (context.getReplicatedLog().isInSnapshot(index)) {
116 ReplicatedLogEntry entry = context.getReplicatedLog().get(index);
117 return entry != null;
121 private void updateInitialSyncStatus(final long currentLeaderCommit, final String newLeaderId) {
122 initialSyncStatusTracker.update(newLeaderId, currentLeaderCommit, context.getCommitIndex());
126 protected RaftActorBehavior handleAppendEntries(final ActorRef sender, final AppendEntries appendEntries) {
128 int numLogEntries = appendEntries.getEntries() != null ? appendEntries.getEntries().size() : 0;
129 if (log.isTraceEnabled()) {
130 log.trace("{}: handleAppendEntries: {}", logName(), appendEntries);
131 } else if (log.isDebugEnabled() && numLogEntries > 0) {
132 log.debug("{}: handleAppendEntries: {}", logName(), appendEntries);
135 // TODO : Refactor this method into a bunch of smaller methods
136 // to make it easier to read. Before refactoring ensure tests
137 // cover the code properly
139 if (snapshotTracker != null && !snapshotTracker.getLeaderId().equals(appendEntries.getLeaderId())) {
140 log.debug("{}: snapshot install is in progress but the prior snapshot leaderId {} does not match the "
141 + "AppendEntries leaderId {}", logName(), snapshotTracker.getLeaderId(), appendEntries.getLeaderId());
142 closeSnapshotTracker();
145 if (snapshotTracker != null || context.getSnapshotManager().isApplying()) {
146 // if snapshot install is in progress, follower should just acknowledge append entries with a reply.
147 AppendEntriesReply reply = new AppendEntriesReply(context.getId(), currentTerm(), true,
148 lastIndex(), lastTerm(), context.getPayloadVersion());
150 log.debug("{}: snapshot install is in progress, replying immediately with {}", logName(), reply);
151 sender.tell(reply, actor());
156 // If we got here then we do appear to be talking to the leader
157 leaderId = appendEntries.getLeaderId();
158 leaderPayloadVersion = appendEntries.getPayloadVersion();
160 // First check if the logs are in sync or not
161 long lastIndex = lastIndex();
163 if (isOutOfSync(appendEntries)) {
164 // We found that the log was out of sync so just send a negative
167 final AppendEntriesReply reply = new AppendEntriesReply(context.getId(), currentTerm(), false, lastIndex,
168 lastTerm(), context.getPayloadVersion());
170 log.info("{}: Follower is out-of-sync so sending negative reply: {}", logName(), reply);
171 updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
172 sender.tell(reply, actor());
176 if (appendEntries.getEntries() != null && appendEntries.getEntries().size() > 0) {
178 log.debug("{}: Number of entries to be appended = {}", logName(),
179 appendEntries.getEntries().size());
181 // 3. If an existing entry conflicts with a new one (same index
182 // but different terms), delete the existing entry and all that
184 int addEntriesFrom = 0;
185 if (context.getReplicatedLog().size() > 0) {
187 // Find the entry up until the one that is not in the follower's log
188 for (int i = 0;i < appendEntries.getEntries().size(); i++, addEntriesFrom++) {
189 ReplicatedLogEntry matchEntry = appendEntries.getEntries().get(i);
191 if (!isLogEntryPresent(matchEntry.getIndex())) {
192 // newEntry not found in the log
196 long existingEntryTerm = getLogEntryTerm(matchEntry.getIndex());
198 log.debug("{}: matchEntry {} is present: existingEntryTerm: {}", logName(), matchEntry,
201 // existingEntryTerm == -1 means it's in the snapshot and not in the log. We don't know
202 // what the term was so we'll assume it matches.
203 if (existingEntryTerm == -1 || existingEntryTerm == matchEntry.getTerm()) {
207 if (!context.getRaftPolicy().applyModificationToStateBeforeConsensus()) {
209 log.info("{}: Removing entries from log starting at {}", logName(), matchEntry.getIndex());
211 // Entries do not match so remove all subsequent entries
212 if (!context.getReplicatedLog().removeFromAndPersist(matchEntry.getIndex())) {
213 // Could not remove the entries - this means the matchEntry index must be in the
214 // snapshot and not the log. In this case the prior entries are part of the state
215 // so we must send back a reply to force a snapshot to completely re-sync the
216 // follower's log and state.
218 log.info("{}: Could not remove entries - sending reply to force snapshot", logName());
219 updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
220 sender.tell(new AppendEntriesReply(context.getId(), currentTerm(), false, lastIndex,
221 lastTerm(), context.getPayloadVersion(), true), actor());
227 updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
228 sender.tell(new AppendEntriesReply(context.getId(), currentTerm(), false, lastIndex,
229 lastTerm(), context.getPayloadVersion(), true), actor());
235 lastIndex = lastIndex();
236 log.debug("{}: After cleanup, lastIndex: {}, entries to be added from: {}", logName(),
237 lastIndex, addEntriesFrom);
239 // When persistence successfully completes for each new log entry appended, we need to determine if we
240 // should capture a snapshot to compact the persisted log. shouldCaptureSnapshot tracks whether or not
241 // one of the log entries has exceeded the log size threshold whereby a snapshot should be taken. However
242 // we don't initiate the snapshot at that log entry but rather after the last log entry has been persisted.
243 // This is done because subsequent log entries after the one that tripped the threshold may have been
244 // applied to the state already, as the persistence callback occurs async, and we want those entries
245 // purged from the persisted log as well.
246 final AtomicBoolean shouldCaptureSnapshot = new AtomicBoolean(false);
247 final Procedure<ReplicatedLogEntry> appendAndPersistCallback = logEntry -> {
248 final ReplicatedLogEntry lastEntryToAppend = appendEntries.getEntries().get(
249 appendEntries.getEntries().size() - 1);
250 if (shouldCaptureSnapshot.get() && logEntry == lastEntryToAppend) {
251 context.getSnapshotManager().capture(context.getReplicatedLog().last(), getReplicatedToAllIndex());
255 // 4. Append any new entries not already in the log
256 for (int i = addEntriesFrom; i < appendEntries.getEntries().size(); i++) {
257 ReplicatedLogEntry entry = appendEntries.getEntries().get(i);
259 log.debug("{}: Append entry to log {}", logName(), entry.getData());
261 context.getReplicatedLog().appendAndPersist(entry, appendAndPersistCallback, false);
263 shouldCaptureSnapshot.compareAndSet(false,
264 context.getReplicatedLog().shouldCaptureSnapshot(entry.getIndex()));
266 if (entry.getData() instanceof ServerConfigurationPayload) {
267 context.updatePeerIds((ServerConfigurationPayload)entry.getData());
271 log.debug("{}: Log size is now {}", logName(), context.getReplicatedLog().size());
274 // 5. If leaderCommit > commitIndex, set commitIndex =
275 // min(leaderCommit, index of last new entry)
277 lastIndex = lastIndex();
278 long prevCommitIndex = context.getCommitIndex();
280 if (appendEntries.getLeaderCommit() > prevCommitIndex) {
281 context.setCommitIndex(Math.min(appendEntries.getLeaderCommit(), lastIndex));
284 if (prevCommitIndex != context.getCommitIndex()) {
285 log.debug("{}: Commit index set to {}", logName(), context.getCommitIndex());
288 AppendEntriesReply reply = new AppendEntriesReply(context.getId(), currentTerm(), true,
289 lastIndex, lastTerm(), context.getPayloadVersion());
291 if (log.isTraceEnabled()) {
292 log.trace("{}: handleAppendEntries returning : {}", logName(), reply);
293 } else if (log.isDebugEnabled() && numLogEntries > 0) {
294 log.debug("{}: handleAppendEntries returning : {}", logName(), reply);
297 // Reply to the leader before applying any previous state so as not to hold up leader consensus.
298 updateInitialSyncStatus(appendEntries.getLeaderCommit(), appendEntries.getLeaderId());
299 sender.tell(reply, actor());
301 // If commitIndex > lastApplied: increment lastApplied, apply
302 // log[lastApplied] to state machine (§5.3)
303 // check if there are any entries to be applied. last-applied can be equal to last-index
304 if (appendEntries.getLeaderCommit() > context.getLastApplied()
305 && context.getLastApplied() < lastIndex) {
306 if (log.isDebugEnabled()) {
307 log.debug("{}: applyLogToStateMachine, appendEntries.getLeaderCommit(): {},"
308 + "context.getLastApplied(): {}, lastIndex(): {}", logName(),
309 appendEntries.getLeaderCommit(), context.getLastApplied(), lastIndex);
312 applyLogToStateMachine(appendEntries.getLeaderCommit());
315 if (!context.getSnapshotManager().isCapturing()) {
316 super.performSnapshotWithoutCapture(appendEntries.getReplicatedToAllIndex());
322 private boolean isOutOfSync(final AppendEntries appendEntries) {
324 long prevLogTerm = getLogEntryTerm(appendEntries.getPrevLogIndex());
325 boolean prevEntryPresent = isLogEntryPresent(appendEntries.getPrevLogIndex());
326 long lastIndex = lastIndex();
327 int numLogEntries = appendEntries.getEntries() != null ? appendEntries.getEntries().size() : 0;
328 boolean outOfSync = true;
330 if (lastIndex == -1 && appendEntries.getPrevLogIndex() != -1) {
332 // The follower's log is out of sync because the leader does have
333 // an entry at prevLogIndex and this follower has no entries in
336 log.info("{}: The followers log is empty and the senders prevLogIndex is {}",
337 logName(), appendEntries.getPrevLogIndex());
338 } else if (lastIndex > -1 && appendEntries.getPrevLogIndex() != -1 && !prevEntryPresent) {
340 // The follower's log is out of sync because the Leader's
341 // prevLogIndex entry was not found in it's log
343 log.info("{}: The log is not empty but the prevLogIndex {} was not found in it - "
344 + "lastIndex: {}, snapshotIndex: {}", logName(), appendEntries.getPrevLogIndex(), lastIndex,
345 context.getReplicatedLog().getSnapshotIndex());
346 } else if (lastIndex > -1 && prevEntryPresent && prevLogTerm != appendEntries.getPrevLogTerm()) {
348 // The follower's log is out of sync because the Leader's
349 // prevLogIndex entry does exist in the follower's log but it has
350 // a different term in it
352 log.info("{}: The prevLogIndex {} was found in the log but the term {} is not equal to the append entries"
353 + "prevLogTerm {} - lastIndex: {}, snapshotIndex: {}", logName(), appendEntries.getPrevLogIndex(),
354 prevLogTerm, appendEntries.getPrevLogTerm(), lastIndex,
355 context.getReplicatedLog().getSnapshotIndex());
356 } else if (appendEntries.getPrevLogIndex() == -1 && appendEntries.getPrevLogTerm() == -1
357 && appendEntries.getReplicatedToAllIndex() != -1
358 && !isLogEntryPresent(appendEntries.getReplicatedToAllIndex())) {
359 // This append entry comes from a leader who has it's log aggressively trimmed and so does not have
360 // the previous entry in it's in-memory journal
362 log.info("{}: Cannot append entries because the replicatedToAllIndex {} does not appear to be in the"
363 + " in-memory journal", logName(), appendEntries.getReplicatedToAllIndex());
364 } else if (appendEntries.getPrevLogIndex() == -1 && appendEntries.getPrevLogTerm() == -1
365 && appendEntries.getReplicatedToAllIndex() != -1 && numLogEntries > 0
366 && !isLogEntryPresent(appendEntries.getEntries().get(0).getIndex() - 1)) {
367 log.info("{}: Cannot append entries because the calculated previousIndex {} was not found in the "
368 + " in-memory journal", logName(), appendEntries.getEntries().get(0).getIndex() - 1);
376 protected RaftActorBehavior handleAppendEntriesReply(final ActorRef sender,
377 final AppendEntriesReply appendEntriesReply) {
382 protected RaftActorBehavior handleRequestVoteReply(final ActorRef sender,
383 final RequestVoteReply requestVoteReply) {
388 public RaftActorBehavior handleMessage(final ActorRef sender, final Object message) {
389 if (message instanceof ElectionTimeout || message instanceof TimeoutNow) {
390 return handleElectionTimeout(message);
393 if (!(message instanceof RaftRPC)) {
394 // The rest of the processing requires the message to be a RaftRPC
398 final RaftRPC rpc = (RaftRPC) message;
399 // If RPC request or response contains term T > currentTerm:
400 // set currentTerm = T, convert to follower (§5.1)
401 // This applies to all RPC messages and responses
402 if (rpc.getTerm() > context.getTermInformation().getCurrentTerm()) {
403 log.info("{}: Term {} in \"{}\" message is greater than follower's term {} - updating term",
404 logName(), rpc.getTerm(), rpc, context.getTermInformation().getCurrentTerm());
406 context.getTermInformation().updateAndPersist(rpc.getTerm(), null);
409 if (rpc instanceof InstallSnapshot) {
410 handleInstallSnapshot(sender, (InstallSnapshot) rpc);
411 restartLastLeaderMessageTimer();
412 scheduleElection(electionDuration());
416 if (!(rpc instanceof RequestVote) || canGrantVote((RequestVote) rpc)) {
417 restartLastLeaderMessageTimer();
418 scheduleElection(electionDuration());
421 return super.handleMessage(sender, rpc);
424 private RaftActorBehavior handleElectionTimeout(final Object message) {
425 // If the message is ElectionTimeout, verify we haven't actually seen a message from the leader
426 // during the election timeout interval. It may that the election timer expired b/c this actor
427 // was busy and messages got delayed, in which case leader messages would be backed up in the
428 // queue but would be processed before the ElectionTimeout message and thus would restart the
429 // lastLeaderMessageTimer.
430 long lastLeaderMessageInterval = lastLeaderMessageTimer.elapsed(TimeUnit.MILLISECONDS);
431 long electionTimeoutInMillis = context.getConfigParams().getElectionTimeOutInterval().toMillis();
432 boolean noLeaderMessageReceived = !lastLeaderMessageTimer.isRunning()
433 || lastLeaderMessageInterval >= electionTimeoutInMillis;
435 if (canStartElection()) {
436 if (message instanceof TimeoutNow) {
437 log.debug("{}: Received TimeoutNow - switching to Candidate", logName());
438 return internalSwitchBehavior(RaftState.Candidate);
439 } else if (noLeaderMessageReceived) {
440 // Check the cluster state to see if the leader is known to be up before we go to Candidate.
441 // However if we haven't heard from the leader in a long time even though the cluster state
442 // indicates it's up then something is wrong - leader might be stuck indefinitely - so switch
444 long maxElectionTimeout = electionTimeoutInMillis * MAX_ELECTION_TIMEOUT_FACTOR;
445 if (isLeaderAvailabilityKnown() && lastLeaderMessageInterval < maxElectionTimeout) {
446 log.debug("{}: Received ElectionTimeout but leader appears to be available", logName());
447 scheduleElection(electionDuration());
449 log.debug("{}: Received ElectionTimeout - switching to Candidate", logName());
450 return internalSwitchBehavior(RaftState.Candidate);
453 log.debug("{}: Received ElectionTimeout but lastLeaderMessageInterval {} < election timeout {}",
454 logName(), lastLeaderMessageInterval, context.getConfigParams().getElectionTimeOutInterval());
455 scheduleElection(electionDuration());
457 } else if (message instanceof ElectionTimeout) {
458 if (noLeaderMessageReceived) {
462 scheduleElection(electionDuration());
468 private boolean isLeaderAvailabilityKnown() {
469 if (leaderId == null) {
473 Optional<Cluster> cluster = context.getCluster();
474 if (!cluster.isPresent()) {
478 ActorSelection leaderActor = context.getPeerActorSelection(leaderId);
479 if (leaderActor == null) {
483 Address leaderAddress = leaderActor.anchorPath().address();
485 CurrentClusterState state = cluster.get().state();
486 Set<Member> unreachable = state.getUnreachable();
488 log.debug("{}: Checking for leader {} in the cluster unreachable set {}", logName(), leaderAddress,
491 for (Member m: unreachable) {
492 if (leaderAddress.equals(m.address())) {
493 log.info("{}: Leader {} is unreachable", logName(), leaderAddress);
498 for (Member m: state.getMembers()) {
499 if (leaderAddress.equals(m.address())) {
500 if (m.status() == MemberStatus.up() || m.status() == MemberStatus.weaklyUp()) {
501 log.debug("{}: Leader {} cluster status is {} - leader is available", logName(),
502 leaderAddress, m.status());
505 log.debug("{}: Leader {} cluster status is {} - leader is unavailable", logName(),
506 leaderAddress, m.status());
512 log.debug("{}: Leader {} not found in the cluster member set", logName(), leaderAddress);
517 private void handleInstallSnapshot(final ActorRef sender, final InstallSnapshot installSnapshot) {
519 log.debug("{}: handleInstallSnapshot: {}", logName(), installSnapshot);
521 leaderId = installSnapshot.getLeaderId();
523 if (snapshotTracker == null) {
524 snapshotTracker = new SnapshotTracker(log, installSnapshot.getTotalChunks(), installSnapshot.getLeaderId(),
528 updateInitialSyncStatus(installSnapshot.getLastIncludedIndex(), installSnapshot.getLeaderId());
531 final InstallSnapshotReply reply = new InstallSnapshotReply(
532 currentTerm(), context.getId(), installSnapshot.getChunkIndex(), true);
534 if (snapshotTracker.addChunk(installSnapshot.getChunkIndex(), installSnapshot.getData(),
535 installSnapshot.getLastChunkHashCode())) {
537 log.info("{}: Snapshot installed from leader: {}", logName(), installSnapshot.getLeaderId());
539 Snapshot snapshot = Snapshot.create(
540 context.getSnapshotManager().convertSnapshot(snapshotTracker.getSnapshotBytes()),
542 installSnapshot.getLastIncludedIndex(),
543 installSnapshot.getLastIncludedTerm(),
544 installSnapshot.getLastIncludedIndex(),
545 installSnapshot.getLastIncludedTerm(),
546 context.getTermInformation().getCurrentTerm(),
547 context.getTermInformation().getVotedFor(),
548 installSnapshot.getServerConfig().orNull());
550 ApplySnapshot.Callback applySnapshotCallback = new ApplySnapshot.Callback() {
552 public void onSuccess() {
553 log.debug("{}: handleInstallSnapshot returning: {}", logName(), reply);
555 sender.tell(reply, actor());
559 public void onFailure() {
560 sender.tell(new InstallSnapshotReply(currentTerm(), context.getId(), -1, false), actor());
564 actor().tell(new ApplySnapshot(snapshot, applySnapshotCallback), actor());
566 closeSnapshotTracker();
568 log.debug("{}: handleInstallSnapshot returning: {}", logName(), reply);
570 sender.tell(reply, actor());
572 } catch (IOException e) {
573 log.debug("{}: Exception in InstallSnapshot of follower", logName(), e);
575 sender.tell(new InstallSnapshotReply(currentTerm(), context.getId(),
576 -1, false), actor());
578 closeSnapshotTracker();
582 private void closeSnapshotTracker() {
583 if (snapshotTracker != null) {
584 snapshotTracker.close();
585 snapshotTracker = null;
590 public void close() {
591 closeSnapshotTracker();
596 SnapshotTracker getSnapshotTracker() {
597 return snapshotTracker;