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;
11 import akka.actor.ActorRef;
12 import akka.actor.ActorSelection;
13 import akka.event.Logging;
14 import akka.event.LoggingAdapter;
15 import akka.japi.Procedure;
16 import akka.persistence.RecoveryCompleted;
17 import akka.persistence.SaveSnapshotFailure;
18 import akka.persistence.SaveSnapshotSuccess;
19 import akka.persistence.SnapshotOffer;
20 import akka.persistence.SnapshotSelectionCriteria;
21 import akka.persistence.UntypedPersistentActor;
22 import org.opendaylight.controller.cluster.raft.behaviors.Candidate;
23 import org.opendaylight.controller.cluster.raft.behaviors.Follower;
24 import org.opendaylight.controller.cluster.raft.behaviors.Leader;
25 import org.opendaylight.controller.cluster.raft.behaviors.RaftActorBehavior;
26 import org.opendaylight.controller.cluster.raft.client.messages.AddRaftPeer;
27 import org.opendaylight.controller.cluster.raft.client.messages.FindLeader;
28 import org.opendaylight.controller.cluster.raft.client.messages.FindLeaderReply;
29 import org.opendaylight.controller.cluster.raft.internal.messages.ApplySnapshot;
30 import org.opendaylight.controller.cluster.raft.client.messages.RemoveRaftPeer;
31 import org.opendaylight.controller.cluster.raft.internal.messages.ApplyState;
32 import org.opendaylight.controller.cluster.raft.internal.messages.Replicate;
34 import java.io.Serializable;
35 import java.util.ArrayList;
36 import java.util.List;
40 * RaftActor encapsulates a state machine that needs to be kept synchronized
41 * in a cluster. It implements the RAFT algorithm as described in the paper
42 * <a href='https://ramcloud.stanford.edu/wiki/download/attachments/11370504/raft.pdf'>
43 * In Search of an Understandable Consensus Algorithm</a>
45 * RaftActor has 3 states and each state has a certain behavior associated
46 * with it. A Raft actor can behave as,
49 * <li> A Follower (or) </li>
50 * <li> A Candidate </li>
54 * A RaftActor MUST be a Leader in order to accept requests from clients to
55 * change the state of it's encapsulated state machine. Once a RaftActor becomes
56 * a Leader it is also responsible for ensuring that all followers ultimately
57 * have the same log and therefore the same state machine as itself.
60 * The current behavior of a RaftActor determines how election for leadership
61 * is initiated and how peer RaftActors react to request for votes.
64 * Each RaftActor also needs to know the current election term. It uses this
65 * information for a couple of things. One is to simply figure out who it
66 * voted for in the last election. Another is to figure out if the message
67 * it received to update it's state is stale.
70 * The RaftActor uses akka-persistence to store it's replicated log.
71 * Furthermore through it's behaviors a Raft Actor determines
74 * <li> when a log entry should be persisted </li>
75 * <li> when a log entry should be applied to the state machine (and) </li>
76 * <li> when a snapshot should be saved </li>
79 public abstract class RaftActor extends UntypedPersistentActor {
80 protected final LoggingAdapter LOG =
81 Logging.getLogger(getContext().system(), this);
84 * The current state determines the current behavior of a RaftActor
85 * A Raft Actor always starts off in the Follower State
87 private RaftActorBehavior currentBehavior;
90 * This context should NOT be passed directly to any other actor it is
91 * only to be consumed by the RaftActorBehaviors
93 private RaftActorContext context;
96 * The in-memory journal
98 private ReplicatedLogImpl replicatedLog = new ReplicatedLogImpl();
101 public RaftActor(String id, Map<String, String> peerAddresses) {
102 context = new RaftActorContextImpl(this.getSelf(),
104 id, new ElectionTermImpl(),
105 -1, -1, replicatedLog, peerAddresses, LOG);
108 @Override public void onReceiveRecover(Object message) {
109 if (message instanceof SnapshotOffer) {
110 SnapshotOffer offer = (SnapshotOffer) message;
111 Snapshot snapshot = (Snapshot) offer.snapshot();
113 // Create a replicated log with the snapshot information
114 // The replicated log can be used later on to retrieve this snapshot
115 // when we need to install it on a peer
116 replicatedLog = new ReplicatedLogImpl(snapshot);
118 // Apply the snapshot to the actors state
119 applySnapshot(snapshot.getState());
121 } else if (message instanceof ReplicatedLogEntry) {
122 replicatedLog.append((ReplicatedLogEntry) message);
123 } else if (message instanceof DeleteEntries) {
124 replicatedLog.removeFrom(((DeleteEntries) message).getFromIndex());
125 } else if (message instanceof UpdateElectionTerm) {
126 context.getTermInformation().update(((UpdateElectionTerm) message).getCurrentTerm(), ((UpdateElectionTerm) message).getVotedFor());
127 } else if (message instanceof RecoveryCompleted) {
129 "Last index in log : " + replicatedLog.lastIndex());
130 currentBehavior = switchBehavior(RaftState.Follower);
134 @Override public void onReceiveCommand(Object message) {
135 if (message instanceof ApplyState){
136 ApplyState applyState = (ApplyState) message;
138 LOG.debug("Applying state for log index {} data {}",
139 applyState.getReplicatedLogEntry().getIndex(),
140 applyState.getReplicatedLogEntry().getData());
142 applyState(applyState.getClientActor(), applyState.getIdentifier(),
143 applyState.getReplicatedLogEntry().getData());
144 } else if(message instanceof ApplySnapshot ) {
145 applySnapshot(((ApplySnapshot) message).getSnapshot());
146 } else if (message instanceof FindLeader) {
149 context.getPeerAddress(currentBehavior.getLeaderId())),
152 } else if (message instanceof SaveSnapshotSuccess) {
153 SaveSnapshotSuccess success = (SaveSnapshotSuccess) message;
155 // TODO: Not sure if we want to be this aggressive with trimming stuff
156 trimPersistentData(success.metadata().sequenceNr());
158 } else if (message instanceof SaveSnapshotFailure) {
160 // TODO: Handle failure in saving the snapshot
161 // Maybe do retries on failure
163 } else if (message instanceof AddRaftPeer){
165 // FIXME : Do not add raft peers like this.
166 // When adding a new Peer we have to ensure that the a majority of
167 // the peers know about the new Peer. Doing it this way may cause
168 // a situation where multiple Leaders may emerge
169 AddRaftPeer arp = (AddRaftPeer)message;
170 context.addToPeers(arp.getName(), arp.getAddress());
172 } else if (message instanceof RemoveRaftPeer){
174 RemoveRaftPeer rrp = (RemoveRaftPeer)message;
175 context.removePeer(rrp.getName());
180 currentBehavior.handleMessage(getSender(), message);
181 currentBehavior = switchBehavior(state);
188 * When a derived RaftActor needs to persist something it must call
195 protected void persistData(ActorRef clientActor, String identifier,
198 ReplicatedLogEntry replicatedLogEntry = new ReplicatedLogImplEntry(
199 context.getReplicatedLog().lastIndex() + 1,
200 context.getTermInformation().getCurrentTerm(), data);
202 LOG.debug("Persist data {}", replicatedLogEntry);
205 .appendAndPersist(clientActor, identifier, replicatedLogEntry);
208 protected String getId() {
209 return context.getId();
213 * Derived actors can call the isLeader method to check if the current
214 * RaftActor is the Leader or not
216 * @return true it this RaftActor is a Leader false otherwise
218 protected boolean isLeader() {
219 return context.getId().equals(currentBehavior.getLeaderId());
223 * Derived actor can call getLeader if they need a reference to the Leader.
224 * This would be useful for example in forwarding a request to an actor
225 * which is the leader
227 * @return A reference to the leader if known, null otherwise
229 protected ActorSelection getLeader(){
230 String leaderId = currentBehavior.getLeaderId();
231 if (leaderId == null) {
234 String peerAddress = context.getPeerAddress(leaderId);
235 LOG.debug("getLeader leaderId = " + leaderId + " peerAddress = "
237 return context.actorSelection(peerAddress);
240 protected RaftState getRaftState() {
241 return currentBehavior.state();
247 * The applyState method will be called by the RaftActor when some data
248 * needs to be applied to the actor's state
250 * @param clientActor A reference to the client who sent this message. This
251 * is the same reference that was passed to persistData
252 * by the derived actor. clientActor may be null when
253 * the RaftActor is behaving as a follower or during
255 * @param identifier The identifier of the persisted data. This is also
256 * the same identifier that was passed to persistData by
257 * the derived actor. identifier may be null when
258 * the RaftActor is behaving as a follower or during
260 * @param data A piece of data that was persisted by the persistData call.
261 * This should NEVER be null.
263 protected abstract void applyState(ActorRef clientActor, String identifier,
267 * This method will be called by the RaftActor when a snapshot needs to be
268 * created. The derived actor should respond with its current state.
270 * During recovery the state that is returned by the derived actor will
271 * be passed back to it by calling the applySnapshot method
273 * @return The current state of the actor
275 protected abstract Object createSnapshot();
278 * This method will be called by the RaftActor during recovery to
279 * reconstruct the state of the actor.
281 * This method may also be called at any other point during normal
282 * operations when the derived actor is out of sync with it's peers
283 * and the only way to bring it in sync is by applying a snapshot
285 * @param snapshot A snapshot of the state of the actor
287 protected abstract void applySnapshot(Object snapshot);
289 private RaftActorBehavior switchBehavior(RaftState state) {
290 if (currentBehavior != null) {
291 if (currentBehavior.state() == state) {
292 return currentBehavior;
294 LOG.info("Switching from state " + currentBehavior.state() + " to "
298 currentBehavior.close();
299 } catch (Exception e) {
301 "Failed to close behavior : " + currentBehavior.state());
305 LOG.info("Switching behavior to " + state);
307 RaftActorBehavior behavior = null;
308 if (state == RaftState.Candidate) {
309 behavior = new Candidate(context);
310 } else if (state == RaftState.Follower) {
311 behavior = new Follower(context);
313 behavior = new Leader(context);
318 private void trimPersistentData(long sequenceNumber) {
320 // FIXME : Not sure how exactly the SnapshotSelectionCriteria is applied
321 // For now guessing that it is ANDed.
322 deleteSnapshots(new SnapshotSelectionCriteria(
323 sequenceNumber - 100000, 43200000));
326 deleteMessages(sequenceNumber);
330 private class ReplicatedLogImpl implements ReplicatedLog {
331 private final List<ReplicatedLogEntry> journal;
332 private final Object snapshot;
333 private long snapshotIndex = -1;
334 private long snapshotTerm = -1;
336 public ReplicatedLogImpl(Snapshot snapshot) {
337 this.snapshot = snapshot.getState();
338 this.snapshotIndex = snapshot.getLastAppliedIndex();
339 this.snapshotTerm = snapshot.getLastAppliedTerm();
341 this.journal = new ArrayList<>(snapshot.getUnAppliedEntries());
344 public ReplicatedLogImpl() {
345 this.snapshot = null;
346 this.journal = new ArrayList<>();
349 @Override public ReplicatedLogEntry get(long index) {
350 int adjustedIndex = adjustedIndex(index);
352 if (adjustedIndex < 0 || adjustedIndex >= journal.size()) {
356 return journal.get(adjustedIndex);
359 @Override public ReplicatedLogEntry last() {
360 if (journal.size() == 0) {
363 return get(journal.size() - 1);
366 @Override public long lastIndex() {
367 if (journal.size() == 0) {
371 return last().getIndex();
374 @Override public long lastTerm() {
375 if (journal.size() == 0) {
379 return last().getTerm();
383 @Override public void removeFrom(long index) {
384 int adjustedIndex = adjustedIndex(index);
386 if (adjustedIndex < 0 || adjustedIndex >= journal.size()) {
390 journal.subList(adjustedIndex , journal.size()).clear();
394 @Override public void removeFromAndPersist(long index) {
395 int adjustedIndex = adjustedIndex(index);
397 if (adjustedIndex < 0 || adjustedIndex >= journal.size()) {
401 // FIXME: Maybe this should be done after the command is saved
402 journal.subList(adjustedIndex , journal.size()).clear();
404 persist(new DeleteEntries(adjustedIndex), new Procedure<DeleteEntries>(){
406 @Override public void apply(DeleteEntries param)
408 //FIXME : Doing nothing for now
415 @Override public void append(
416 final ReplicatedLogEntry replicatedLogEntry) {
417 journal.add(replicatedLogEntry);
420 @Override public List<ReplicatedLogEntry> getFrom(long index) {
421 int adjustedIndex = adjustedIndex(index);
423 List<ReplicatedLogEntry> entries = new ArrayList<>(100);
424 if (adjustedIndex < 0 || adjustedIndex >= journal.size()) {
429 for (int i = adjustedIndex;
430 i < journal.size(); i++) {
431 entries.add(journal.get(i));
436 @Override public void appendAndPersist(
437 final ReplicatedLogEntry replicatedLogEntry) {
438 appendAndPersist(null, null, replicatedLogEntry);
441 public void appendAndPersist(final ActorRef clientActor,
442 final String identifier,
443 final ReplicatedLogEntry replicatedLogEntry) {
444 context.getLogger().debug(
445 "Append log entry and persist " + replicatedLogEntry);
446 // FIXME : By adding the replicated log entry to the in-memory journal we are not truly ensuring durability of the logs
447 journal.add(replicatedLogEntry);
449 // When persisting events with persist it is guaranteed that the
450 // persistent actor will not receive further commands between the
451 // persist call and the execution(s) of the associated event
452 // handler. This also holds for multiple persist calls in context
453 // of a single command.
454 persist(replicatedLogEntry,
455 new Procedure<ReplicatedLogEntry>() {
456 public void apply(ReplicatedLogEntry evt) throws Exception {
457 // FIXME : Tentatively create a snapshot every hundred thousand entries. To be tuned.
458 if (size() > 100000) {
459 ReplicatedLogEntry lastAppliedEntry =
460 get(context.getLastApplied());
461 long lastAppliedIndex = -1;
462 long lastAppliedTerm = -1;
463 if (lastAppliedEntry != null) {
464 lastAppliedIndex = lastAppliedEntry.getIndex();
465 lastAppliedTerm = lastAppliedEntry.getTerm();
468 saveSnapshot(Snapshot.create(createSnapshot(),
469 getFrom(context.getLastApplied() + 1),
470 lastIndex(), lastTerm(), lastAppliedIndex,
473 // Send message for replication
474 if (clientActor != null) {
475 currentBehavior.handleMessage(getSelf(),
476 new Replicate(clientActor, identifier,
485 @Override public long size() {
486 return journal.size() + snapshotIndex + 1;
489 @Override public boolean isPresent(long index) {
490 int adjustedIndex = adjustedIndex(index);
492 if (adjustedIndex < 0 || adjustedIndex >= journal.size()) {
498 @Override public boolean isInSnapshot(long index) {
499 return index <= snapshotIndex;
502 @Override public Object getSnapshot() {
506 @Override public long getSnapshotIndex() {
507 return snapshotIndex;
510 @Override public long getSnapshotTerm() {
514 private int adjustedIndex(long index) {
515 if(snapshotIndex < 0){
518 return (int) (index - snapshotIndex);
523 private static class ReplicatedLogImplEntry implements ReplicatedLogEntry,
526 private final long index;
527 private final long term;
528 private final Object payload;
530 public ReplicatedLogImplEntry(long index, long term, Object payload) {
534 this.payload = payload;
537 @Override public Object getData() {
541 @Override public long getTerm() {
545 @Override public long getIndex() {
549 @Override public String toString() {
557 private static class DeleteEntries implements Serializable {
558 private final int fromIndex;
561 public DeleteEntries(int fromIndex) {
562 this.fromIndex = fromIndex;
565 public int getFromIndex() {
571 private static class Snapshot implements Serializable {
572 private final Object state;
573 private final List<ReplicatedLogEntry> unAppliedEntries;
574 private final long lastIndex;
575 private final long lastTerm;
576 private final long lastAppliedIndex;
577 private final long lastAppliedTerm;
579 private Snapshot(Object state,
580 List<ReplicatedLogEntry> unAppliedEntries, long lastIndex,
581 long lastTerm, long lastAppliedIndex, long lastAppliedTerm) {
583 this.unAppliedEntries = unAppliedEntries;
584 this.lastIndex = lastIndex;
585 this.lastTerm = lastTerm;
586 this.lastAppliedIndex = lastAppliedIndex;
587 this.lastAppliedTerm = lastAppliedTerm;
591 public static Snapshot create(Object state,
592 List<ReplicatedLogEntry> entries, long lastIndex, long lastTerm,
593 long lastAppliedIndex, long lastAppliedTerm) {
594 return new Snapshot(state, entries, lastIndex, lastTerm,
595 lastAppliedIndex, lastAppliedTerm);
598 public Object getState() {
602 public List<ReplicatedLogEntry> getUnAppliedEntries() {
603 return unAppliedEntries;
606 public long getLastTerm() {
610 public long getLastAppliedIndex() {
611 return lastAppliedIndex;
614 public long getLastAppliedTerm() {
615 return lastAppliedTerm;
619 private class ElectionTermImpl implements ElectionTerm {
621 * Identifier of the actor whose election term information this is
623 private long currentTerm = 0;
624 private String votedFor = null;
626 public long getCurrentTerm() {
630 public String getVotedFor() {
634 @Override public void update(long currentTerm, String votedFor) {
635 LOG.info("Set currentTerm={}, votedFor={}", currentTerm, votedFor);
637 this.currentTerm = currentTerm;
638 this.votedFor = votedFor;
642 public void updateAndPersist(long currentTerm, String votedFor){
643 update(currentTerm, votedFor);
644 // FIXME : Maybe first persist then update the state
645 persist(new UpdateElectionTerm(this.currentTerm, this.votedFor), new Procedure<UpdateElectionTerm>(){
647 @Override public void apply(UpdateElectionTerm param)
655 private static class UpdateElectionTerm implements Serializable {
656 private final long currentTerm;
657 private final String votedFor;
659 public UpdateElectionTerm(long currentTerm, String votedFor) {
660 this.currentTerm = currentTerm;
661 this.votedFor = votedFor;
664 public long getCurrentTerm() {
668 public String getVotedFor() {