* terms of the Eclipse Public License v1.0 which accompanies this distribution,
* and is available at http://www.eclipse.org/legal/epl-v10.html
*/
-
package org.opendaylight.controller.cluster.raft.behaviors;
+import static java.util.Objects.requireNonNull;
+
import akka.actor.ActorRef;
import akka.actor.ActorSelection;
-import akka.actor.Cancellable;
-import com.google.common.base.Preconditions;
-import org.opendaylight.controller.cluster.raft.ClientRequestTracker;
-import org.opendaylight.controller.cluster.raft.ClientRequestTrackerImpl;
+import com.google.common.annotations.VisibleForTesting;
+import com.google.common.base.Optional;
+import com.google.common.base.Stopwatch;
+import java.util.concurrent.TimeUnit;
+import org.eclipse.jdt.annotation.NonNull;
+import org.eclipse.jdt.annotation.Nullable;
import org.opendaylight.controller.cluster.raft.FollowerLogInformation;
-import org.opendaylight.controller.cluster.raft.FollowerLogInformationImpl;
import org.opendaylight.controller.cluster.raft.RaftActorContext;
+import org.opendaylight.controller.cluster.raft.RaftActorLeadershipTransferCohort;
import org.opendaylight.controller.cluster.raft.RaftState;
-import org.opendaylight.controller.cluster.raft.ReplicatedLogEntry;
-import org.opendaylight.controller.cluster.raft.base.messages.Replicate;
-import org.opendaylight.controller.cluster.raft.base.messages.SendHeartBeat;
-import org.opendaylight.controller.cluster.raft.base.messages.SendInstallSnapshot;
-import org.opendaylight.controller.cluster.raft.messages.AppendEntries;
+import org.opendaylight.controller.cluster.raft.base.messages.TimeoutNow;
import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
-import org.opendaylight.controller.cluster.raft.messages.InstallSnapshot;
-import org.opendaylight.controller.cluster.raft.messages.InstallSnapshotReply;
-import org.opendaylight.controller.cluster.raft.messages.RaftRPC;
-import org.opendaylight.controller.cluster.raft.messages.RequestVoteReply;
-import scala.concurrent.duration.FiniteDuration;
-
-import java.util.ArrayList;
-import java.util.Collections;
-import java.util.HashMap;
-import java.util.List;
-import java.util.Map;
-import java.util.Set;
-import java.util.concurrent.TimeUnit;
-import java.util.concurrent.atomic.AtomicLong;
/**
- * The behavior of a RaftActor when it is in the Leader state
- * <p/>
+ * The behavior of a RaftActor when it is in the Leader state.
+ *
+ * <p>
* Leaders:
* <ul>
* <li> Upon election: send initial empty AppendEntries RPCs
* respond after entry applied to state machine (§5.3)
* <li> If last log index ≥ nextIndex for a follower: send
* AppendEntries RPC with log entries starting at nextIndex
- * <ul>
* <li> If successful: update nextIndex and matchIndex for
* follower (§5.3)
* <li> If AppendEntries fails because of log inconsistency:
* decrement nextIndex and retry (§5.3)
- * </ul>
- * <li> If there exists an N such that N > commitIndex, a majority
+ * <li> If there exists an N such that N > commitIndex, a majority
* of matchIndex[i] ≥ N, and log[N].term == currentTerm:
* set commitIndex = N (§5.3, §5.4).
+ * </ul>
*/
-public class Leader extends AbstractRaftActorBehavior {
-
-
- private final Map<String, FollowerLogInformation> followerToLog =
- new HashMap();
-
- private final Set<String> followers;
-
- private Cancellable heartbeatSchedule = null;
- private Cancellable appendEntriesSchedule = null;
- private Cancellable installSnapshotSchedule = null;
-
- private List<ClientRequestTracker> trackerList = new ArrayList<>();
-
- private final int minReplicationCount;
-
- public Leader(RaftActorContext context) {
- super(context);
-
- if (lastIndex() >= 0) {
- context.setCommitIndex(lastIndex());
- }
-
- followers = context.getPeerAddresses().keySet();
-
- for (String followerId : followers) {
- FollowerLogInformation followerLogInformation =
- new FollowerLogInformationImpl(followerId,
- new AtomicLong(lastIndex()),
- new AtomicLong(-1));
-
- followerToLog.put(followerId, followerLogInformation);
- }
-
- context.getLogger().debug("Election:Leader has following peers:"+ followers);
-
- if (followers.size() > 0) {
- minReplicationCount = (followers.size() + 1) / 2 + 1;
- } else {
- minReplicationCount = 0;
- }
-
-
- // Immediately schedule a heartbeat
- // Upon election: send initial empty AppendEntries RPCs
- // (heartbeat) to each server; repeat during idle periods to
- // prevent election timeouts (§5.2)
- scheduleHeartBeat(new FiniteDuration(0, TimeUnit.SECONDS));
+public class Leader extends AbstractLeader {
+ /**
+ * Internal message sent to periodically check if this leader has become isolated and should transition
+ * to {@link IsolatedLeader}.
+ */
+ @VisibleForTesting
+ static final Object ISOLATED_LEADER_CHECK = new Object();
- scheduleInstallSnapshotCheck(
- new FiniteDuration(context.getConfigParams().getHeartBeatInterval().length() * 1000,
- context.getConfigParams().getHeartBeatInterval().unit())
- );
+ private final Stopwatch isolatedLeaderCheck = Stopwatch.createStarted();
+ private @Nullable LeadershipTransferContext leadershipTransferContext;
+ Leader(RaftActorContext context, @Nullable AbstractLeader initializeFromLeader) {
+ super(context, RaftState.Leader, initializeFromLeader);
}
- @Override protected RaftState handleAppendEntries(ActorRef sender,
- AppendEntries appendEntries) {
-
- context.getLogger().debug(appendEntries.toString());
-
- return state();
+ public Leader(RaftActorContext context) {
+ this(context, null);
}
- @Override protected RaftState handleAppendEntriesReply(ActorRef sender,
- AppendEntriesReply appendEntriesReply) {
-
- if(! appendEntriesReply.isSuccess()) {
- context.getLogger()
- .debug(appendEntriesReply.toString());
- }
-
- // Update the FollowerLogInformation
- String followerId = appendEntriesReply.getFollowerId();
- FollowerLogInformation followerLogInformation =
- followerToLog.get(followerId);
-
- if(followerLogInformation == null){
- context.getLogger().error("Unknown follower {}", followerId);
- return state();
- }
-
- if (appendEntriesReply.isSuccess()) {
- followerLogInformation
- .setMatchIndex(appendEntriesReply.getLogLastIndex());
- followerLogInformation
- .setNextIndex(appendEntriesReply.getLogLastIndex() + 1);
- } else {
-
- // TODO: When we find that the follower is out of sync with the
- // Leader we simply decrement that followers next index by 1.
- // Would it be possible to do better than this? The RAFT spec
- // does not explicitly deal with it but may be something for us to
- // think about
-
- followerLogInformation.decrNextIndex();
- }
-
- // Now figure out if this reply warrants a change in the commitIndex
- // If there exists an N such that N > commitIndex, a majority
- // of matchIndex[i] ≥ N, and log[N].term == currentTerm:
- // set commitIndex = N (§5.3, §5.4).
- for (long N = context.getCommitIndex() + 1; ; N++) {
- int replicatedCount = 1;
-
- for (FollowerLogInformation info : followerToLog.values()) {
- if (info.getMatchIndex().get() >= N) {
- replicatedCount++;
- }
- }
+ @Override
+ public RaftActorBehavior handleMessage(ActorRef sender, Object originalMessage) {
+ requireNonNull(sender, "sender should not be null");
- if (replicatedCount >= minReplicationCount) {
- ReplicatedLogEntry replicatedLogEntry =
- context.getReplicatedLog().get(N);
- if (replicatedLogEntry != null
- && replicatedLogEntry.getTerm()
- == currentTerm()) {
- context.setCommitIndex(N);
- }
+ if (ISOLATED_LEADER_CHECK.equals(originalMessage)) {
+ if (isLeaderIsolated()) {
+ log.warn("{}: At least {} followers need to be active, Switching {} from Leader to IsolatedLeader",
+ context.getId(), getMinIsolatedLeaderPeerCount(), getLeaderId());
+ return internalSwitchBehavior(new IsolatedLeader(context, this));
} else {
- break;
+ return this;
}
+ } else {
+ return super.handleMessage(sender, originalMessage);
}
-
- // Apply the change to the state machine
- if (context.getCommitIndex() > context.getLastApplied()) {
- applyLogToStateMachine(context.getCommitIndex());
- }
-
- return state();
}
- protected ClientRequestTracker findClientRequestTracker(long logIndex) {
- for (ClientRequestTracker tracker : trackerList) {
- if (tracker.getIndex() == logIndex) {
- return tracker;
- }
+ @Override
+ protected void beforeSendHeartbeat() {
+ if (isolatedLeaderCheck.elapsed(TimeUnit.MILLISECONDS)
+ > context.getConfigParams().getIsolatedCheckIntervalInMillis()) {
+ context.getActor().tell(ISOLATED_LEADER_CHECK, context.getActor());
+ isolatedLeaderCheck.reset().start();
}
- return null;
+ if (leadershipTransferContext != null && leadershipTransferContext.isExpired(
+ context.getConfigParams().getElectionTimeOutInterval().toMillis())) {
+ log.debug("{}: Leadership transfer expired", logName());
+ leadershipTransferContext = null;
+ }
}
- @Override protected RaftState handleRequestVoteReply(ActorRef sender,
- RequestVoteReply requestVoteReply) {
- return state();
+ @Override
+ protected RaftActorBehavior handleAppendEntriesReply(ActorRef sender, AppendEntriesReply appendEntriesReply) {
+ RaftActorBehavior returnBehavior = super.handleAppendEntriesReply(sender, appendEntriesReply);
+ tryToCompleteLeadershipTransfer(appendEntriesReply.getFollowerId());
+ return returnBehavior;
}
- @Override public RaftState state() {
- return RaftState.Leader;
- }
+ /**
+ * Attempts to transfer leadership to a follower as per the raft paper (§3.10) as follows:
+ * <ul>
+ * <li>Start a timer (Stopwatch).</li>
+ * <li>Send an initial AppendEntries heartbeat to all followers.</li>
+ * <li>On AppendEntriesReply, check if the follower's new match Index matches the leader's last index</li>
+ * <li>If it matches,
+ * <ul>
+ * <li>Send an additional AppendEntries to ensure the follower has applied all its log entries to its state.</li>
+ * <li>Send an ElectionTimeout to the follower to immediately start an election.</li>
+ * <li>Notify {@link RaftActorLeadershipTransferCohort#transferComplete}.</li>
+ * </ul></li>
+ * <li>Otherwise if the election time out period elapses, notify
+ * {@link RaftActorLeadershipTransferCohort#abortTransfer}.</li>
+ * </ul>
+ *
+ * @param leadershipTransferCohort the cohort participating in the leadership transfer
+ */
+ public void transferLeadership(@NonNull RaftActorLeadershipTransferCohort leadershipTransferCohort) {
+ log.debug("{}: Attempting to transfer leadership", logName());
- @Override public RaftState handleMessage(ActorRef sender, Object originalMessage) {
- Preconditions.checkNotNull(sender, "sender should not be null");
+ leadershipTransferContext = new LeadershipTransferContext(leadershipTransferCohort);
- Object message = fromSerializableMessage(originalMessage);
+ // Send an immediate heart beat to the followers.
+ sendAppendEntries(0, false);
+ }
- if (message instanceof RaftRPC) {
- RaftRPC rpc = (RaftRPC) message;
- // If RPC request or response contains term T > currentTerm:
- // set currentTerm = T, convert to follower (§5.1)
- // This applies to all RPC messages and responses
- if (rpc.getTerm() > context.getTermInformation().getCurrentTerm()) {
- context.getTermInformation().updateAndPersist(rpc.getTerm(), null);
- return RaftState.Follower;
- }
+ private void tryToCompleteLeadershipTransfer(String followerId) {
+ if (leadershipTransferContext == null) {
+ return;
}
- try {
- if (message instanceof SendHeartBeat) {
- return sendHeartBeat();
- } else if(message instanceof SendInstallSnapshot) {
- installSnapshotIfNeeded();
- } else if (message instanceof Replicate) {
- replicate((Replicate) message);
- } else if (message instanceof InstallSnapshotReply){
- handleInstallSnapshotReply(
- (InstallSnapshotReply) message);
- }
- } finally {
- scheduleHeartBeat(context.getConfigParams().getHeartBeatInterval());
+ final Optional<String> requestedFollowerIdOptional
+ = leadershipTransferContext.transferCohort.getRequestedFollowerId();
+ if (requestedFollowerIdOptional.isPresent() && !requestedFollowerIdOptional.get().equals(followerId)) {
+ // we want to transfer leadership to specific follower
+ return;
}
- return super.handleMessage(sender, message);
- }
-
- private void handleInstallSnapshotReply(InstallSnapshotReply message) {
- InstallSnapshotReply reply = message;
- String followerId = reply.getFollowerId();
- FollowerLogInformation followerLogInformation =
- followerToLog.get(followerId);
+ FollowerLogInformation followerInfo = getFollower(followerId);
+ if (followerInfo == null) {
+ return;
+ }
- followerLogInformation
- .setMatchIndex(context.getReplicatedLog().getSnapshotIndex());
- followerLogInformation
- .setNextIndex(context.getReplicatedLog().getSnapshotIndex() + 1);
- }
+ long lastIndex = context.getReplicatedLog().lastIndex();
+ boolean isVoting = context.getPeerInfo(followerId).isVoting();
- private void replicate(Replicate replicate) {
- long logIndex = replicate.getReplicatedLogEntry().getIndex();
+ log.debug("{}: tryToCompleteLeadershipTransfer: followerId: {}, matchIndex: {}, lastIndex: {}, isVoting: {}",
+ logName(), followerId, followerInfo.getMatchIndex(), lastIndex, isVoting);
- context.getLogger().debug("Replicate message " + logIndex);
+ if (isVoting && followerInfo.getMatchIndex() == lastIndex) {
+ log.debug("{}: Follower's log matches - sending ElectionTimeout", logName());
- // Create a tracker entry we will use this later to notify the
- // client actor
- trackerList.add(
- new ClientRequestTrackerImpl(replicate.getClientActor(),
- replicate.getIdentifier(),
- logIndex)
- );
+ // We can't be sure if the follower has applied all its log entries to its state so send an
+ // additional AppendEntries with the latest commit index.
+ sendAppendEntries(0, false);
- if (followers.size() == 0) {
- context.setCommitIndex(logIndex);
- applyLogToStateMachine(logIndex);
- } else {
- sendAppendEntries();
- }
- }
+ // Now send a TimeoutNow message to the matching follower to immediately start an election.
+ ActorSelection followerActor = context.getPeerActorSelection(followerId);
+ followerActor.tell(TimeoutNow.INSTANCE, context.getActor());
- private void sendAppendEntries() {
- // Send an AppendEntries to all followers
- for (String followerId : followers) {
- ActorSelection followerActor =
- context.getPeerActorSelection(followerId);
-
- if (followerActor != null) {
- FollowerLogInformation followerLogInformation =
- followerToLog.get(followerId);
-
- long nextIndex = followerLogInformation.getNextIndex().get();
-
- List<ReplicatedLogEntry> entries = Collections.emptyList();
-
- if (context.getReplicatedLog().isPresent(nextIndex)) {
- // FIXME : Sending one entry at a time
- entries =
- context.getReplicatedLog().getFrom(nextIndex, 1);
- }
-
- followerActor.tell(
- new AppendEntries(currentTerm(), context.getId(),
- prevLogIndex(nextIndex),
- prevLogTerm(nextIndex), entries,
- context.getCommitIndex()).toSerializable(),
- actor()
- );
- }
- }
- }
+ log.debug("{}: Leader transfer complete", logName());
- /**
- * An installSnapshot is scheduled at a interval that is a multiple of
- * a HEARTBEAT_INTERVAL. This is to avoid the need to check for installing
- * snapshots at every heartbeat.
- */
- private void installSnapshotIfNeeded(){
- for (String followerId : followers) {
- ActorSelection followerActor =
- context.getPeerActorSelection(followerId);
-
- if(followerActor != null) {
- FollowerLogInformation followerLogInformation =
- followerToLog.get(followerId);
-
- long nextIndex = followerLogInformation.getNextIndex().get();
-
- if (!context.getReplicatedLog().isPresent(nextIndex) && context
- .getReplicatedLog().isInSnapshot(nextIndex)) {
- followerActor.tell(
- new InstallSnapshot(currentTerm(), context.getId(),
- context.getReplicatedLog().getSnapshotIndex(),
- context.getReplicatedLog().getSnapshotTerm(),
- context.getReplicatedLog().getSnapshot()
- ),
- actor()
- );
- }
- }
+ leadershipTransferContext.transferCohort.transferComplete();
+ leadershipTransferContext = null;
}
}
- private RaftState sendHeartBeat() {
- if (followers.size() > 0) {
- sendAppendEntries();
+ @Override
+ public void close() {
+ if (leadershipTransferContext != null) {
+ LeadershipTransferContext localLeadershipTransferContext = leadershipTransferContext;
+ leadershipTransferContext = null;
+ localLeadershipTransferContext.transferCohort.abortTransfer();
}
- return state();
- }
- private void stopHeartBeat() {
- if (heartbeatSchedule != null && !heartbeatSchedule.isCancelled()) {
- heartbeatSchedule.cancel();
- }
+ super.close();
}
- private void stopInstallSnapshotSchedule() {
- if (installSnapshotSchedule != null && !installSnapshotSchedule.isCancelled()) {
- installSnapshotSchedule.cancel();
- }
+ @VisibleForTesting
+ void markFollowerActive(String followerId) {
+ getFollower(followerId).markFollowerActive();
}
- private void scheduleHeartBeat(FiniteDuration interval) {
- if(followers.size() == 0){
- // Optimization - do not bother scheduling a heartbeat as there are
- // no followers
- return;
- }
-
- stopHeartBeat();
-
- // Schedule a heartbeat. When the scheduler triggers a SendHeartbeat
- // message is sent to itself.
- // Scheduling the heartbeat only once here because heartbeats do not
- // need to be sent if there are other messages being sent to the remote
- // actor.
- heartbeatSchedule =
- context.getActorSystem().scheduler().scheduleOnce(
- interval,
- context.getActor(), new SendHeartBeat(),
- context.getActorSystem().dispatcher(), context.getActor());
+ @VisibleForTesting
+ void markFollowerInActive(String followerId) {
+ getFollower(followerId).markFollowerInActive();
}
+ private static class LeadershipTransferContext {
+ RaftActorLeadershipTransferCohort transferCohort;
+ Stopwatch timer = Stopwatch.createStarted();
- private void scheduleInstallSnapshotCheck(FiniteDuration interval) {
- if(followers.size() == 0){
- // Optimization - do not bother scheduling a heartbeat as there are
- // no followers
- return;
+ LeadershipTransferContext(RaftActorLeadershipTransferCohort transferCohort) {
+ this.transferCohort = transferCohort;
}
- stopInstallSnapshotSchedule();
-
- // Schedule a message to send append entries to followers that can
- // accept an append entries with some data in it
- installSnapshotSchedule =
- context.getActorSystem().scheduler().scheduleOnce(
- interval,
- context.getActor(), new SendInstallSnapshot(),
- context.getActorSystem().dispatcher(), context.getActor());
- }
-
-
-
- @Override public void close() throws Exception {
- stopHeartBeat();
- }
+ boolean isExpired(long timeout) {
+ if (timer.elapsed(TimeUnit.MILLISECONDS) >= timeout) {
+ transferCohort.abortTransfer();
+ return true;
+ }
- @Override public String getLeaderId() {
- return context.getId();
+ return false;
+ }
}
-
}