import akka.actor.ActorRef;
import akka.actor.ActorSelection;
+import java.util.ArrayList;
+import java.util.Collection;
+import org.opendaylight.controller.cluster.raft.PeerInfo;
import org.opendaylight.controller.cluster.raft.RaftActorContext;
import org.opendaylight.controller.cluster.raft.RaftState;
+import org.opendaylight.controller.cluster.raft.ReplicatedLogEntry;
+import org.opendaylight.controller.cluster.raft.base.messages.ApplyState;
import org.opendaylight.controller.cluster.raft.base.messages.ElectionTimeout;
import org.opendaylight.controller.cluster.raft.messages.AppendEntries;
import org.opendaylight.controller.cluster.raft.messages.AppendEntriesReply;
import org.opendaylight.controller.cluster.raft.messages.RaftRPC;
import org.opendaylight.controller.cluster.raft.messages.RequestVote;
import org.opendaylight.controller.cluster.raft.messages.RequestVoteReply;
-
-import java.util.Set;
+import scala.concurrent.duration.FiniteDuration;
/**
- * The behavior of a RaftActor when it is in the CandidateState
- * <p/>
+ * The behavior of a RaftActor when it is in the Candidate raft state.
+ *
+ * <p>
* Candidates (§5.2):
* <ul>
* <li> On conversion to candidate, start election:
private final int votesRequired;
- private final Set<String> peers;
-
- public Candidate(RaftActorContext context) {
- super(context);
+ private final Collection<String> votingPeers = new ArrayList<>();
- peers = context.getPeerAddresses().keySet();
+ public Candidate(final RaftActorContext context) {
+ super(context, RaftState.Candidate);
- if(LOG.isDebugEnabled()) {
- LOG.debug("Election:Candidate has following peers: {}", peers);
+ for (PeerInfo peer: context.getPeers()) {
+ if (peer.isVoting()) {
+ votingPeers.add(peer.getId());
+ }
}
- if(peers.size() > 0) {
- // Votes are required from a majority of the peers including self.
- // The votesRequired field therefore stores a calculated value
- // of the number of votes required for this candidate to win an
- // election based on it's known peers.
- // If a peer was added during normal operation and raft replicas
- // came to know about them then the new peer would also need to be
- // taken into consideration when calculating this value.
- // Here are some examples for what the votesRequired would be for n
- // peers
- // 0 peers = 1 votesRequired (0 + 1) / 2 + 1 = 1
- // 2 peers = 2 votesRequired (2 + 1) / 2 + 1 = 2
- // 4 peers = 3 votesRequired (4 + 1) / 2 + 1 = 3
- int noOfPeers = peers.size();
- int self = 1;
- votesRequired = (noOfPeers + self) / 2 + 1;
+ log.debug("{}: Election: Candidate has following voting peers: {}", logName(), votingPeers);
+
+ votesRequired = getMajorityVoteCount(votingPeers.size());
+
+ startNewTerm();
+
+ if (votingPeers.isEmpty()) {
+ actor().tell(ElectionTimeout.INSTANCE, actor());
} else {
- votesRequired = 0;
+ scheduleElection(electionDuration());
}
+ }
- startNewTerm();
- scheduleElection(electionDuration());
+ @Override
+ public final String getLeaderId() {
+ return null;
+ }
+
+ @Override
+ public final short getLeaderPayloadVersion() {
+ return -1;
}
- @Override protected RaftActorBehavior handleAppendEntries(ActorRef sender,
- AppendEntries appendEntries) {
+ @Override
+ protected RaftActorBehavior handleAppendEntries(final ActorRef sender, final AppendEntries appendEntries) {
+
+ log.debug("{}: handleAppendEntries: {}", logName(), appendEntries);
- if(LOG.isDebugEnabled()) {
- LOG.debug(appendEntries.toString());
+ // Some other candidate for the same term became a leader and sent us an append entry
+ if (currentTerm() == appendEntries.getTerm()) {
+ log.info("{}: New Leader {} sent an AppendEntries to Candidate for term {} - will switch to Follower",
+ logName(), appendEntries.getLeaderId(), currentTerm());
+
+ return switchBehavior(new Follower(context));
}
return this;
}
- @Override protected RaftActorBehavior handleAppendEntriesReply(ActorRef sender,
- AppendEntriesReply appendEntriesReply) {
-
+ @Override
+ protected RaftActorBehavior handleAppendEntriesReply(final ActorRef sender,
+ final AppendEntriesReply appendEntriesReply) {
return this;
}
- @Override protected RaftActorBehavior handleRequestVoteReply(ActorRef sender,
- RequestVoteReply requestVoteReply) {
+ @Override
+ protected RaftActorBehavior handleRequestVoteReply(final ActorRef sender, final RequestVoteReply requestVoteReply) {
+ log.debug("{}: handleRequestVoteReply: {}, current voteCount: {}", logName(), requestVoteReply, voteCount);
if (requestVoteReply.isVoteGranted()) {
voteCount++;
}
if (voteCount >= votesRequired) {
- return switchBehavior(new Leader(context));
+ if (context.getLastApplied() < context.getReplicatedLog().lastIndex()) {
+ log.info("{}: LastApplied index {} is behind last index {} - switching to PreLeader",
+ logName(), context.getLastApplied(), context.getReplicatedLog().lastIndex());
+ return internalSwitchBehavior(RaftState.PreLeader);
+ } else {
+ return internalSwitchBehavior(RaftState.Leader);
+ }
}
return this;
}
- @Override public RaftState state() {
- return RaftState.Candidate;
+ @Override
+ protected FiniteDuration electionDuration() {
+ return super.electionDuration().$div(context.getConfigParams().getCandidateElectionTimeoutDivisor());
}
+
@Override
- public RaftActorBehavior handleMessage(ActorRef sender, Object originalMessage) {
+ final ApplyState getApplyStateFor(final ReplicatedLogEntry entry) {
+ throw new IllegalStateException("A candidate should never attempt to apply " + entry);
+ }
- Object message = fromSerializableMessage(originalMessage);
+ @Override
+ public RaftActorBehavior handleMessage(final ActorRef sender, final Object message) {
+ if (message instanceof ElectionTimeout) {
+ log.debug("{}: Received ElectionTimeout", logName());
+
+ if (votesRequired == 0) {
+ // If there are no peers then we should be a Leader
+ // We wait for the election timeout to occur before declare
+ // ourselves the leader. This gives enough time for a leader
+ // who we do not know about (as a peer)
+ // to send a message to the candidate
+
+ return internalSwitchBehavior(RaftState.Leader);
+ }
+
+ startNewTerm();
+ scheduleElection(electionDuration());
+ return this;
+ }
if (message instanceof RaftRPC) {
RaftRPC rpc = (RaftRPC) message;
- if(LOG.isDebugEnabled()) {
- LOG.debug("RaftRPC message received {} my term is {}", rpc, context.getTermInformation().getCurrentTerm());
- }
+ log.debug("{}: RaftRPC message received {}, my term is {}", logName(), rpc,
+ context.getTermInformation().getCurrentTerm());
// 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);
+ log.info("{}: Term {} in \"{}\" message is greater than Candidate's term {} - switching to Follower",
+ logName(), rpc.getTerm(), rpc, context.getTermInformation().getCurrentTerm());
- return switchBehavior(new Follower(context));
- }
- }
+ context.getTermInformation().updateAndPersist(rpc.getTerm(), null);
- if (message instanceof ElectionTimeout) {
- if (votesRequired == 0) {
- // If there are no peers then we should be a Leader
- // We wait for the election timeout to occur before declare
- // ourselves the leader. This gives enough time for a leader
- // who we do not know about (as a peer)
- // to send a message to the candidate
+ // The raft paper does not say whether or not a Candidate can/should process a RequestVote in
+ // this case but doing so gains quicker convergence when the sender's log is more up-to-date.
+ if (message instanceof RequestVote) {
+ super.handleMessage(sender, message);
+ }
- return switchBehavior(new Leader(context));
+ return internalSwitchBehavior(RaftState.Follower);
}
- startNewTerm();
- scheduleElection(electionDuration());
- return this;
}
return super.handleMessage(sender, message);
}
-
private void startNewTerm() {
-
-
// set voteCount back to 1 (that is voting for self)
voteCount = 1;
// Increment the election term and vote for self
long currentTerm = context.getTermInformation().getCurrentTerm();
- context.getTermInformation().updateAndPersist(currentTerm + 1,
- context.getId());
+ long newTerm = currentTerm + 1;
+ context.getTermInformation().updateAndPersist(newTerm, context.getId());
- if(LOG.isDebugEnabled()) {
- LOG.debug("Starting new term {}", (currentTerm + 1));
- }
+ log.info("{}: Starting new election term {}", logName(), newTerm);
// Request for a vote
// TODO: Retry request for vote if replies do not arrive in a reasonable
// amount of time TBD
- for (String peerId : peers) {
+ for (String peerId : votingPeers) {
ActorSelection peerActor = context.getPeerActorSelection(peerId);
- if(peerActor != null) {
- peerActor.tell(new RequestVote(
+ if (peerActor != null) {
+ RequestVote requestVote = new RequestVote(
context.getTermInformation().getCurrentTerm(),
context.getId(),
context.getReplicatedLog().lastIndex(),
- context.getReplicatedLog().lastTerm()),
- context.getActor()
- );
- }
- }
+ context.getReplicatedLog().lastTerm());
+ log.debug("{}: Sending {} to peer {}", logName(), requestVote, peerId);
+ peerActor.tell(requestVote, context.getActor());
+ }
+ }
}
- @Override public void close() throws Exception {
+ @Override
+ public void close() {
stopElection();
}
}
Code Review / controller.git / blobdiff