Package org.elasticsearch.discovery.zen

Source Code of org.elasticsearch.discovery.zen.ZenDiscovery$ProcessClusterState

/*
* Licensed to Elasticsearch under one or more contributor
* license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright
* ownership. Elasticsearch licenses this file to you under
* the Apache License, Version 2.0 (the "License"); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
*    http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing,
* software distributed under the License is distributed on an
* "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
* KIND, either express or implied.  See the License for the
* specific language governing permissions and limitations
* under the License.
*/

package org.elasticsearch.discovery.zen;

import com.google.common.base.Objects;
import com.google.common.collect.Lists;
import com.google.common.collect.Sets;
import org.elasticsearch.ElasticsearchException;
import org.elasticsearch.ElasticsearchIllegalArgumentException;
import org.elasticsearch.ElasticsearchIllegalStateException;
import org.elasticsearch.ExceptionsHelper;
import org.elasticsearch.cluster.*;
import org.elasticsearch.cluster.block.ClusterBlocks;
import org.elasticsearch.cluster.metadata.IndexMetaData;
import org.elasticsearch.cluster.metadata.MetaData;
import org.elasticsearch.cluster.node.DiscoveryNode;
import org.elasticsearch.cluster.node.DiscoveryNodeService;
import org.elasticsearch.cluster.node.DiscoveryNodes;
import org.elasticsearch.cluster.routing.allocation.AllocationService;
import org.elasticsearch.cluster.routing.allocation.RoutingAllocation;
import org.elasticsearch.cluster.service.InternalClusterService;
import org.elasticsearch.cluster.settings.ClusterDynamicSettings;
import org.elasticsearch.cluster.settings.DynamicSettings;
import org.elasticsearch.cluster.settings.Validator;
import org.elasticsearch.common.Priority;
import org.elasticsearch.common.collect.Tuple;
import org.elasticsearch.common.component.AbstractLifecycleComponent;
import org.elasticsearch.common.component.Lifecycle;
import org.elasticsearch.common.inject.Inject;
import org.elasticsearch.common.inject.internal.Nullable;
import org.elasticsearch.common.io.stream.StreamInput;
import org.elasticsearch.common.io.stream.StreamOutput;
import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.common.unit.TimeValue;
import org.elasticsearch.common.util.concurrent.ConcurrentCollections;
import org.elasticsearch.discovery.Discovery;
import org.elasticsearch.discovery.DiscoverySettings;
import org.elasticsearch.discovery.InitialStateDiscoveryListener;
import org.elasticsearch.discovery.zen.elect.ElectMasterService;
import org.elasticsearch.discovery.zen.fd.MasterFaultDetection;
import org.elasticsearch.discovery.zen.fd.NodesFaultDetection;
import org.elasticsearch.discovery.zen.membership.MembershipAction;
import org.elasticsearch.discovery.zen.ping.PingContextProvider;
import org.elasticsearch.discovery.zen.ping.ZenPing;
import org.elasticsearch.discovery.zen.ping.ZenPingService;
import org.elasticsearch.discovery.zen.publish.PublishClusterStateAction;
import org.elasticsearch.node.service.NodeService;
import org.elasticsearch.node.settings.NodeSettingsService;
import org.elasticsearch.threadpool.ThreadPool;
import org.elasticsearch.transport.*;

import java.io.IOException;
import java.util.ArrayList;
import java.util.List;
import java.util.Set;
import java.util.concurrent.BlockingQueue;
import java.util.concurrent.CopyOnWriteArrayList;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.atomic.AtomicLong;
import java.util.concurrent.atomic.AtomicReference;

import static com.google.common.collect.Lists.newArrayList;
import static org.elasticsearch.common.unit.TimeValue.timeValueSeconds;

/**
*
*/
public class ZenDiscovery extends AbstractLifecycleComponent<Discovery> implements Discovery, PingContextProvider {

    public final static String SETTING_REJOIN_ON_MASTER_GONE = "discovery.zen.rejoin_on_master_gone";
    public final static String SETTING_PING_TIMEOUT = "discovery.zen.ping.timeout";
    public final static String SETTING_JOIN_TIMEOUT = "discovery.zen.join_timeout";
    public final static String SETTING_JOIN_RETRY_ATTEMPTS = "discovery.zen.join_retry_attempts";
    public final static String SETTING_JOIN_RETRY_DELAY = "discovery.zen.join_retry_delay";
    public final static String SETTING_MAX_PINGS_FROM_ANOTHER_MASTER = "discovery.zen.max_pings_from_another_master";
    public final static String SETTING_SEND_LEAVE_REQUEST = "discovery.zen.send_leave_request";
    public final static String SETTING_MASTER_ELECTION_FILTER_CLIENT = "discovery.zen.master_election.filter_client";
    public final static String SETTING_MASTER_ELECTION_FILTER_DATA = "discovery.zen.master_election.filter_data";

    public static final String DISCOVERY_REJOIN_ACTION_NAME = "internal:discovery/zen/rejoin";

    private final TransportService transportService;
    private final ClusterService clusterService;
    private AllocationService allocationService;
    private final ClusterName clusterName;
    private final DiscoveryNodeService discoveryNodeService;
    private final DiscoverySettings discoverySettings;
    private final ZenPingService pingService;
    private final MasterFaultDetection masterFD;
    private final NodesFaultDetection nodesFD;
    private final PublishClusterStateAction publishClusterState;
    private final MembershipAction membership;

    private final TimeValue pingTimeout;
    private final TimeValue joinTimeout;

    /** how many retry attempts to perform if join request failed with an retriable error */
    private final int joinRetryAttempts;
    /** how long to wait before performing another join attempt after a join request failed with an retriable error */
    private final TimeValue joinRetryDelay;

    /** how many pings from *another* master to tolerate before forcing a rejoin on other or local master */
    private final int maxPingsFromAnotherMaster;

    // a flag that should be used only for testing
    private final boolean sendLeaveRequest;

    private final ElectMasterService electMaster;

    private final boolean masterElectionFilterClientNodes;
    private final boolean masterElectionFilterDataNodes;


    private final CopyOnWriteArrayList<InitialStateDiscoveryListener> initialStateListeners = new CopyOnWriteArrayList<>();

    private final JoinThreadControl joinThreadControl;

    private final AtomicBoolean initialStateSent = new AtomicBoolean();

    private volatile boolean rejoinOnMasterGone;

    /** counts the time this node has joined the cluster or have elected it self as master */
    private final AtomicLong clusterJoinsCounter = new AtomicLong();

    @Nullable
    private NodeService nodeService;

    private final BlockingQueue<Tuple<DiscoveryNode, MembershipAction.JoinCallback>> processJoinRequests = ConcurrentCollections.newBlockingQueue();

    @Inject
    public ZenDiscovery(Settings settings, ClusterName clusterName, ThreadPool threadPool,
                        TransportService transportService, final ClusterService clusterService, NodeSettingsService nodeSettingsService,
                        DiscoveryNodeService discoveryNodeService, ZenPingService pingService, ElectMasterService electMasterService,
                        DiscoverySettings discoverySettings, @ClusterDynamicSettings DynamicSettings dynamicSettings) {
        super(settings);
        this.clusterName = clusterName;
        this.clusterService = clusterService;
        this.transportService = transportService;
        this.discoveryNodeService = discoveryNodeService;
        this.discoverySettings = discoverySettings;
        this.pingService = pingService;
        this.electMaster = electMasterService;

        // keep using componentSettings for BWC, in case this class gets extended.
        TimeValue pingTimeout = componentSettings.getAsTime("initial_ping_timeout", timeValueSeconds(3));
        pingTimeout = componentSettings.getAsTime("ping_timeout", pingTimeout);
        pingTimeout = settings.getAsTime("discovery.zen.ping_timeout", pingTimeout);
        this.pingTimeout = settings.getAsTime(SETTING_PING_TIMEOUT, pingTimeout);

        this.joinTimeout = settings.getAsTime(SETTING_JOIN_TIMEOUT, TimeValue.timeValueMillis(pingTimeout.millis() * 20));
        this.joinRetryAttempts = settings.getAsInt(SETTING_JOIN_RETRY_ATTEMPTS, 3);
        this.joinRetryDelay = settings.getAsTime(SETTING_JOIN_RETRY_DELAY, TimeValue.timeValueMillis(100));
        this.maxPingsFromAnotherMaster = settings.getAsInt(SETTING_MAX_PINGS_FROM_ANOTHER_MASTER, 3);
        this.sendLeaveRequest = settings.getAsBoolean(SETTING_SEND_LEAVE_REQUEST, true);

        this.masterElectionFilterClientNodes = settings.getAsBoolean(SETTING_MASTER_ELECTION_FILTER_CLIENT, true);
        this.masterElectionFilterDataNodes = settings.getAsBoolean(SETTING_MASTER_ELECTION_FILTER_DATA, false);
        this.rejoinOnMasterGone = settings.getAsBoolean(SETTING_REJOIN_ON_MASTER_GONE, true);

        if (this.joinRetryAttempts < 1) {
            throw new ElasticsearchIllegalArgumentException("'" + SETTING_JOIN_RETRY_ATTEMPTS + "' must be a positive number. got [" + this.SETTING_JOIN_RETRY_ATTEMPTS + "]");
        }
        if (this.maxPingsFromAnotherMaster < 1) {
            throw new ElasticsearchIllegalArgumentException("'" + SETTING_MAX_PINGS_FROM_ANOTHER_MASTER + "' must be a positive number. got [" + this.maxPingsFromAnotherMaster + "]");
        }

        logger.debug("using ping.timeout [{}], join.timeout [{}], master_election.filter_client [{}], master_election.filter_data [{}]", pingTimeout, joinTimeout, masterElectionFilterClientNodes, masterElectionFilterDataNodes);

        nodeSettingsService.addListener(new ApplySettings());

        this.masterFD = new MasterFaultDetection(settings, threadPool, transportService, clusterName, clusterService);
        this.masterFD.addListener(new MasterNodeFailureListener());

        this.nodesFD = new NodesFaultDetection(settings, threadPool, transportService, clusterName);
        this.nodesFD.addListener(new NodeFaultDetectionListener());

        this.publishClusterState = new PublishClusterStateAction(settings, transportService, this, new NewClusterStateListener(), discoverySettings, clusterName);
        this.pingService.setPingContextProvider(this);
        this.membership = new MembershipAction(settings, clusterService, transportService, this, new MembershipListener());

        this.joinThreadControl = new JoinThreadControl(threadPool);

        transportService.registerHandler(DISCOVERY_REJOIN_ACTION_NAME, new RejoinClusterRequestHandler());

        dynamicSettings.addDynamicSetting(ElectMasterService.DISCOVERY_ZEN_MINIMUM_MASTER_NODES, new Validator() {
            @Override
            public String validate(String setting, String value) {
                int intValue;
                try {
                    intValue = Integer.parseInt(value);
                } catch (NumberFormatException ex) {
                    return "cannot parse value [" + value + "] as an integer";
                }
                int masterNodes = clusterService.state().nodes().masterNodes().size();
                if (intValue > masterNodes) {
                    return "cannot set " + ElectMasterService.DISCOVERY_ZEN_MINIMUM_MASTER_NODES + " to more than the current master nodes count [" + masterNodes + "]";
                }
                return null;
            }
        });

    }

    @Override
    public void setNodeService(@Nullable NodeService nodeService) {
        this.nodeService = nodeService;
    }

    @Override
    public void setAllocationService(AllocationService allocationService) {
        this.allocationService = allocationService;
    }

    @Override
    protected void doStart() throws ElasticsearchException {
        nodesFD.setLocalNode(clusterService.localNode());
        joinThreadControl.start();
        pingService.start();

        // start the join thread from a cluster state update. See {@link JoinThreadControl} for details.
        clusterService.submitStateUpdateTask("initial_join", new ClusterStateNonMasterUpdateTask() {
            @Override
            public ClusterState execute(ClusterState currentState) throws Exception {
                // do the join on a different thread, the DiscoveryService waits for 30s anyhow till it is discovered
                joinThreadControl.startNewThreadIfNotRunning();
                return currentState;
            }

            @Override
            public void onFailure(String source, @org.elasticsearch.common.Nullable Throwable t) {
                logger.warn("failed to start initial join process", t);
            }
        });
    }

    @Override
    protected void doStop() throws ElasticsearchException {
        joinThreadControl.stop();
        pingService.stop();
        masterFD.stop("zen disco stop");
        nodesFD.stop();
        initialStateSent.set(false);
        DiscoveryNodes nodes = nodes();
        if (sendLeaveRequest) {
            if (nodes.masterNode() == null) {
                // if we don't know who the master is, nothing to do here
            } else if (!nodes.localNodeMaster()) {
                try {
                    membership.sendLeaveRequestBlocking(nodes.masterNode(), nodes.localNode(), TimeValue.timeValueSeconds(1));
                } catch (Exception e) {
                    logger.debug("failed to send leave request to master [{}]", e, nodes.masterNode());
                }
            } else {
                // we're master -> let other potential master we left and start a master election now rather then wait for masterFD
                DiscoveryNode[] possibleMasters = electMaster.nextPossibleMasters(nodes.nodes().values(), 5);
                for (DiscoveryNode possibleMaster : possibleMasters) {
                    if (nodes.localNode().equals(possibleMaster)) {
                        continue;
                    }
                    try {
                        membership.sendLeaveRequest(nodes.localNode(), possibleMaster);
                    } catch (Exception e) {
                        logger.debug("failed to send leave request from master [{}] to possible master [{}]", e, nodes.masterNode(), possibleMaster);
                    }
                }
            }
        }
    }

    @Override
    protected void doClose() throws ElasticsearchException {
        masterFD.close();
        nodesFD.close();
        publishClusterState.close();
        membership.close();
        pingService.close();
    }

    @Override
    public DiscoveryNode localNode() {
        return clusterService.localNode();
    }

    @Override
    public void addListener(InitialStateDiscoveryListener listener) {
        this.initialStateListeners.add(listener);
    }

    @Override
    public void removeListener(InitialStateDiscoveryListener listener) {
        this.initialStateListeners.remove(listener);
    }

    @Override
    public String nodeDescription() {
        return clusterName.value() + "/" + clusterService.localNode().id();
    }

    /** start of {@link org.elasticsearch.discovery.zen.ping.PingContextProvider } implementation */
    @Override
    public DiscoveryNodes nodes() {
        return clusterService.state().nodes();
    }

    @Override
    public NodeService nodeService() {
        return this.nodeService;
    }

    @Override
    public boolean nodeHasJoinedClusterOnce() {
        return clusterJoinsCounter.get() > 0;
    }

    /** end of {@link org.elasticsearch.discovery.zen.ping.PingContextProvider } implementation */


    @Override
    public void publish(ClusterState clusterState, AckListener ackListener) {
        if (!clusterState.getNodes().localNodeMaster()) {
            throw new ElasticsearchIllegalStateException("Shouldn't publish state when not master");
        }
        nodesFD.updateNodesAndPing(clusterState);
        publishClusterState.publish(clusterState, ackListener);
    }

    /**
     * returns true if zen discovery is started and there is a currently a background thread active for (re)joining
     * the cluster used for testing.
     */
    public boolean joiningCluster() {
        return joinThreadControl.joinThreadActive();
    }

    /**
     * the main function of a join thread. This function is guaranteed to join the cluster
     * or spawn a new join thread upon failure to do so.
     */
    private void innerJoinCluster() {
        DiscoveryNode masterNode = null;
        final Thread currentThread = Thread.currentThread();
        while (masterNode == null && joinThreadControl.joinThreadActive(currentThread)) {
            masterNode = findMaster();
        }

        if (!joinThreadControl.joinThreadActive(currentThread)) {
            logger.trace("thread is no longer in currentJoinThread. Stopping.");
            return;
        }

        if (clusterService.localNode().equals(masterNode)) {
            clusterService.submitStateUpdateTask("zen-disco-join (elected_as_master)", Priority.IMMEDIATE, new ProcessedClusterStateNonMasterUpdateTask() {
                @Override
                public ClusterState execute(ClusterState currentState) {
                    // Take into account the previous known nodes, if they happen not to be available
                    // then fault detection will remove these nodes.

                    if (currentState.nodes().masterNode() != null) {
                        // TODO can we tie break here? we don't have a remote master cluster state version to decide on
                        logger.trace("join thread elected local node as master, but there is already a master in place: {}", currentState.nodes().masterNode());
                        return currentState;
                    }

                    DiscoveryNodes.Builder builder = new DiscoveryNodes.Builder(currentState.nodes()).masterNodeId(currentState.nodes().localNode().id());
                    // update the fact that we are the master...
                    ClusterBlocks clusterBlocks = ClusterBlocks.builder().blocks(currentState.blocks()).removeGlobalBlock(discoverySettings.getNoMasterBlock()).build();
                    currentState = ClusterState.builder(currentState).nodes(builder).blocks(clusterBlocks).build();

                    // eagerly run reroute to remove dead nodes from routing table
                    RoutingAllocation.Result result = allocationService.reroute(currentState);
                    return ClusterState.builder(currentState).routingResult(result).build();
                }

                @Override
                public void onFailure(String source, Throwable t) {
                    logger.error("unexpected failure during [{}]", t, source);
                    joinThreadControl.markThreadAsDoneAndStartNew(currentThread);
                }

                @Override
                public void clusterStateProcessed(String source, ClusterState oldState, ClusterState newState) {
                    if (newState.nodes().localNodeMaster()) {
                        // we only starts nodesFD if we are master (it may be that we received a cluster state while pinging)
                        joinThreadControl.markThreadAsDone(currentThread);
                        nodesFD.updateNodesAndPing(newState); // start the nodes FD
                    } else {
                        // if we're not a master it means another node published a cluster state while we were pinging
                        // make sure we go through another pinging round and actively join it
                        joinThreadControl.markThreadAsDoneAndStartNew(currentThread);
                    }
                    sendInitialStateEventIfNeeded();
                    long count = clusterJoinsCounter.incrementAndGet();
                    logger.trace("cluster joins counter set to [{}] (elected as master)", count);

                }
            });
        } else {
            // send join request
            final boolean success = joinElectedMaster(masterNode);

            // finalize join through the cluster state update thread
            final DiscoveryNode finalMasterNode = masterNode;
            clusterService.submitStateUpdateTask("finalize_join (" + masterNode + ")", new ClusterStateNonMasterUpdateTask() {
                @Override
                public ClusterState execute(ClusterState currentState) throws Exception {
                    if (!success) {
                        // failed to join. Try again...
                        joinThreadControl.markThreadAsDoneAndStartNew(currentThread);
                        return currentState;
                    }

                    if (currentState.getNodes().masterNode() == null) {
                        // Post 1.3.0, the master should publish a new cluster state before acking our join request. we now should have
                        // a valid master.
                        logger.debug("no master node is set, despite of join request completing. retrying pings.");
                        joinThreadControl.markThreadAsDoneAndStartNew(currentThread);
                        return currentState;
                    }

                    if (!currentState.getNodes().masterNode().equals(finalMasterNode)) {
                        return joinThreadControl.stopRunningThreadAndRejoin(currentState, "master_switched_while_finalizing_join");
                    }

                    // Note: we do not have to start master fault detection here because it's set at {@link #handleNewClusterStateFromMaster }
                    // when the first cluster state arrives.
                    joinThreadControl.markThreadAsDone(currentThread);
                    return currentState;
                }

                @Override
                public void onFailure(String source, @Nullable Throwable t) {
                    logger.error("unexpected error while trying to finalize cluster join", t);
                    joinThreadControl.markThreadAsDoneAndStartNew(currentThread);
                }
            });
        }
    }

    /**
     * Join a newly elected master.
     *
     * @return true if successful
     */
    private boolean joinElectedMaster(DiscoveryNode masterNode) {
        try {
            // first, make sure we can connect to the master
            transportService.connectToNode(masterNode);
        } catch (Exception e) {
            logger.warn("failed to connect to master [{}], retrying...", e, masterNode);
            return false;
        }
        int joinAttempt = 0; // we retry on illegal state if the master is not yet ready
        while (true) {
            try {
                logger.trace("joining master {}", masterNode);
                membership.sendJoinRequestBlocking(masterNode, clusterService.localNode(), joinTimeout);
                return true;
            } catch (Throwable t) {
                Throwable unwrap = ExceptionsHelper.unwrapCause(t);
                if (unwrap instanceof ElasticsearchIllegalStateException) {
                    if (++joinAttempt == this.joinRetryAttempts) {
                        logger.info("failed to send join request to master [{}], reason [{}], tried [{}] times", masterNode, ExceptionsHelper.detailedMessage(t), joinAttempt);
                        return false;
                    } else {
                        logger.trace("master {} failed with [{}]. retrying... (attempts done: [{}])", masterNode, ExceptionsHelper.detailedMessage(t), joinAttempt);
                    }
                } else {
                    if (logger.isTraceEnabled()) {
                        logger.trace("failed to send join request to master [{}]", t, masterNode);
                    } else {
                        logger.info("failed to send join request to master [{}], reason [{}]", masterNode, ExceptionsHelper.detailedMessage(t));
                    }
                    return false;
                }
            }

            try {
                Thread.sleep(this.joinRetryDelay.millis());
            } catch (InterruptedException e) {
                Thread.currentThread().interrupt();
            }
        }
    }

    private void handleLeaveRequest(final DiscoveryNode node) {
        if (lifecycleState() != Lifecycle.State.STARTED) {
            // not started, ignore a node failure
            return;
        }
        if (localNodeMaster()) {
            clusterService.submitStateUpdateTask("zen-disco-node_left(" + node + ")", Priority.IMMEDIATE, new ClusterStateUpdateTask() {
                @Override
                public ClusterState execute(ClusterState currentState) {
                    DiscoveryNodes.Builder builder = DiscoveryNodes.builder(currentState.nodes()).remove(node.id());
                    currentState = ClusterState.builder(currentState).nodes(builder).build();
                    // check if we have enough master nodes, if not, we need to move into joining the cluster again
                    if (!electMaster.hasEnoughMasterNodes(currentState.nodes())) {
                        return rejoin(currentState, "not enough master nodes");
                    }
                    // eagerly run reroute to remove dead nodes from routing table
                    RoutingAllocation.Result routingResult = allocationService.reroute(ClusterState.builder(currentState).build());
                    return ClusterState.builder(currentState).routingResult(routingResult).build();
                }

                @Override
                public void onNoLongerMaster(String source) {
                    // ignoring (already logged)
                }

                @Override
                public void onFailure(String source, Throwable t) {
                    logger.error("unexpected failure during [{}]", t, source);
                }
            });
        } else if (node.equals(nodes().masterNode())) {
            handleMasterGone(node, "shut_down");
        }
    }

    private void handleNodeFailure(final DiscoveryNode node, String reason) {
        if (lifecycleState() != Lifecycle.State.STARTED) {
            // not started, ignore a node failure
            return;
        }
        if (!localNodeMaster()) {
            // nothing to do here...
            return;
        }
        clusterService.submitStateUpdateTask("zen-disco-node_failed(" + node + "), reason " + reason, Priority.IMMEDIATE, new ProcessedClusterStateUpdateTask() {
            @Override
            public ClusterState execute(ClusterState currentState) {
                DiscoveryNodes.Builder builder = DiscoveryNodes.builder(currentState.nodes())
                        .remove(node.id());
                currentState = ClusterState.builder(currentState).nodes(builder).build();
                // check if we have enough master nodes, if not, we need to move into joining the cluster again
                if (!electMaster.hasEnoughMasterNodes(currentState.nodes())) {
                    return rejoin(currentState, "not enough master nodes");
                }
                // eagerly run reroute to remove dead nodes from routing table
                RoutingAllocation.Result routingResult = allocationService.reroute(ClusterState.builder(currentState).build());
                return ClusterState.builder(currentState).routingResult(routingResult).build();
            }

            @Override
            public void onNoLongerMaster(String source) {
                // already logged
            }

            @Override
            public void onFailure(String source, Throwable t) {
                logger.error("unexpected failure during [{}]", t, source);
            }

            @Override
            public void clusterStateProcessed(String source, ClusterState oldState, ClusterState newState) {
                sendInitialStateEventIfNeeded();
            }
        });
    }

    private void handleMinimumMasterNodesChanged(final int minimumMasterNodes) {
        if (lifecycleState() != Lifecycle.State.STARTED) {
            // not started, ignore a node failure
            return;
        }
        final int prevMinimumMasterNode = ZenDiscovery.this.electMaster.minimumMasterNodes();
        ZenDiscovery.this.electMaster.minimumMasterNodes(minimumMasterNodes);
        if (!localNodeMaster()) {
            // We only set the new value. If the master doesn't see enough nodes it will revoke it's mastership.
            return;
        }
        clusterService.submitStateUpdateTask("zen-disco-minimum_master_nodes_changed", Priority.IMMEDIATE, new ProcessedClusterStateUpdateTask() {
            @Override
            public ClusterState execute(ClusterState currentState) {
                // check if we have enough master nodes, if not, we need to move into joining the cluster again
                if (!electMaster.hasEnoughMasterNodes(currentState.nodes())) {
                    return rejoin(currentState, "not enough master nodes on change of minimum_master_nodes from [" + prevMinimumMasterNode + "] to [" + minimumMasterNodes + "]");
                }
                return currentState;
            }


            @Override
            public void onNoLongerMaster(String source) {
                // ignoring (already logged)
            }

            @Override
            public void onFailure(String source, Throwable t) {
                logger.error("unexpected failure during [{}]", t, source);
            }

            @Override
            public void clusterStateProcessed(String source, ClusterState oldState, ClusterState newState) {
                sendInitialStateEventIfNeeded();
            }
        });
    }

    private void handleMasterGone(final DiscoveryNode masterNode, final String reason) {
        if (lifecycleState() != Lifecycle.State.STARTED) {
            // not started, ignore a master failure
            return;
        }
        if (localNodeMaster()) {
            // we might get this on both a master telling us shutting down, and then the disconnect failure
            return;
        }

        logger.info("master_left [{}], reason [{}]", masterNode, reason);

        clusterService.submitStateUpdateTask("zen-disco-master_failed (" + masterNode + ")", Priority.IMMEDIATE, new ProcessedClusterStateNonMasterUpdateTask() {
            @Override
            public ClusterState execute(ClusterState currentState) {
                if (!masterNode.id().equals(currentState.nodes().masterNodeId())) {
                    // master got switched on us, no need to send anything
                    return currentState;
                }

                DiscoveryNodes discoveryNodes = DiscoveryNodes.builder(currentState.nodes())
                        // make sure the old master node, which has failed, is not part of the nodes we publish
                        .remove(masterNode.id())
                        .masterNodeId(null).build();

                // flush any pending cluster states from old master, so it will not be set as master again
                ArrayList<ProcessClusterState> pendingNewClusterStates = new ArrayList<>();
                processNewClusterStates.drainTo(pendingNewClusterStates);
                logger.trace("removed [{}] pending cluster states", pendingNewClusterStates.size());

                if (rejoinOnMasterGone) {
                    return rejoin(ClusterState.builder(currentState).nodes(discoveryNodes).build(), "master left (reason = " + reason + ")");
                }

                if (!electMaster.hasEnoughMasterNodes(discoveryNodes)) {
                    return rejoin(ClusterState.builder(currentState).nodes(discoveryNodes).build(), "not enough master nodes after master left (reason = " + reason + ")");
                }

                final DiscoveryNode electedMaster = electMaster.electMaster(discoveryNodes); // elect master
                final DiscoveryNode localNode = currentState.nodes().localNode();
                if (localNode.equals(electedMaster)) {
                    masterFD.stop("got elected as new master since master left (reason = " + reason + ")");
                    discoveryNodes = DiscoveryNodes.builder(discoveryNodes).masterNodeId(localNode.id()).build();
                    ClusterState newState = ClusterState.builder(currentState).nodes(discoveryNodes).build();
                    nodesFD.updateNodesAndPing(newState);
                    return newState;

                } else {
                    nodesFD.stop();
                    if (electedMaster != null) {
                        discoveryNodes = DiscoveryNodes.builder(discoveryNodes).masterNodeId(electedMaster.id()).build();
                        masterFD.restart(electedMaster, "possible elected master since master left (reason = " + reason + ")");
                        return ClusterState.builder(currentState)
                                .nodes(discoveryNodes)
                                .build();
                    } else {
                        return rejoin(ClusterState.builder(currentState).nodes(discoveryNodes).build(), "master_left and no other node elected to become master");
                    }
                }
            }

            @Override
            public void onFailure(String source, Throwable t) {
                logger.error("unexpected failure during [{}]", t, source);
            }

            @Override
            public void clusterStateProcessed(String source, ClusterState oldState, ClusterState newState) {
                sendInitialStateEventIfNeeded();
            }

        });
    }

    static class ProcessClusterState {
        final ClusterState clusterState;
        final PublishClusterStateAction.NewClusterStateListener.NewStateProcessed newStateProcessed;
        volatile boolean processed;

        ProcessClusterState(ClusterState clusterState, PublishClusterStateAction.NewClusterStateListener.NewStateProcessed newStateProcessed) {
            this.clusterState = clusterState;
            this.newStateProcessed = newStateProcessed;
        }
    }

    private final BlockingQueue<ProcessClusterState> processNewClusterStates = ConcurrentCollections.newBlockingQueue();

    void handleNewClusterStateFromMaster(ClusterState newClusterState, final PublishClusterStateAction.NewClusterStateListener.NewStateProcessed newStateProcessed) {
        final ClusterName incomingClusterName = newClusterState.getClusterName();
        /* The cluster name can still be null if the state comes from a node that is prev 1.1.1*/
        if (incomingClusterName != null && !incomingClusterName.equals(this.clusterName)) {
            logger.warn("received cluster state from [{}] which is also master but with a different cluster name [{}]", newClusterState.nodes().masterNode(), incomingClusterName);
            newStateProcessed.onNewClusterStateFailed(new ElasticsearchIllegalStateException("received state from a node that is not part of the cluster"));
            return;
        }
        if (localNodeMaster()) {
            logger.debug("received cluster state from [{}] which is also master with cluster name [{}]", newClusterState.nodes().masterNode(), incomingClusterName);
            final ClusterState newState = newClusterState;
            clusterService.submitStateUpdateTask("zen-disco-master_receive_cluster_state_from_another_master [" + newState.nodes().masterNode() + "]", Priority.URGENT, new ProcessedClusterStateUpdateTask() {
                @Override
                public ClusterState execute(ClusterState currentState) {
                    return handleAnotherMaster(currentState, newState.nodes().masterNode(), newState.version(), "via a new cluster state");
                }

                @Override
                public void clusterStateProcessed(String source, ClusterState oldState, ClusterState newState) {
                    newStateProcessed.onNewClusterStateProcessed();
                }

                @Override
                public void onFailure(String source, Throwable t) {
                    logger.error("unexpected failure during [{}]", t, source);
                    newStateProcessed.onNewClusterStateFailed(t);
                }

            });
        } else {
            if (newClusterState.nodes().localNode() == null) {
                logger.warn("received a cluster state from [{}] and not part of the cluster, should not happen", newClusterState.nodes().masterNode());
                newStateProcessed.onNewClusterStateFailed(new ElasticsearchIllegalStateException("received state from a node that is not part of the cluster"));
            } else {

                final ProcessClusterState processClusterState = new ProcessClusterState(newClusterState, newStateProcessed);
                processNewClusterStates.add(processClusterState);


                assert newClusterState.nodes().masterNode() != null : "received a cluster state without a master";
                assert !newClusterState.blocks().hasGlobalBlock(discoverySettings.getNoMasterBlock()) : "received a cluster state with a master block";

                clusterService.submitStateUpdateTask("zen-disco-receive(from master [" + newClusterState.nodes().masterNode() + "])", Priority.URGENT, new ProcessedClusterStateNonMasterUpdateTask() {
                    @Override
                    public ClusterState execute(ClusterState currentState) {
                        // we already processed it in a previous event
                        if (processClusterState.processed) {
                            return currentState;
                        }

                        // TODO: once improvement that we can do is change the message structure to include version and masterNodeId
                        // at the start, this will allow us to keep the "compressed bytes" around, and only parse the first page
                        // to figure out if we need to use it or not, and only once we picked the latest one, parse the whole state


                        // try and get the state with the highest version out of all the ones with the same master node id
                        ProcessClusterState stateToProcess = processNewClusterStates.poll();
                        if (stateToProcess == null) {
                            return currentState;
                        }
                        stateToProcess.processed = true;
                        while (true) {
                            ProcessClusterState potentialState = processNewClusterStates.peek();
                            // nothing else in the queue, bail
                            if (potentialState == null) {
                                break;
                            }
                            // if its not from the same master, then bail
                            if (!Objects.equal(stateToProcess.clusterState.nodes().masterNodeId(), potentialState.clusterState.nodes().masterNodeId())) {
                                break;
                            }

                            // we are going to use it for sure, poll (remove) it
                            potentialState = processNewClusterStates.poll();
                            if (potentialState == null) {
                                // might happen if the queue is drained
                                break;
                            }

                            potentialState.processed = true;

                            if (potentialState.clusterState.version() > stateToProcess.clusterState.version()) {
                                // we found a new one
                                stateToProcess = potentialState;
                            }
                        }

                        ClusterState updatedState = stateToProcess.clusterState;

                        // if the new state has a smaller version, and it has the same master node, then no need to process it
                        if (updatedState.version() < currentState.version() && Objects.equal(updatedState.nodes().masterNodeId(), currentState.nodes().masterNodeId())) {
                            return currentState;
                        }

                        // we don't need to do this, since we ping the master, and get notified when it has moved from being a master
                        // because it doesn't have enough master nodes...
                        //if (!electMaster.hasEnoughMasterNodes(newState.nodes())) {
                        //    return disconnectFromCluster(newState, "not enough master nodes on new cluster state received from [" + newState.nodes().masterNode() + "]");
                        //}

                        // check to see that we monitor the correct master of the cluster
                        if (masterFD.masterNode() == null || !masterFD.masterNode().equals(updatedState.nodes().masterNode())) {
                            masterFD.restart(updatedState.nodes().masterNode(), "new cluster state received and we are monitoring the wrong master [" + masterFD.masterNode() + "]");
                        }

                        if (currentState.blocks().hasGlobalBlock(discoverySettings.getNoMasterBlock())) {
                            // its a fresh update from the master as we transition from a start of not having a master to having one
                            logger.debug("got first state from fresh master [{}]", updatedState.nodes().masterNodeId());
                            long count = clusterJoinsCounter.incrementAndGet();
                            logger.trace("updated cluster join cluster to [{}]", count);

                            return updatedState;
                        }


                        // some optimizations to make sure we keep old objects where possible
                        ClusterState.Builder builder = ClusterState.builder(updatedState);

                        // if the routing table did not change, use the original one
                        if (updatedState.routingTable().version() == currentState.routingTable().version()) {
                            builder.routingTable(currentState.routingTable());
                        }
                        // same for metadata
                        if (updatedState.metaData().version() == currentState.metaData().version()) {
                            builder.metaData(currentState.metaData());
                        } else {
                            // if its not the same version, only copy over new indices or ones that changed the version
                            MetaData.Builder metaDataBuilder = MetaData.builder(updatedState.metaData()).removeAllIndices();
                            for (IndexMetaData indexMetaData : updatedState.metaData()) {
                                IndexMetaData currentIndexMetaData = currentState.metaData().index(indexMetaData.index());
                                if (currentIndexMetaData == null || currentIndexMetaData.version() != indexMetaData.version()) {
                                    metaDataBuilder.put(indexMetaData, false);
                                } else {
                                    metaDataBuilder.put(currentIndexMetaData, false);
                                }
                            }
                            builder.metaData(metaDataBuilder);
                        }

                        return builder.build();
                    }

                    @Override
                    public void onFailure(String source, Throwable t) {
                        logger.error("unexpected failure during [{}]", t, source);
                        newStateProcessed.onNewClusterStateFailed(t);
                    }

                    @Override
                    public void clusterStateProcessed(String source, ClusterState oldState, ClusterState newState) {
                        sendInitialStateEventIfNeeded();
                        newStateProcessed.onNewClusterStateProcessed();
                    }
                });
            }
        }
    }

    private void handleJoinRequest(final DiscoveryNode node, final MembershipAction.JoinCallback callback) {

        if (!transportService.addressSupported(node.address().getClass())) {
            // TODO, what should we do now? Maybe inform that node that its crap?
            logger.warn("received a wrong address type from [{}], ignoring...", node);
        } else {
            // try and connect to the node, if it fails, we can raise an exception back to the client...
            transportService.connectToNode(node);

            // validate the join request, will throw a failure if it fails, which will get back to the
            // node calling the join request
            membership.sendValidateJoinRequestBlocking(node, joinTimeout);
            processJoinRequests.add(new Tuple<>(node, callback));
            clusterService.submitStateUpdateTask("zen-disco-receive(join from node[" + node + "])", Priority.URGENT, new ProcessedClusterStateUpdateTask() {

                private final List<Tuple<DiscoveryNode, MembershipAction.JoinCallback>> drainedJoinRequests = new ArrayList<>();

                @Override
                public ClusterState execute(ClusterState currentState) {
                    processJoinRequests.drainTo(drainedJoinRequests);
                    if (drainedJoinRequests.isEmpty()) {
                        return currentState;
                    }

                    boolean modified = false;
                    DiscoveryNodes.Builder nodesBuilder = DiscoveryNodes.builder(currentState.nodes());
                    for (Tuple<DiscoveryNode, MembershipAction.JoinCallback> task : drainedJoinRequests) {
                        DiscoveryNode node = task.v1();
                        if (currentState.nodes().nodeExists(node.id())) {
                            logger.debug("received a join request for an existing node [{}]", node);
                        } else {
                            modified = true;
                            nodesBuilder.put(node);
                            for (DiscoveryNode existingNode : currentState.nodes()) {
                                if (node.address().equals(existingNode.address())) {
                                    nodesBuilder.remove(existingNode.id());
                                    logger.warn("received join request from node [{}], but found existing node {} with same address, removing existing node", node, existingNode);
                                }
                            }
                        }
                    }

                    ClusterState.Builder stateBuilder = ClusterState.builder(currentState);
                    if (modified) {
                        stateBuilder.nodes(nodesBuilder);
                    }
                    return stateBuilder.build();
                }

                @Override
                public void onNoLongerMaster(String source) {
                    // we are rejected, so drain all pending task (execute never run)
                    processJoinRequests.drainTo(drainedJoinRequests);
                    Exception e = new ElasticsearchIllegalStateException("Node [" + clusterService.localNode() + "] not master for join request from [" + node + "]");
                    innerOnFailure(e);
                }

                void innerOnFailure(Throwable t) {
                    for (Tuple<DiscoveryNode, MembershipAction.JoinCallback> drainedTask : drainedJoinRequests) {
                        try {
                            drainedTask.v2().onFailure(t);
                        } catch (Exception e) {
                            logger.error("error during task failure", e);
                        }
                    }
                }

                @Override
                public void onFailure(String source, Throwable t) {
                    logger.error("unexpected failure during [{}]", t, source);
                    innerOnFailure(t);
                }

                @Override
                public void clusterStateProcessed(String source, ClusterState oldState, ClusterState newState) {
                    for (Tuple<DiscoveryNode, MembershipAction.JoinCallback> drainedTask : drainedJoinRequests) {
                        try {
                            drainedTask.v2().onSuccess();
                        } catch (Exception e) {
                            logger.error("unexpected error during [{}]", e, source);
                        }
                    }
                }
            });
        }
    }

    private DiscoveryNode findMaster() {
        logger.trace("starting to ping");
        ZenPing.PingResponse[] fullPingResponses = pingService.pingAndWait(pingTimeout);
        if (fullPingResponses == null) {
            logger.trace("No full ping responses");
            return null;
        }
        if (logger.isTraceEnabled()) {
            StringBuilder sb = new StringBuilder("full ping responses:");
            if (fullPingResponses.length == 0) {
                sb.append(" {none}");
            } else {
                for (ZenPing.PingResponse pingResponse : fullPingResponses) {
                    sb.append("\n\t--> ").append(pingResponse);
                }
            }
            logger.trace(sb.toString());
        }

        // filter responses
        List<ZenPing.PingResponse> pingResponses = Lists.newArrayList();
        for (ZenPing.PingResponse pingResponse : fullPingResponses) {
            DiscoveryNode node = pingResponse.node();
            if (masterElectionFilterClientNodes && (node.clientNode() || (!node.masterNode() && !node.dataNode()))) {
                // filter out the client node, which is a client node, or also one that is not data and not master (effectively, client)
            } else if (masterElectionFilterDataNodes && (!node.masterNode() && node.dataNode())) {
                // filter out data node that is not also master
            } else {
                pingResponses.add(pingResponse);
            }
        }

        if (logger.isDebugEnabled()) {
            StringBuilder sb = new StringBuilder("filtered ping responses: (filter_client[").append(masterElectionFilterClientNodes).append("], filter_data[").append(masterElectionFilterDataNodes).append("])");
            if (pingResponses.isEmpty()) {
                sb.append(" {none}");
            } else {
                for (ZenPing.PingResponse pingResponse : pingResponses) {
                    sb.append("\n\t--> ").append(pingResponse);
                }
            }
            logger.debug(sb.toString());
        }

        final DiscoveryNode localNode = clusterService.localNode();
        List<DiscoveryNode> pingMasters = newArrayList();
        for (ZenPing.PingResponse pingResponse : pingResponses) {
            if (pingResponse.master() != null) {
                // We can't include the local node in pingMasters list, otherwise we may up electing ourselves without
                // any check / verifications from other nodes in ZenDiscover#innerJoinCluster()
                if (!localNode.equals(pingResponse.master())) {
                    pingMasters.add(pingResponse.master());
                }
            }
        }

        // nodes discovered during pinging
        Set<DiscoveryNode> activeNodes = Sets.newHashSet();
        // nodes discovered who has previously been part of the cluster and do not ping for the very first time
        Set<DiscoveryNode> joinedOnceActiveNodes = Sets.newHashSet();
        if (localNode.masterNode()) {
            activeNodes.add(localNode);
            long joinsCounter = clusterJoinsCounter.get();
            if (joinsCounter > 0) {
                logger.trace("adding local node to the list of active nodes who has previously joined the cluster (joins counter is [{}})", joinsCounter);
                joinedOnceActiveNodes.add(localNode);
            }
        }
        for (ZenPing.PingResponse pingResponse : pingResponses) {
            activeNodes.add(pingResponse.node());
            if (pingResponse.hasJoinedOnce()) {
                joinedOnceActiveNodes.add(pingResponse.node());
            }
        }

        if (pingMasters.isEmpty()) {
            if (electMaster.hasEnoughMasterNodes(activeNodes)) {
                // we give preference to nodes who have previously already joined the cluster. Those will
                // have a cluster state in memory, including an up to date routing table (which is not persistent to disk
                // by the gateway)
                DiscoveryNode master = electMaster.electMaster(joinedOnceActiveNodes);
                if (master != null) {
                    return master;
                }
                return electMaster.electMaster(activeNodes);
            } else {
                // if we don't have enough master nodes, we bail, because there are not enough master to elect from
                logger.trace("not enough master nodes [{}]", activeNodes);
                return null;
            }
        } else {

            assert !pingMasters.contains(localNode) : "local node should never be elected as master when other nodes indicate an active master";
            // lets tie break between discovered nodes
            return electMaster.electMaster(pingMasters);
        }
    }

    private ClusterState rejoin(ClusterState clusterState, String reason) {

        // *** called from within an cluster state update task *** //
        assert Thread.currentThread().getName().contains(InternalClusterService.UPDATE_THREAD_NAME);

        logger.warn(reason + ", current nodes: {}", clusterState.nodes());
        nodesFD.stop();
        masterFD.stop(reason);


        ClusterBlocks clusterBlocks = ClusterBlocks.builder().blocks(clusterState.blocks())
                .addGlobalBlock(discoverySettings.getNoMasterBlock())
                .build();

        // clean the nodes, we are now not connected to anybody, since we try and reform the cluster
        DiscoveryNodes discoveryNodes = new DiscoveryNodes.Builder(clusterState.nodes()).masterNodeId(null).build();

        // TODO: do we want to force a new thread if we actively removed the master? this is to give a full pinging cycle
        // before a decision is made.
        joinThreadControl.startNewThreadIfNotRunning();

        return ClusterState.builder(clusterState)
                .blocks(clusterBlocks)
                .nodes(discoveryNodes)
                .build();
    }

    private boolean localNodeMaster() {
        return nodes().localNodeMaster();
    }

    private ClusterState handleAnotherMaster(ClusterState localClusterState, final DiscoveryNode otherMaster, long otherClusterStateVersion, String reason) {
        assert localClusterState.nodes().localNodeMaster() : "handleAnotherMaster called but current node is not a master";
        assert Thread.currentThread().getName().contains(InternalClusterService.UPDATE_THREAD_NAME) : "not called from the cluster state update thread";

        if (otherClusterStateVersion > localClusterState.version()) {
            return rejoin(localClusterState, "zen-disco-discovered another master with a new cluster_state [" + otherMaster + "][" + reason + "]");
        } else {
            logger.warn("discovered [{}] which is also master but with an older cluster_state, telling [{}] to rejoin the cluster ([{}])", otherMaster, otherMaster, reason);
            try {
                // make sure we're connected to this node (connect to node does nothing if we're already connected)
                // since the network connections are asymmetric, it may be that we received a state but have disconnected from the node
                // in the past (after a master failure, for example)
                transportService.connectToNode(otherMaster);
                transportService.sendRequest(otherMaster, DISCOVERY_REJOIN_ACTION_NAME, new RejoinClusterRequest(localClusterState.nodes().localNodeId()), new EmptyTransportResponseHandler(ThreadPool.Names.SAME) {

                    @Override
                    public void handleException(TransportException exp) {
                        logger.warn("failed to send rejoin request to [{}]", exp, otherMaster);
                    }
                });
            } catch (Exception e) {
                logger.warn("failed to send rejoin request to [{}]", e, otherMaster);
            }
            return localClusterState;
        }
    }

    private void sendInitialStateEventIfNeeded() {
        if (initialStateSent.compareAndSet(false, true)) {
            for (InitialStateDiscoveryListener listener : initialStateListeners) {
                listener.initialStateProcessed();
            }
        }
    }

    private class NewClusterStateListener implements PublishClusterStateAction.NewClusterStateListener {

        @Override
        public void onNewClusterState(ClusterState clusterState, NewStateProcessed newStateProcessed) {
            handleNewClusterStateFromMaster(clusterState, newStateProcessed);
        }
    }

    private class MembershipListener implements MembershipAction.MembershipListener {
        @Override
        public void onJoin(DiscoveryNode node, MembershipAction.JoinCallback callback) {
            handleJoinRequest(node, callback);
        }

        @Override
        public void onLeave(DiscoveryNode node) {
            handleLeaveRequest(node);
        }
    }

    private class NodeFaultDetectionListener extends NodesFaultDetection.Listener {

        private final AtomicInteger pingsWhileMaster = new AtomicInteger(0);

        @Override
        public void onNodeFailure(DiscoveryNode node, String reason) {
            handleNodeFailure(node, reason);
        }

        @Override
        public void onPingReceived(final NodesFaultDetection.PingRequest pingRequest) {
            // if we are master, we don't expect any fault detection from another node. If we get it
            // means we potentially have two masters in the cluster.
            if (!localNodeMaster()) {
                pingsWhileMaster.set(0);
                return;
            }

            // nodes pre 1.4.0 do not send this information
            if (pingRequest.masterNode() == null) {
                return;
            }

            if (pingsWhileMaster.incrementAndGet() < maxPingsFromAnotherMaster) {
                logger.trace("got a ping from another master {}. current ping count: [{}]", pingRequest.masterNode(), pingsWhileMaster.get());
                return;
            }
            logger.debug("got a ping from another master {}. resolving who should rejoin. current ping count: [{}]", pingRequest.masterNode(), pingsWhileMaster.get());
            clusterService.submitStateUpdateTask("ping from another master", Priority.IMMEDIATE, new ClusterStateUpdateTask() {

                @Override
                public ClusterState execute(ClusterState currentState) throws Exception {
                    pingsWhileMaster.set(0);
                    return handleAnotherMaster(currentState, pingRequest.masterNode(), pingRequest.clusterStateVersion(), "node fd ping");
                }

                @Override
                public void onFailure(String source, Throwable t) {
                    logger.debug("unexpected error during cluster state update task after pings from another master", t);
                }
            });
        }
    }

    private class MasterNodeFailureListener implements MasterFaultDetection.Listener {

        @Override
        public void onMasterFailure(DiscoveryNode masterNode, String reason) {
            handleMasterGone(masterNode, reason);
        }
    }

    boolean isRejoinOnMasterGone() {
        return rejoinOnMasterGone;
    }

    static class RejoinClusterRequest extends TransportRequest {

        private String fromNodeId;

        RejoinClusterRequest(String fromNodeId) {
            this.fromNodeId = fromNodeId;
        }

        RejoinClusterRequest() {
        }

        @Override
        public void readFrom(StreamInput in) throws IOException {
            super.readFrom(in);
            fromNodeId = in.readOptionalString();
        }

        @Override
        public void writeTo(StreamOutput out) throws IOException {
            super.writeTo(out);
            out.writeOptionalString(fromNodeId);
        }
    }

    class RejoinClusterRequestHandler extends BaseTransportRequestHandler<RejoinClusterRequest> {

        @Override
        public RejoinClusterRequest newInstance() {
            return new RejoinClusterRequest();
        }

        @Override
        public void messageReceived(final RejoinClusterRequest request, final TransportChannel channel) throws Exception {
            clusterService.submitStateUpdateTask("received a request to rejoin the cluster from [" + request.fromNodeId + "]", Priority.IMMEDIATE, new ClusterStateNonMasterUpdateTask() {
                @Override
                public ClusterState execute(ClusterState currentState) {
                    try {
                        channel.sendResponse(TransportResponse.Empty.INSTANCE);
                    } catch (Exception e) {
                        logger.warn("failed to send response on rejoin cluster request handling", e);
                    }
                    return rejoin(currentState, "received a request to rejoin the cluster from [" + request.fromNodeId + "]");
                }

                @Override
                public void onNoLongerMaster(String source) {
                    // already logged
                }

                @Override
                public void onFailure(String source, Throwable t) {
                    logger.error("unexpected failure during [{}]", t, source);
                }
            });
        }

        @Override
        public String executor() {
            return ThreadPool.Names.SAME;
        }
    }

    class ApplySettings implements NodeSettingsService.Listener {
        @Override
        public void onRefreshSettings(Settings settings) {
            int minimumMasterNodes = settings.getAsInt(ElectMasterService.DISCOVERY_ZEN_MINIMUM_MASTER_NODES,
                    ZenDiscovery.this.electMaster.minimumMasterNodes());
            if (minimumMasterNodes != ZenDiscovery.this.electMaster.minimumMasterNodes()) {
                logger.info("updating {} from [{}] to [{}]", ElectMasterService.DISCOVERY_ZEN_MINIMUM_MASTER_NODES,
                        ZenDiscovery.this.electMaster.minimumMasterNodes(), minimumMasterNodes);
                handleMinimumMasterNodesChanged(minimumMasterNodes);
            }

            boolean rejoinOnMasterGone = settings.getAsBoolean(SETTING_REJOIN_ON_MASTER_GONE, ZenDiscovery.this.rejoinOnMasterGone);
            if (rejoinOnMasterGone != ZenDiscovery.this.rejoinOnMasterGone) {
                logger.info("updating {} from [{}] to [{}]", SETTING_REJOIN_ON_MASTER_GONE, ZenDiscovery.this.rejoinOnMasterGone, rejoinOnMasterGone);
                ZenDiscovery.this.rejoinOnMasterGone = rejoinOnMasterGone;
            }
        }
    }


    /**
     * All control of the join thread should happen under the cluster state update task thread.
     * This is important to make sure that the background joining process is always in sync with any cluster state updates
     * like master loss, failure to join, received cluster state while joining etc.
     */
    private class JoinThreadControl {

        private final ThreadPool threadPool;
        private final AtomicBoolean running = new AtomicBoolean(false);
        private final AtomicReference<Thread> currentJoinThread = new AtomicReference<>();

        public JoinThreadControl(ThreadPool threadPool) {
            this.threadPool = threadPool;
        }

        /** returns true if join thread control is started and there is currently an active join thread */
        public boolean joinThreadActive() {
            Thread currentThread = currentJoinThread.get();
            return running.get() && currentThread != null && currentThread.isAlive();
        }

        /** returns true if join thread control is started and the supplied thread is the currently active joinThread */
        public boolean joinThreadActive(Thread joinThread) {
            return running.get() && joinThread.equals(currentJoinThread.get());
        }

        /** cleans any running joining thread and calls {@link #rejoin} */
        public ClusterState stopRunningThreadAndRejoin(ClusterState clusterState, String reason) {
            assertClusterStateThread();
            currentJoinThread.set(null);
            return rejoin(clusterState, reason);
        }

        /** starts a new joining thread if there is no currently active one and join thread controlling is started */
        public void startNewThreadIfNotRunning() {
            assertClusterStateThread();
            if (joinThreadActive()) {
                return;
            }
            threadPool.generic().execute(new Runnable() {
                @Override
                public void run() {
                    Thread currentThread = Thread.currentThread();
                    if (!currentJoinThread.compareAndSet(null, currentThread)) {
                        return;
                    }
                    while (running.get() && joinThreadActive(currentThread)) {
                        try {
                            innerJoinCluster();
                            return;
                        } catch (Exception e) {
                            logger.error("unexpected error while joining cluster, trying again", e);
                            // Because we catch any exception here, we want to know in
                            // tests if an uncaught exception got to this point and the test infra uncaught exception
                            // leak detection can catch this. In practise no uncaught exception should leak
                            assert ExceptionsHelper.reThrowIfNotNull(e);
                        }
                    }
                    // cleaning the current thread from currentJoinThread is done by explicit calls.
                }
            });
        }

        /**
         * marks the given joinThread as completed and makes sure another thread is running (starting one if needed)
         * If the given thread is not the currently running join thread, the command is ignored.
         */
        public void markThreadAsDoneAndStartNew(Thread joinThread) {
            assertClusterStateThread();
            if (!markThreadAsDone(joinThread)) {
                return;
            }
            startNewThreadIfNotRunning();
        }

        /** marks the given joinThread as completed. Returns false if the supplied thread is not the currently active join thread */
        public boolean markThreadAsDone(Thread joinThread) {
            assertClusterStateThread();
            return currentJoinThread.compareAndSet(joinThread, null);
        }

        public void stop() {
            running.set(false);
            Thread joinThread = currentJoinThread.getAndSet(null);
            if (joinThread != null) {
                joinThread.interrupt();
            }
        }

        public void start() {
            running.set(true);
        }

        private void assertClusterStateThread() {
            assert Thread.currentThread().getName().contains(InternalClusterService.UPDATE_THREAD_NAME) : "not called from the cluster state update thread";
        }

    }
}
TOP

Related Classes of org.elasticsearch.discovery.zen.ZenDiscovery$ProcessClusterState

TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.