Elasticsearch新的Discovery---Coordinator

简介

Elasticsearch目前在使用的服务发现算法，是基于zen的服务发现的第二代实现

节点模式

节点有三种模式，分别是CANDIDATE, LEADER, FOLLOWER。

流程分析

doStart

主要是服务发现的基本配置的工作，做了几件事：

1. 获取持久化的协调状态
2. 初始化节点发现器（探测器）
3. 选举配置
4. 初始化集群状态

protected void doStart() {
        synchronized (mutex) {
            // 获取持久化的协调状态
            CoordinationState.PersistedState persistedState = persistedStateSupplier.get();
            // 当前协调状态
            coordinationState.set(new CoordinationState(settings, getLocalNode(), persistedState));
            // 节点发现的数量
            peerFinder.setCurrentTerm(getCurrentTerm());
            // hosts解析器
            configuredHostsResolver.start();
            // 最近一次的状态
            final ClusterState lastAcceptedState = coordinationState.get().getLastAcceptedState();
            if (lastAcceptedState.metaData().clusterUUIDCommitted()) {
                logger.info("cluster UUID [{}]", lastAcceptedState.metaData().clusterUUID());
            }
            // 选举的节点的id，当集群状态变更的时候用到的配置 这里获取最近保存的配置信息
            final VotingConfiguration votingConfiguration = lastAcceptedState.getLastCommittedConfiguration();
            // 单节点发现，而且选举列表不为空，同时没有足够的选举节点就会引发异常
            if (singleNodeDiscovery &&
                votingConfiguration.isEmpty() == false &&
                votingConfiguration.hasQuorum(Collections.singleton(getLocalNode().getId())) == false) {
                throw new IllegalStateException("cannot start with [" + DiscoveryModule.DISCOVERY_TYPE_SETTING.getKey() + "] set to [" +
                    DiscoveryModule.SINGLE_NODE_DISCOVERY_TYPE + "] when local node " + getLocalNode() +
                    " does not have quorum in voting configuration " + votingConfiguration);
            }
            // 初始化集群状态
            ClusterState initialState = ClusterState.builder(ClusterName.CLUSTER_NAME_SETTING.get(settings))
                .blocks(ClusterBlocks.builder()
                    .addGlobalBlock(STATE_NOT_RECOVERED_BLOCK)
                    .addGlobalBlock(noMasterBlockService.getNoMasterBlock()))
                .nodes(DiscoveryNodes.builder().add(getLocalNode()).localNodeId(getLocalNode().getId()))
                .build();
            applierState = initialState;
            clusterApplier.setInitialState(initialState);
        }
    }

startInitialJoin

主要用于第一次启动时进行节点发现及选举加入集群。

  synchronized (mutex) {
            // 成为候选人，Mode初始化是null，也即是什么都不是，这里会改成CANDIDATE.
            becomeCandidate("startInitialJoin");
        }
        // 如果是master，启动集群
        clusterBootstrapService.scheduleUnconfiguredBootstrap();

becomeCandidate(String method)

void becomeCandidate(String method) {
        assert Thread.holdsLock(mutex) : "Coordinator mutex not held";
        logger.debug("{}: coordinator becoming CANDIDATE in term {} (was {}, lastKnownLeader was [{}])",
            method, getCurrentTerm(), mode, lastKnownLeader);
        // 第一次启动节点mode是null
        if (mode != Mode.CANDIDATE) {
            final Mode prevMode = mode;
            mode = Mode.CANDIDATE;
            cancelActivePublication("become candidate: " + method);
            joinAccumulator.close(mode);
            joinAccumulator = joinHelper.new CandidateJoinAccumulator();

            peerFinder.activate(coordinationState.get().getLastAcceptedState().nodes());
            clusterFormationFailureHelper.start();

            if (getCurrentTerm() == ZEN1_BWC_TERM) {
                discoveryUpgradeService.activate(lastKnownLeader, coordinationState.get().getLastAcceptedState());
            }

            leaderChecker.setCurrentNodes(DiscoveryNodes.EMPTY_NODES);
            leaderChecker.updateLeader(null);

            followersChecker.clearCurrentNodes();
            followersChecker.updateFastResponseState(getCurrentTerm(), mode);
            lagDetector.clearTrackedNodes();

            if (prevMode == Mode.LEADER) {
                cleanMasterService();
            }

            if (applierState.nodes().getMasterNodeId() != null) {
                applierState = clusterStateWithNoMasterBlock(applierState);
                clusterApplier.onNewClusterState("becoming candidate: " + method, () -> applierState, (source, e) -> {
                });
            }
        }

        preVoteCollector.update(getPreVoteResponse(), null);
    }