91超碰碰碰碰久久久久久综合_超碰av人澡人澡人澡人澡人掠_国产黄大片在线观看画质优化_txt小说免费全本

溫馨提示×

溫馨提示×

您好,登錄后才能下訂單哦!

密碼登錄×
登錄注冊×
其他方式登錄
點擊 登錄注冊 即表示同意《億速云用戶服務條款》

nacos RaftCore中signalPublish的原理和使用方法

發布時間:2021-06-24 10:33:48 來源:億速云 閱讀:190 作者:chen 欄目:大數據

這篇文章主要講解了“nacos RaftCore中signalPublish的原理和使用方法”,文中的講解內容簡單清晰,易于學習與理解,下面請大家跟著小編的思路慢慢深入,一起來研究和學習“nacos RaftCore中signalPublish的原理和使用方法”吧!

本文主要研究一下nacos RaftCore的signalPublish

RaftCore

nacos-1.1.3/naming/src/main/java/com/alibaba/nacos/naming/consistency/persistent/raft/RaftCore.java

@Component
public class RaftCore {

    public static final String API_VOTE = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/vote";

    public static final String API_BEAT = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/beat";

    public static final String API_PUB = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum";

    public static final String API_DEL = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum";

    public static final String API_GET = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum";

    public static final String API_ON_PUB = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum/commit";

    public static final String API_ON_DEL = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum/commit";

    public static final String API_GET_PEER = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/peer";

    private ScheduledExecutorService executor = new ScheduledThreadPoolExecutor(1, new ThreadFactory() {
        @Override
        public Thread newThread(Runnable r) {
            Thread t = new Thread(r);

            t.setDaemon(true);
            t.setName("com.alibaba.nacos.naming.raft.notifier");

            return t;
        }
    });

    public static final Lock OPERATE_LOCK = new ReentrantLock();

    public static final int PUBLISH_TERM_INCREASE_COUNT = 100;

    private volatile Map<String, List<RecordListener>> listeners = new ConcurrentHashMap<>();

    private volatile ConcurrentMap<String, Datum> datums = new ConcurrentHashMap<>();

    @Autowired
    private RaftPeerSet peers;

    @Autowired
    private SwitchDomain switchDomain;

    @Autowired
    private GlobalConfig globalConfig;

    @Autowired
    private RaftProxy raftProxy;

    @Autowired
    private RaftStore raftStore;

    public volatile Notifier notifier = new Notifier();

    private boolean initialized = false;

    @PostConstruct
    public void init() throws Exception {

        Loggers.RAFT.info("initializing Raft sub-system");

        executor.submit(notifier);

        long start = System.currentTimeMillis();

        raftStore.loadDatums(notifier, datums);

        setTerm(NumberUtils.toLong(raftStore.loadMeta().getProperty("term"), 0L));

        Loggers.RAFT.info("cache loaded, datum count: {}, current term: {}", datums.size(), peers.getTerm());

        while (true) {
            if (notifier.tasks.size() <= 0) {
                break;
            }
            Thread.sleep(1000L);
        }

        initialized = true;

        Loggers.RAFT.info("finish to load data from disk, cost: {} ms.", (System.currentTimeMillis() - start));

        GlobalExecutor.registerMasterElection(new MasterElection());
        GlobalExecutor.registerHeartbeat(new HeartBeat());

        Loggers.RAFT.info("timer started: leader timeout ms: {}, heart-beat timeout ms: {}",
            GlobalExecutor.LEADER_TIMEOUT_MS, GlobalExecutor.HEARTBEAT_INTERVAL_MS);
    }

    public Map<String, List<RecordListener>> getListeners() {
        return listeners;
    }

    public void signalPublish(String key, Record value) throws Exception {

        if (!isLeader()) {
            JSONObject params = new JSONObject();
            params.put("key", key);
            params.put("value", value);
            Map<String, String> parameters = new HashMap<>(1);
            parameters.put("key", key);

            raftProxy.proxyPostLarge(getLeader().ip, API_PUB, params.toJSONString(), parameters);
            return;
        }

        try {
            OPERATE_LOCK.lock();
            long start = System.currentTimeMillis();
            final Datum datum = new Datum();
            datum.key = key;
            datum.value = value;
            if (getDatum(key) == null) {
                datum.timestamp.set(1L);
            } else {
                datum.timestamp.set(getDatum(key).timestamp.incrementAndGet());
            }

            JSONObject json = new JSONObject();
            json.put("datum", datum);
            json.put("source", peers.local());

            onPublish(datum, peers.local());

            final String content = JSON.toJSONString(json);

            final CountDownLatch latch = new CountDownLatch(peers.majorityCount());
            for (final String server : peers.allServersIncludeMyself()) {
                if (isLeader(server)) {
                    latch.countDown();
                    continue;
                }
                final String url = buildURL(server, API_ON_PUB);
                HttpClient.asyncHttpPostLarge(url, Arrays.asList("key=" + key), content, new AsyncCompletionHandler<Integer>() {
                    @Override
                    public Integer onCompleted(Response response) throws Exception {
                        if (response.getStatusCode() != HttpURLConnection.HTTP_OK) {
                            Loggers.RAFT.warn("[RAFT] failed to publish data to peer, datumId={}, peer={}, http code={}",
                                datum.key, server, response.getStatusCode());
                            return 1;
                        }
                        latch.countDown();
                        return 0;
                    }

                    @Override
                    public STATE onContentWriteCompleted() {
                        return STATE.CONTINUE;
                    }
                });

            }

            if (!latch.await(UtilsAndCommons.RAFT_PUBLISH_TIMEOUT, TimeUnit.MILLISECONDS)) {
                // only majority servers return success can we consider this update success
                Loggers.RAFT.error("data publish failed, caused failed to notify majority, key={}", key);
                throw new IllegalStateException("data publish failed, caused failed to notify majority, key=" + key);
            }

            long end = System.currentTimeMillis();
            Loggers.RAFT.info("signalPublish cost {} ms, key: {}", (end - start), key);
        } finally {
            OPERATE_LOCK.unlock();
        }
    }

    //......
}
  • signalPublish方法判斷當前節點是否是leader,如果不是則轉發publish到leader節點的/v1/ns/raft/datum接口

  • 如果是leader則構造datum以及peers.majorityCount()大小的CountDownLatch,然后遍歷peers.allServersIncludeMyself(),對于leader節點直接latch.countDown(),對于非leader節點則發送異步請求,請求/v1/ns/raft/datum/commit接口,在onCompleted的時候,如果請求成功執行latch.countDown()

  • 最后對于CountDownLatch未能在RAFT_PUBLISH_TIMEOUT返回的,拋出IllegalStateException

RaftController

nacos-1.1.3/naming/src/main/java/com/alibaba/nacos/naming/controllers/RaftController.java

@RestController
@RequestMapping({UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft",
    UtilsAndCommons.NACOS_SERVER_CONTEXT + UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft"})
public class RaftController {

    @Autowired
    private RaftConsistencyServiceImpl raftConsistencyService;

    @Autowired
    private ServiceManager serviceManager;

    @Autowired
    private RaftCore raftCore;

    //......

    @NeedAuth
    @RequestMapping(value = "/datum/commit", method = RequestMethod.POST)
    public String onPublish(HttpServletRequest request, HttpServletResponse response) throws Exception {

        response.setHeader("Content-Type", "application/json; charset=" + getAcceptEncoding(request));
        response.setHeader("Cache-Control", "no-cache");
        response.setHeader("Content-Encode", "gzip");

        String entity = IOUtils.toString(request.getInputStream(), "UTF-8");
        String value = URLDecoder.decode(entity, "UTF-8");
        JSONObject jsonObject = JSON.parseObject(value);
        String key = "key";

        RaftPeer source = JSON.parseObject(jsonObject.getString("source"), RaftPeer.class);
        JSONObject datumJson = jsonObject.getJSONObject("datum");

        Datum datum = null;
        if (KeyBuilder.matchInstanceListKey(datumJson.getString(key))) {
            datum = JSON.parseObject(jsonObject.getString("datum"), new TypeReference<Datum<Instances>>() {
            });
        } else if (KeyBuilder.matchSwitchKey(datumJson.getString(key))) {
            datum = JSON.parseObject(jsonObject.getString("datum"), new TypeReference<Datum<SwitchDomain>>() {
            });
        } else if (KeyBuilder.matchServiceMetaKey(datumJson.getString(key))) {
            datum = JSON.parseObject(jsonObject.getString("datum"), new TypeReference<Datum<Service>>() {
            });
        }

        raftConsistencyService.onPut(datum, source);
        return "ok";
    }

    //......
}
  • onPublish方法主要是執行raftConsistencyService.onPut(datum, source)

RaftConsistencyServiceImpl

nacos-1.1.3/naming/src/main/java/com/alibaba/nacos/naming/consistency/persistent/raft/RaftConsistencyServiceImpl.java

@Service
public class RaftConsistencyServiceImpl implements PersistentConsistencyService {

    @Autowired
    private RaftCore raftCore;

    @Autowired
    private RaftPeerSet peers;

    @Autowired
    private SwitchDomain switchDomain;

    //......

    public void onPut(Datum datum, RaftPeer source) throws NacosException {
        try {
            raftCore.onPublish(datum, source);
        } catch (Exception e) {
            Loggers.RAFT.error("Raft onPut failed.", e);
            throw new NacosException(NacosException.SERVER_ERROR, "Raft onPut failed, datum:" + datum + ", source: " + source, e);
        }
    }

    //......
}
  • onPut方法執行的是raftCore.onPublish(datum, source)

RaftCore

nacos-1.1.3/naming/src/main/java/com/alibaba/nacos/naming/consistency/persistent/raft/RaftCore.java

@Component
public class RaftCore {

    public static final String API_VOTE = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/vote";

    public static final String API_BEAT = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/beat";

    public static final String API_PUB = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum";

    public static final String API_DEL = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum";

    public static final String API_GET = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum";

    public static final String API_ON_PUB = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum/commit";

    public static final String API_ON_DEL = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/datum/commit";

    public static final String API_GET_PEER = UtilsAndCommons.NACOS_NAMING_CONTEXT + "/raft/peer";

    private ScheduledExecutorService executor = new ScheduledThreadPoolExecutor(1, new ThreadFactory() {
        @Override
        public Thread newThread(Runnable r) {
            Thread t = new Thread(r);

            t.setDaemon(true);
            t.setName("com.alibaba.nacos.naming.raft.notifier");

            return t;
        }
    });

    public static final Lock OPERATE_LOCK = new ReentrantLock();

    public static final int PUBLISH_TERM_INCREASE_COUNT = 100;

    private volatile Map<String, List<RecordListener>> listeners = new ConcurrentHashMap<>();

    private volatile ConcurrentMap<String, Datum> datums = new ConcurrentHashMap<>();

    @Autowired
    private RaftPeerSet peers;

    @Autowired
    private SwitchDomain switchDomain;

    @Autowired
    private GlobalConfig globalConfig;

    @Autowired
    private RaftProxy raftProxy;

    @Autowired
    private RaftStore raftStore;

    public volatile Notifier notifier = new Notifier();

    private boolean initialized = false;

    //......

    public void onPublish(Datum datum, RaftPeer source) throws Exception {
        RaftPeer local = peers.local();
        if (datum.value == null) {
            Loggers.RAFT.warn("received empty datum");
            throw new IllegalStateException("received empty datum");
        }

        if (!peers.isLeader(source.ip)) {
            Loggers.RAFT.warn("peer {} tried to publish data but wasn't leader, leader: {}",
                JSON.toJSONString(source), JSON.toJSONString(getLeader()));
            throw new IllegalStateException("peer(" + source.ip + ") tried to publish " +
                "data but wasn't leader");
        }

        if (source.term.get() < local.term.get()) {
            Loggers.RAFT.warn("out of date publish, pub-term: {}, cur-term: {}",
                JSON.toJSONString(source), JSON.toJSONString(local));
            throw new IllegalStateException("out of date publish, pub-term:"
                + source.term.get() + ", cur-term: " + local.term.get());
        }

        local.resetLeaderDue();

        // if data should be persistent, usually this is always true:
        if (KeyBuilder.matchPersistentKey(datum.key)) {
            raftStore.write(datum);
        }

        datums.put(datum.key, datum);

        if (isLeader()) {
            local.term.addAndGet(PUBLISH_TERM_INCREASE_COUNT);
        } else {
            if (local.term.get() + PUBLISH_TERM_INCREASE_COUNT > source.term.get()) {
                //set leader term:
                getLeader().term.set(source.term.get());
                local.term.set(getLeader().term.get());
            } else {
                local.term.addAndGet(PUBLISH_TERM_INCREASE_COUNT);
            }
        }
        raftStore.updateTerm(local.term.get());

        notifier.addTask(datum.key, ApplyAction.CHANGE);

        Loggers.RAFT.info("data added/updated, key={}, term={}", datum.key, local.term);
    }

    //......
}
  • onPublish方法首先判斷請求的節點是否是leader,不是則拋出IllegalStateException;對于source.term小于local.term的拋出IllegalStateException

  • 之后執行local.resetLeaderDue(),以及raftStore.write(datum),datums.put(datum.key, datum);對于leader節點執行local.term.addAndGet(PUBLISH_TERM_INCREASE_COUNT),非leader節點則更新leader term以及local.term

  • 最后執行raftStore.updateTerm(local.term.get())以及notifier.addTask(datum.key, ApplyAction.CHANGE)

小結

  • signalPublish方法判斷當前節點是否是leader,如果不是則轉發publish到leader節點的/v1/ns/raft/datum接口

  • 如果是leader則構造datum以及peers.majorityCount()大小的CountDownLatch,然后遍歷peers.allServersIncludeMyself(),對于leader節點直接latch.countDown(),對于非leader節點則發送異步請求,請求/v1/ns/raft/datum/commit接口,在onCompleted的時候,如果請求成功執行latch.countDown()

  • 最后對于CountDownLatch未能在RAFT_PUBLISH_TIMEOUT返回的,拋出IllegalStateException

感謝各位的閱讀,以上就是“nacos RaftCore中signalPublish的原理和使用方法”的內容了,經過本文的學習后,相信大家對nacos RaftCore中signalPublish的原理和使用方法這一問題有了更深刻的體會,具體使用情況還需要大家實踐驗證。這里是億速云,小編將為大家推送更多相關知識點的文章,歡迎關注!

向AI問一下細節

免責聲明:本站發布的內容(圖片、視頻和文字)以原創、轉載和分享為主,文章觀點不代表本網站立場,如果涉及侵權請聯系站長郵箱:is@yisu.com進行舉報,并提供相關證據,一經查實,將立刻刪除涉嫌侵權內容。

AI

商河县| 平谷区| 临猗县| 财经| 溧阳市| 年辖:市辖区| 宿松县| 慈利县| 灵宝市| 桂东县| 于都县| 尉氏县| 昌图县| 慈利县| 祁阳县| 咸宁市| 明星| 改则县| 南宁市| 翁牛特旗| 庆城县| 临朐县| 东光县| 南岸区| 邯郸市| 广德县| 鹤山市| 谢通门县| 大名县| 凉城县| 巍山| 土默特左旗| 博罗县| 怀集县| 渑池县| 都昌县| 郸城县| 卫辉市| 望江县| 新巴尔虎右旗| 莆田市|