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》

NacosNamingService中selectOneHealthyInstance的原理及作用是什么

發布時間:2021-06-26 09:39:58 來源:億速云 閱讀:325 作者:chen 欄目:大數據

本篇內容介紹了“NacosNamingService中selectOneHealthyInstance的原理及作用是什么”的有關知識,在實際案例的操作過程中,不少人都會遇到這樣的困境,接下來就讓小編帶領大家學習一下如何處理這些情況吧!希望大家仔細閱讀,能夠學有所成!

本文主要研究一下NacosNamingService的selectOneHealthyInstance

NacosNamingService

nacos-1.1.3/client/src/main/java/com/alibaba/nacos/client/naming/NacosNamingService.java

public class NacosNamingService implements NamingService {
    private static final String DEFAULT_PORT = "8080";
    private static final long DEFAULT_HEART_BEAT_INTERVAL = TimeUnit.SECONDS.toMillis(5);

    /**
     * Each Naming instance should have different namespace.
     */
    private String namespace;

    private String endpoint;

    private String serverList;

    private String cacheDir;

    private String logName;

    private HostReactor hostReactor;

    private BeatReactor beatReactor;

    private EventDispatcher eventDispatcher;

    private NamingProxy serverProxy;

    //......

    @Override
    public Instance selectOneHealthyInstance(String serviceName) throws NacosException {
        return selectOneHealthyInstance(serviceName, new ArrayList<String>());
    }

    @Override
    public Instance selectOneHealthyInstance(String serviceName, String groupName) throws NacosException {
        return selectOneHealthyInstance(serviceName, groupName, true);
    }

    @Override
    public Instance selectOneHealthyInstance(String serviceName, boolean subscribe) throws NacosException {
        return selectOneHealthyInstance(serviceName, new ArrayList<String>(), subscribe);
    }

    @Override
    public Instance selectOneHealthyInstance(String serviceName, String groupName, boolean subscribe) throws NacosException {
        return selectOneHealthyInstance(serviceName, groupName, new ArrayList<String>(), subscribe);
    }

    @Override
    public Instance selectOneHealthyInstance(String serviceName, List<String> clusters) throws NacosException {
        return selectOneHealthyInstance(serviceName, clusters, true);
    }

    @Override
    public Instance selectOneHealthyInstance(String serviceName, String groupName, List<String> clusters) throws NacosException {
        return selectOneHealthyInstance(serviceName, groupName, clusters, true);
    }

    @Override
    public Instance selectOneHealthyInstance(String serviceName, List<String> clusters, boolean subscribe)
        throws NacosException {
        return selectOneHealthyInstance(serviceName, Constants.DEFAULT_GROUP, clusters, subscribe);
    }

    @Override
    public Instance selectOneHealthyInstance(String serviceName, String groupName, List<String> clusters, boolean subscribe) throws NacosException {

        if (subscribe) {
            return Balancer.RandomByWeight.selectHost(
                hostReactor.getServiceInfo(NamingUtils.getGroupedName(serviceName, groupName), StringUtils.join(clusters, ",")));
        } else {
            return Balancer.RandomByWeight.selectHost(
                hostReactor.getServiceInfoDirectlyFromServer(NamingUtils.getGroupedName(serviceName, groupName), StringUtils.join(clusters, ",")));
        }
    }

    //......
}
  • selectOneHealthyInstance跟selectInstances類似,只不過它返回的是單個instance;selectOneHealthyInstance也是先從hostReactor獲取serviceInfo

  • 如果subscribe為true,則執行hostReactor.getServiceInfo獲取serviceInfo,否則執行hostReactor.getServiceInfoDirectlyFromServer獲取serviceInfo

  • 獲取到serviceInfo之后,selectOneHealthyInstance通過Balancer.RandomByWeight.selectHost方法來選取單個healthy的instance

Balancer

nacos-1.1.3/client/src/main/java/com/alibaba/nacos/client/naming/core/Balancer.java

public class Balancer {

    /**
     * report status to server
     */
    public final static List<String> UNCONSISTENT_SERVICE_WITH_ADDRESS_SERVER = new CopyOnWriteArrayList<String>();

    public static class RandomByWeight {

        public static List<Instance> selectAll(ServiceInfo serviceInfo) {
            List<Instance> hosts = serviceInfo.getHosts();

            if (CollectionUtils.isEmpty(hosts)) {
                throw new IllegalStateException("no host to srv for serviceInfo: " + serviceInfo.getName());
            }

            return hosts;
        }

        public static Instance selectHost(ServiceInfo dom) {

            List<Instance> hosts = selectAll(dom);

            if (CollectionUtils.isEmpty(hosts)) {
                throw new IllegalStateException("no host to srv for service: " + dom.getName());
            }

            return getHostByRandomWeight(hosts);
        }
    }

    /**
     * Return one host from the host list by random-weight.
     *
     * @param hosts The list of the host.
     * @return The random-weight result of the host
     */
    protected static Instance getHostByRandomWeight(List<Instance> hosts) {
        NAMING_LOGGER.debug("entry randomWithWeight");
        if (hosts == null || hosts.size() == 0) {
            NAMING_LOGGER.debug("hosts == null || hosts.size() == 0");
            return null;
        }

        Chooser<String, Instance> vipChooser = new Chooser<String, Instance>("www.taobao.com");

        NAMING_LOGGER.debug("new Chooser");

        List<Pair<Instance>> hostsWithWeight = new ArrayList<Pair<Instance>>();
        for (Instance host : hosts) {
            if (host.isHealthy()) {
                hostsWithWeight.add(new Pair<Instance>(host, host.getWeight()));
            }
        }
        NAMING_LOGGER.debug("for (Host host : hosts)");
        vipChooser.refresh(hostsWithWeight);
        NAMING_LOGGER.debug("vipChooser.refresh");
        return vipChooser.randomWithWeight();
    }
}
  • Balancer的RandomByWeight提供了selectAll及selectHost方法;selectAll針對serviceInfo.getHosts()進行了空判斷,空的話會拋出IllegalStateException

  • selectHost方法內部調用了selectAll方法,其最后通過getHostByRandomWeight來選取單個healthy的instance

  • getHostByRandomWeight方法首先創建一個Chooser,然后選取healthy的instance構造hostsWithWeight,再通過vipChooser.refresh(hostsWithWeight)進行refresh,最后通過vipChooser.randomWithWeight()選取單個healthy的instance

Chooser

nacos-1.1.3/client/src/main/java/com/alibaba/nacos/client/naming/utils/Chooser.java

public class Chooser<K, T> {

    private K uniqueKey;

    private volatile Ref<T> ref;

    public T random() {
        List<T> items = ref.items;
        if (items.size() == 0) {
            return null;
        }
        if (items.size() == 1) {
            return items.get(0);
        }
        return items.get(ThreadLocalRandom.current().nextInt(items.size()));
    }

    public T randomWithWeight() {
        Ref<T> ref = this.ref;
        double random = ThreadLocalRandom.current().nextDouble(0, 1);
        int index = Arrays.binarySearch(ref.weights, random);
        if (index < 0) {
            index = -index - 1;
        } else {
            return ref.items.get(index);
        }

        if (index >= 0 && index < ref.weights.length) {
            if (random < ref.weights[index]) {
                return ref.items.get(index);
            }
        }

        /* This should never happen, but it ensures we will return a correct
         * object in case there is some floating point inequality problem
         * wrt the cumulative probabilities. */
        return ref.items.get(ref.items.size() - 1);
    }

    public Chooser(K uniqueKey) {
        this(uniqueKey, new ArrayList<Pair<T>>());
    }

    public Chooser(K uniqueKey, List<Pair<T>> pairs) {
        Ref<T> ref = new Ref<T>(pairs);
        ref.refresh();
        this.uniqueKey = uniqueKey;
        this.ref = ref;
    }

    public K getUniqueKey() {
        return uniqueKey;
    }

    public Ref<T> getRef() {
        return ref;
    }

    public void refresh(List<Pair<T>> itemsWithWeight) {
        Ref<T> newRef = new Ref<T>(itemsWithWeight);
        newRef.refresh();
        newRef.poller = this.ref.poller.refresh(newRef.items);
        this.ref = newRef;
    }

    //......
}
  • Chooser的refresh方法會根據itemsWithWeight創建Ref,然后執行Ref的refresh方法;randomWithWeight方法通過Arrays.binarySearch(ref.weights, random)創建初始index,然后根據index從ref.items獲取元素

Ref

nacos-1.1.3/client/src/main/java/com/alibaba/nacos/client/naming/utils/Chooser.java

    public class Ref<T> {
        private List<Pair<T>> itemsWithWeight = new ArrayList<Pair<T>>();
        private List<T> items = new ArrayList<T>();
        private Poller<T> poller = new GenericPoller<T>(items);
        private double[] weights;

        @SuppressWarnings("unchecked")
        public Ref(List<Pair<T>> itemsWithWeight) {
            this.itemsWithWeight = itemsWithWeight;
        }

        public void refresh() {
            Double originWeightSum = (double) 0;

            for (Pair<T> item : itemsWithWeight) {

                double weight = item.weight();
                //ignore item which weight is zero.see test_randomWithWeight_weight0 in ChooserTest
                if (weight <= 0) {
                    continue;
                }

                items.add(item.item());
                if (Double.isInfinite(weight)) {
                    weight = 10000.0D;
                }
                if (Double.isNaN(weight)) {
                    weight = 1.0D;
                }
                originWeightSum += weight;
            }

            double[] exactWeights = new double[items.size()];
            int index = 0;
            for (Pair<T> item : itemsWithWeight) {
                double singleWeight = item.weight();
                //ignore item which weight is zero.see test_randomWithWeight_weight0 in ChooserTest
                if (singleWeight <= 0) {
                    continue;
                }
                exactWeights[index++] = singleWeight / originWeightSum;
            }

            weights = new double[items.size()];
            double randomRange = 0D;
            for (int i = 0; i < index; i++) {
                weights[i] = randomRange + exactWeights[i];
                randomRange += exactWeights[i];
            }

            double doublePrecisionDelta = 0.0001;

            if (index == 0 || (Math.abs(weights[index - 1] - 1) < doublePrecisionDelta)) {
                return;
            }
            throw new IllegalStateException("Cumulative Weight caculate wrong , the sum of probabilities does not equals 1.");
        }

        //......
    }
  • Ref的refresh方法主要是初始化items及weights

小結

  • selectOneHealthyInstance跟selectInstances類似,只不過它返回的是單個instance;selectOneHealthyInstance也是先從hostReactor獲取serviceInfo

  • 如果subscribe為true,則執行hostReactor.getServiceInfo獲取serviceInfo,否則執行hostReactor.getServiceInfoDirectlyFromServer獲取serviceInfo

  • 獲取到serviceInfo之后,selectOneHealthyInstance通過Balancer.RandomByWeight.selectHost方法來選取單個healthy的instance

“NacosNamingService中selectOneHealthyInstance的原理及作用是什么”的內容就介紹到這里了,感謝大家的閱讀。如果想了解更多行業相關的知識可以關注億速云網站,小編將為大家輸出更多高質量的實用文章!

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

新宾| 兴文县| 扶风县| 馆陶县| 龙南县| 上思县| 尉氏县| 怀化市| 江口县| 顺昌县| 图们市| 汨罗市| 崇仁县| 司法| 哈密市| 金川县| 沂源县| 蒲城县| 大化| 云南省| 丽江市| 威信县| 中宁县| 新平| 汕头市| 乌苏市| 肃南| 云阳县| 盘山县| 双桥区| 镇巴县| 盐源县| 永平县| 成都市| 随州市| 吉安市| 故城县| 西宁市| 资兴市| 廉江市| 密云县|