這篇文章將為大家詳細(xì)講解有關(guān)ribbon中ServerListSubsetFilter的作用是什么，文章內(nèi)容質(zhì)量較高，因此小編分享給大家做個(gè)參考，希望大家閱讀完這篇文章后對(duì)相關(guān)知識(shí)有一定的了解。

創(chuàng)新互聯(lián)公司是專(zhuān)業(yè)的中牟網(wǎng)站建設(shè)公司，中牟接單;提供成都網(wǎng)站設(shè)計(jì)、做網(wǎng)站,網(wǎng)頁(yè)設(shè)計(jì),網(wǎng)站設(shè)計(jì),建網(wǎng)站,PHP網(wǎng)站建設(shè)等專(zhuān)業(yè)做網(wǎng)站服務(wù);采用PHP框架,可快速的進(jìn)行中牟網(wǎng)站開(kāi)發(fā)網(wǎng)頁(yè)制作和功能擴(kuò)展;專(zhuān)業(yè)做搜索引擎喜愛(ài)的網(wǎng)站,專(zhuān)業(yè)的做網(wǎng)站團(tuán)隊(duì),希望更多企業(yè)前來(lái)合作!

ServerListSubsetFilter

ribbon-loadbalancer-2.3.0-sources.jar!/com/netflix/loadbalancer/ServerListSubsetFilter.java

public class ServerListSubsetFilter extends ZoneAffinityServerListFilter implements IClientConfigAware, Comparator{

    private Random random = new Random();
    private volatile Set currentSubset = Sets.newHashSet(); 
    private DynamicIntProperty sizeProp = new DynamicIntProperty(DefaultClientConfigImpl.DEFAULT_PROPERTY_NAME_SPACE + ".ServerListSubsetFilter.size", 20);
    private DynamicFloatProperty eliminationPercent = 
            new DynamicFloatProperty(DefaultClientConfigImpl.DEFAULT_PROPERTY_NAME_SPACE + ".ServerListSubsetFilter.forceEliminatePercent", 0.1f);
    private DynamicIntProperty eliminationFailureCountThreshold = 
            new DynamicIntProperty(DefaultClientConfigImpl.DEFAULT_PROPERTY_NAME_SPACE + ".ServerListSubsetFilter.eliminationFailureThresold", 0);
    private DynamicIntProperty eliminationConnectionCountThreshold = 
            new DynamicIntProperty(DefaultClientConfigImpl.DEFAULT_PROPERTY_NAME_SPACE + ".ServerListSubsetFilter.eliminationConnectionThresold", 0);
    
    @Override
    public void initWithNiwsConfig(IClientConfig clientConfig) {
        super.initWithNiwsConfig(clientConfig);
        sizeProp = new DynamicIntProperty(clientConfig.getClientName() + "." + clientConfig.getNameSpace() + ".ServerListSubsetFilter.size", 20);
        eliminationPercent = 
                new DynamicFloatProperty(clientConfig.getClientName() + "." + clientConfig.getNameSpace() + ".ServerListSubsetFilter.forceEliminatePercent", 0.1f);
        eliminationFailureCountThreshold = new DynamicIntProperty( clientConfig.getClientName() + "." + clientConfig.getNameSpace()
                + ".ServerListSubsetFilter.eliminationFailureThresold", 0);
        eliminationConnectionCountThreshold = new DynamicIntProperty(clientConfig.getClientName() + "." + clientConfig.getNameSpace()
                + ".ServerListSubsetFilter.eliminationConnectionThresold", 0);
    }
        

    @Override
    public List getFilteredListOfServers(List servers) {
        List zoneAffinityFiltered = super.getFilteredListOfServers(servers);
        Set candidates = Sets.newHashSet(zoneAffinityFiltered);
        Set newSubSet = Sets.newHashSet(currentSubset);
        LoadBalancerStats lbStats = getLoadBalancerStats();
        for (T server: currentSubset) {
            // this server is either down or out of service
            if (!candidates.contains(server)) {
                newSubSet.remove(server);
            } else {
                ServerStats stats = lbStats.getSingleServerStat(server);
                // remove the servers that do not meet health criteria
                if (stats.getActiveRequestsCount() > eliminationConnectionCountThreshold.get()
                        || stats.getFailureCount() > eliminationFailureCountThreshold.get()) {
                    newSubSet.remove(server);
                    // also remove from the general pool to avoid selecting them again
                    candidates.remove(server);
                }
            }
        }
        int targetedListSize = sizeProp.get();
        int numEliminated = currentSubset.size() - newSubSet.size();
        int minElimination = (int) (targetedListSize * eliminationPercent.get());
        int numToForceEliminate = 0;
        if (targetedListSize < newSubSet.size()) {
            // size is shrinking
            numToForceEliminate = newSubSet.size() - targetedListSize;
        } else if (minElimination > numEliminated) {
            numToForceEliminate = minElimination - numEliminated; 
        }
        
        if (numToForceEliminate > newSubSet.size()) {
            numToForceEliminate = newSubSet.size();
        }

        if (numToForceEliminate > 0) {
            List sortedSubSet = Lists.newArrayList(newSubSet);           
            Collections.sort(sortedSubSet, this);
            List forceEliminated = sortedSubSet.subList(0, numToForceEliminate);
            newSubSet.removeAll(forceEliminated);
            candidates.removeAll(forceEliminated);
        }
        
        // after forced elimination or elimination of unhealthy instances,
        // the size of the set may be less than the targeted size,
        // then we just randomly add servers from the big pool
        if (newSubSet.size() < targetedListSize) {
            int numToChoose = targetedListSize - newSubSet.size();
            candidates.removeAll(newSubSet);
            if (numToChoose > candidates.size()) {
                // Not enough healthy instances to choose, fallback to use the
                // total server pool
                candidates = Sets.newHashSet(zoneAffinityFiltered);
                candidates.removeAll(newSubSet);
            }
            List chosen = randomChoose(Lists.newArrayList(candidates), numToChoose);
            for (T server: chosen) {
                newSubSet.add(server);
            }
        }
        currentSubset = newSubSet;       
        return Lists.newArrayList(newSubSet);            
    }

    /**
     * Randomly shuffle the beginning portion of server list (according to the number passed into the method) 
     * and return them.
     *  
     * @param servers
     * @param toChoose
     * @return
     */
    private List randomChoose(List servers, int toChoose) {
        int size = servers.size();
        if (toChoose >= size || toChoose < 0) {
            return servers;
        } 
        for (int i = 0; i < toChoose; i++) {
            int index = random.nextInt(size);
            T tmp = servers.get(index);
            servers.set(index, servers.get(i));
            servers.set(i, tmp);
        }
        return servers.subList(0, toChoose);        
    }

    /**
     * Function to sort the list by server health condition, with
     * unhealthy servers before healthy servers. The servers are first sorted by
     * failures count, and then concurrent connection count.
     */
    @Override
    public int compare(T server1, T server2) {
        LoadBalancerStats lbStats = getLoadBalancerStats();
        ServerStats stats1 = lbStats.getSingleServerStat(server1);
        ServerStats stats2 = lbStats.getSingleServerStat(server2);
        int failuresDiff = (int) (stats2.getFailureCount() - stats1.getFailureCount());
        if (failuresDiff != 0) {
            return failuresDiff;
        } else {
            return (stats2.getActiveRequestsCount() - stats1.getActiveRequestsCount());
        }
    }
}

• ServerListSubsetFilter繼承了ZoneAffinityServerListFilter，實(shí)現(xiàn)了IClientConfigAware、Comparator接口

• initWithNiwsConfig方法從IClientConfig讀取了ServerListSubsetFilter.size、ServerListSubsetFilter.forceEliminatePercent、ServerListSubsetFilter.eliminationFailureThresold、ServerListSubsetFilter.eliminationConnectionThresold配置

• getFilteredListOfServers方法會(huì)先調(diào)用父類(lèi)ZoneAffinityServerListFilter的getFilteredListOfServers先過(guò)濾出zoneAffinityFiltered作為candidates，然后遍歷currentSubset根據(jù)ServerStats剔除activeRequestsCount及failureCount超出閾值的server；然后根據(jù)numToForceEliminate，以及failureCount排序，使用subList方法得出forceEliminated，然后移除掉，最后randomChoose出來(lái)新的newSubSet，然后重置currentSubset

小結(jié)

在server list非常多的場(chǎng)景下，沒(méi)有必要在連接池的保持這么多的連接，ServerListSubsetFilter可以在這種場(chǎng)景下對(duì)server list進(jìn)行精簡(jiǎn)，通過(guò)剔除相對(duì)不健康(failureCount、activeRequestCount)的server來(lái)達(dá)到此目標(biāo)

關(guān)于ribbon中ServerListSubsetFilter的作用是什么就分享到這里了，希望以上內(nèi)容可以對(duì)大家有一定的幫助，可以學(xué)到更多知識(shí)。如果覺(jué)得文章不錯(cuò)，可以把它分享出去讓更多的人看到。