vip 02 zookeeper客戶端使用與集群特性

ZookeeperJava客戶端ZookeeperJava publicpublicclassTest3 privatestaticfinalString5 privatestaticfinalintSESSION_TIMEOUT=7 privatestaticZooKeeper9 privatestaticfinalStringpublicvoidinit()throwsIOException,InterruptedExceptionfinalCountDownLatchcountDownLatch=newzooKeeper=newZooKeeper(ZK_ADDRESS,SESSION_TIMEOUT,event‐>eventevent.getType()==} }26ZooKeeperZooKeeper(StringconnectString,intsessionTimeout,WatcherZooKeeper(StringconnectString,intsessionTimeout,Watcherwatcher,ZKZooKeeper(StringconnectString,intsessionTimeout,Watcherwatcher,booleananBeReadOnly,HostProvider)ZooKeeper(StringconnectString,intsessionTimeout,Watcherwatcher,booleananBeReadOnly,HostProvider,ZK ZooKeeper(StringconnectString,intsessionTimeout,Watcherwatcher,ZooKeeper(StringconnectString,intsessionTimeout,Watcherwatcher,booleananBeReadOnly,ZK ZooKeeper(StringconnectString,intsessionTimeout,Watcherwatcher,long,byZooKeeper(StringconnectString,intsessionTimeout,Watcherwatcher,long,e[],boolean,ZooKeeper(StringconnectString,intsessionTimeout,Watcherwatcher,long,bye[],boolean,HostProvider,ZK ZooKeeper(StringconnectString,intsessionTimeout,Watcherwatcher,long,byte[],boolean)后的 客戶端在構(gòu)造方法中傳入一個(gè)接口watcher(org.apache.zookeeper.數(shù)可以設(shè)置為null以表明不需要設(shè)置默認(rèn)的Watcher處理器。boeedn(只oKepe這就是ZooKeeper的“read-only”模式。DoKepeZoeper實(shí)longgetSessionId();publicvoidcreateTest()throwsKeeperException,InterruptedExceptionStringpath=zooKeeper.create(ZK_NODE,"data".getBytes(),CL_UNSAFE,("createdpath:5publicvoidcreateAsycTest()throwsInterruptedException(rc,path,ctx,name)‐>("rc{},path{},ctx7publicvoidsetTest()throwsKeeperException,InterruptedException33Statstat=newbyte[]data=zooKeeper.getData(ZK_NODE,false,("修改前,newbyte[]dataAfter=zooKeeper.getData(ZK_NODE,false,("修改后,new}Curator是一套flix公司開源的，Java語言編程的ZooKeeper客戶端框架，Curator項(xiàng)目用，Curator把 平時(shí)常用的很多ZooKeeper服務(wù)開發(fā)功能做了封裝，例如Leader、分布式計(jì)數(shù)器、分布式鎖。這就減少了技術(shù)人員在使用ZooKeeper時(shí)的大部分底層細(xì)節(jié)開發(fā)工作。在會話重新連接、Watch反復(fù)、多種異常處理等使用場景中，用原生的ZooKeeper處理比較復(fù)雜。而在使用Curator時(shí)，由于其對這些功能都做了高度的封裝，使用起來更加簡Curator 以Maven工程為例，首先要引入Curator框架相關(guān)的開發(fā)包，這里為了方便測試引入了junit，lombok，由于Zookeeper本身以來了log4j日志框架，所以這里可以創(chuàng)建對應(yīng)的log4j配置文件后直接使用。如下面的代碼所示， 通過將Curator相關(guān)的包配置到Maven工程的pom文件中，將Curaotr框架到工程項(xiàng)目里，在配置文件中分別了兩個(gè)Curator相關(guān)的包，第一個(gè)是curator-framework包，該包是對ZooKeeper底層API的一些封裝。另一個(gè)是curator-recipes包，該包封裝了一些ZooKeeper服務(wù)的高級特性，如：Cache、、分布式鎖、分布式Barrier。 =newExponentialBackoffRetry(1000,3nString,= fluent112=newExponentialBackoffRetry(1000,= 789義一個(gè)ZooKeeper客戶端對象，并在的上下文中使用。在定義CuratorFramework對象 使用了CuratorFrameworkFactory工廠方法，并指定了connectionString服務(wù)器地址列表、retry重試策略、sessionTimeoutMs會話超時(shí)時(shí)間、connectionTimeoutMs會話創(chuàng)建超時(shí)時(shí)間。下面 是多個(gè)地址。如果是多個(gè)地址，那么每個(gè)服務(wù)器地址列表用逗號分隔,如host1:port1,host2:port2,host3；port3。重新連接ZooKeeper服務(wù)端。而Curator提供了一次重試、多次重試等不同種類的實(shí)現(xiàn)方式。在Curator，可以通過判斷服務(wù)器返回的keeperException的狀態(tài)代碼來判斷是否進(jìn)OKSYSTEMERROR超時(shí)時(shí)間：Curator客戶端創(chuàng)建過程中，有兩個(gè)超時(shí)時(shí)間的設(shè)置。一個(gè)是sessionTimeoutMs會話超時(shí)時(shí)間，用來設(shè)置該條會話在ZooKeeper服務(wù)端的失效時(shí)間。另一個(gè)是ZooKeeper服務(wù)端應(yīng)答的時(shí)間。sessionTimeoutMs作用在服務(wù)端，而connectionTimeoutMs作用在客戶端。 22publicvoidtestCreate()ceptionStringpath=("curatorcreatenode:{}6CuratorcreatewithMode點(diǎn)，調(diào)用forPath函數(shù)來指定節(jié)點(diǎn)的路徑和數(shù)據(jù)信息。publicvoidtestCreateWithParent()ceptionStringStringpath=curatorFramework.create().creatingParentsIfNeeded().forPath(pa("curatorcreatenode:{}6publicvoidtestGetData()ceptionbyte[]bytes=("getdatafromnode:{}successfully.",new5后邊，通過forPath函數(shù)來指定更新的數(shù)據(jù)節(jié)點(diǎn)路徑以及要更新的數(shù)據(jù)。byte[]bytes=("getdatafromnode/curator‐node:{}successfully.",newString(bytceptionpublicvoidtestSetData()66publicvoidtestDelete()ceptionString5ZooKeeper服務(wù)端被刪除。在異步線程中調(diào)用，默認(rèn)在EventThread中調(diào)用，也可以自定義線程池。publicpublicinterface{ *Calledwhentheasyncbackgroundoperation** *@parameventoperation 89*ceptionpublicvoid},CuratorEventevent) inBackground異步處理默認(rèn)在EventThread中執(zhí)行publicvoidtest()ception ("background:{}",6 8publicvoidtest()ception ExecutorServiceexecutorService=4("background:{}",8}1*Receivesnotificationsabouterrorsandbackgroundpublicinterface{ *Calledwhenabackgroundtaskhascompletedorawatchhas***@parameventthe*ceptionanypublicvoid},CuratorEventevent)CuratorCurator引入了Cache來實(shí)現(xiàn)對Zookeeper服務(wù)端，Cache可以理解為一node112345publicString‐path‐pathto,publicclassNodeCacheTest3 publicstaticfinalString5publicvoidtestNodeCacheTest() ception8NodeCachenodeCache=newNodeCache(curatorFramework,nodeCache.getListenable().addListener(newNodeCacheListener()publicvoidnodeChanged() ("{}pathnodeChanged:}} publicvoidprintNodeData()ceptionpublicvoidaddListener(NodeCacheListenerAddachangelistener‐thebyte[]byte[] ("data:{}",new}27path11456publicStringpath,booleancacheData)‐path‐pathto,237cacheData‐iftrue,nodecontentsarecachedinadditiontothepublicclassPathCacheTest3 publicstaticfinalString5publicvoidtestPathCache()ception8PathChildrenCachepathChildrenCache=newPathChildrenCache(curatorFramework,PATH,true);PathChildrenCacheListener(){publicvoidchildEvent(CuratorFramework ,PathChildrenCacheEventevent)throw ception{("event:} //publicvoidaddListener(PathChildrenCacheListenerAddachangelistener‐the}21treeTreeCache使用一個(gè) 了映射。所以TreeCache可以當(dāng)前節(jié)點(diǎn)下所有節(jié)點(diǎn)的。11456publicStringpath,booleancacheData)‐path‐pathto,237cacheData‐iftrue,nodecontentsarecachedinadditiontothepublicclassTreeCacheTest3 publicstaticfinalString5678publicvoidtestTreeCache()throwceptionTreeCachetreeCache=newTreeCache(curatorFramework,publicvoidchildEvent(CuratorFramework ,TreeCacheEventevent)throwsException{("treecache:}}18publicvoidaddListener(TreeCacheListenerAddachangelistener‐theZookeeper Follower：只能處理讀請求，同時(shí)作為Leader的候選節(jié)點(diǎn)，即如果Leader宕機(jī)，F(xiàn)ollower節(jié)點(diǎn) 1java1java1wgettar‐zxvfcd重命名 cpzoo_sample.cfg1#1#vim23server.1=:2001:3001:participantparticipant可以不用寫，默認(rèn)就是part 己對外提供服務(wù)的起始狀態(tài)。Leader服務(wù)器允許Follower再initLimit時(shí)間內(nèi)完成這個(gè)工與dataDir：顧名思義就是Zookeeper保存數(shù)據(jù)的 cpconf/zoo1.cfgcp cpconf/zoo1.cfgcpconf/zoo1.cfgcpconf/zoo1.cfg4vim vimvim vim Step5：標(biāo)識Server myidserverid，即cdcdvim1cdvim2cdvim3cdvim4////////////binzkServer.shstartconfbinzkServer.shstartconfbinzkServer.shstartconf4Step7：檢測集群狀態(tài)，也可以直接用命令zkServer.sh statusconf/zoo1.cfg進(jìn)行每臺服務(wù)1serverzookeeper/configZookeeper3.5.0Zookeeper3.5.0以前，Zookeeper集群角色要發(fā)生改變的話，只能通過停掉所有的應(yīng)對高可用需求，Zookeeper3.5.0提供了支持動態(tài)擴(kuò)容/縮容的新特性。但是通過客戶端API可以變更服務(wù)端集群狀態(tài)是件很的事情，所以在zookeeper3.5.3版本要用動態(tài)配置，需要開啟超級管理員驗(yàn)證模式ACLs。如果是在一個(gè)安全的環(huán)境也可以通過配置系統(tǒng)參數(shù)-echo‐ng