分佈式管理員zookeeper

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"晚上好，我是盧卡，發現很多技術學的不太紮實，比如經典的zookeeper，Apache zookeeper開源的分佈式協調通知，對於電商系統，高併發，以及微服務的系統服務適用信息都比較高。重新再學習了一下，總結下自己的知識點，對知識點提煉和自己的理解很有幫助，技術人生從重學技術點開始。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/9d/9de77b5af1a851581df9530929d0d2a8.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"zookeeper的定位","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Zookeeper 是apache Zookeeper開源的項目,","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"Zookeeper的主要作用","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"對於微服務分佈式的協調管理,元數據的管理,Master選舉, 實際使用場景下,主要是三類項目","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"1.zookeeper的適用方向","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"① 後端以Java爲主的電商類管理系統","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Zookeeper主要作用於 分佈式鎖,註冊中心 ,這裏Zookeeper主要保持CAP理論中的CP(一致性和分區容錯性),也就是說Zookeeper保證的是強一致性","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"②對於大數據存儲,數據管理,主從數據切換等系統","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這裏我們熟知的Kafla,canal 等在使用Zookeeper的元數據管理 ,通過master選舉出具體的主從,ZAB原子一致性協議,","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"③大型互聯網公司,自研的分佈式組件","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一般以zookeeper爲原型, zookeeper已經被大規模的工業級適用於主要的分佈式系統,做元數據管理,以及註冊中心,一般使用於 dubbo+Zookeeper 做一個基本的微服務架構來實現基本的數據調用","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"小總結:","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"zookeeper 分佈式協調系統, 封裝了分佈式架構中所有的核心和主流的需求和功能;","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"分佈式鎖, 元數據管理, master選舉, 分佈式協調和通知","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"zookeeper的基本特性：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 在瞭解zookeeper的基本架構之前，我們來了解一下，zookeeper爲什麼可以實現分佈式系統的協調通知， 元數據管理等；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 熟悉zookeeper技術棧的都比較瞭解，它本身是處於通知協調機制，數據同步方面有很強的處理能力，這也就是爲什麼很多自研的框架，底層都用zookeeper的原因，好了，廢話不多說，我們開幹；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"背景：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"集羣環境下，zookeeper集羣搭建了3臺物理機器；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"1.順序一致性","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"java"},"content":[{"type":"text","text":" 多個請求，不管是讀取數據的，還是說寫入數據的，都要按照順序執行；\n zookeeper內部要處理的請求，leader角色會給每一個請求分配一個全局唯一且自增的ZxID，用於請求數據\n \n //這個唯一的全局ID就保證了順序一致性\n這裏的zxid，先簡單瞭解，作爲一個請求的排列序號，後面會重點講解的\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"2.原子性","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"java"},"content":[{"type":"text","text":"要麼全部機器都成功，要麼全部機器都不成功，在同步數據操作時\n //保證每次同步操作結果都是一致的\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"3.高可用性","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"java"},"content":[{"type":"text","text":"集羣化部署，避免宕機，\n\n如果宕機，也可自己選舉leader，重新提供服務\n //從恢復模式到消息廣播模式\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"4.數據一致性：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"java"},"content":[{"type":"text","text":" 集羣情況下，每臺機器上的數據都是一致的，但是如果說，在數據同步的時候，宕機了，\n 其他機器會自動將數據保存到磁盤日誌中，等到leader角色提交之後，纔會同步到znode節點中，真正實現數據在每臺機器都是一致的\n \n //這裏說的數據一致性，是強一致性，也是zookeeper的核心的特性，根據分佈式系統保持的CAP理論，只能支持CP,\n //其中C-consistency ，一致性對於，數據的強一致性，p-爲系統的分區容錯性 A-高可用性，其eureka就是支持AP特性\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"5.高性能性","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"java"},"content":[{"type":"text","text":"高性能，這裏提出一個zookeeper的數據模型，znode節點\n//znode節點，基於純內存的，速度很快，樹形結構，類似於Linux的文件系統，每個節點都有數據\n \n 存儲的數據，已經可以處理的能力也就比較強，後面會仔細講的，接着看下去吧\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"小總結：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"基於zookeeper的集羣架構的特點圖：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/43/43cac96288517b5848106972357feb65.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"zookeeper的集羣化：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"背景：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"和上述的場景一致，都是3臺機器，自動化部署的zookeeper集羣，現在我們來具體拆分下每個機器的主要功能","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"Zookeeper的同步數據，是經過主leader提交commit然後同步到各個follow角色","attrs":{}},{"type":"text","text":" ：2PC提交事務--留個念想，後面細聊","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"1.Zookeeper集羣化角色劃分：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Leader（領導）","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"集羣啓動後，根據半數選舉成功的機器，成爲Leader角色，支持讀寫，主要用於寫操作，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Follower（候選人）","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"剩餘的選舉不成功，爲follower角色，支持讀取數據和同步數據，（leader宕機，提供不了寫數據操作時，其他Follower中選舉出新的L earner來實現寫數據的操作）","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Observer","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"只是支持讀取數據，不參與leader的競爭","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這裏面角色都是自身理解，也可以說成模式等等；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當每個角色確定功能之後，我們可以將數據直接開始同步，提供服務；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/13/134cca7af0ad695ce58f47cfc237f9aa.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"1.1注意事項：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 當一個寫操作的請求直接訪問follower角色的zookeeper的進程時，follower角色處理不了寫操作，因爲自身沒有權限，只有Leader角色處理寫操作的請求，這樣的話，follower角色就會轉發請求到Leader角色處理此請求，處理之後，開始數據同步，然後返回到其他follower角色，完成數據同步，阿虎；","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"2.Zookeeper集羣與客戶端訪問的過程：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"背景：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ Zookeeper集羣化，處理請求，可能來源於消息隊列，或者是單獨的發生請求讀取數據，等等；那麼每個請求到底是怎麼連接到Zookeeper集羣建立連接的呢，這節我們來具體聊聊？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":"當Zookeeper的集羣環境下啓動成功，然後我們開始分配角色，然後我們集羣於客戶端開始建立TCP的長連接，\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當建立連接後，就會產生一個服務端的session ，形成一個會話的機制，相應的也會有sessionTimeout的存在","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果當前環境下，突然連接中斷了，在sessionTimeout時間內連接成功不會影響此次長連接的使用；","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"3.Zookeeper集羣化的數據一致性是怎麼保證的？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"背景：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 上述說到Zookeeper支持CAP理論中的CP ,主要是consistency，一致性； 他保證數據強一致性，就很有自己的架構特點，我們來一步步揭開Zookeeper怎麼保證強一致性的面紗？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"Zookeeper的強一致性是通過: ZAB(Zookeeper Atomic Broadcast) 協議： 原子廣播協議","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這個協議是Zookeeper的數據操作的核心，也用於元數據管理的關鍵；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"協議來進行Zookeeper集羣之間的數據同步，保證數據的強一致性；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"ZAB： zookeeper Atomic broadcast 協議的思想：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"劃分角色： leader和follower","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"發送一個Proposal（提議），leader給所有的follower都同步一個proposal（提議）如果半數以上的follower，都收到事務的proposal提議，每個follower都會返回一個ack","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"每個提議Pproposal ->先不會寫入follower中znode數據中，而是會往一個自己本地磁盤日誌文件中寫入，然後返回給leader一個ackleader會發送一個commit提交事務","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"半數提交：2PC也就是兩個階段提交；leader如果異常宕機，會從follower中重新選舉；","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"4.Zookeeper集羣化數據同步的過程：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"劃分的比較細，堅持看完一定會有收穫的⛽️","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1⃣️當集羣啓動的時候，選舉算法開始進行在機器中分配leader和follower角色，通過半數選舉機制成功選到leader角色的機器","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2⃣️剩下的當然就是follower角色的機器了，然後可以向外提供服務了","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3⃣️當一個（寫操作）請求經過Zookeeper集羣后，leader角色機器會優先分配一個zxid用於創建節點或者變更節點的全局唯一自增ID，然後將發起一個提議Proposal（只是一個提議，相當於告訴其他人來活了，準備一下），將提議都放入，之前給每個follow角色準備好的隊列中，（這裏到可以保證順序一致性）。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4⃣️每個follower角色的機器，拿到提議proposal,然後將數據放入自身的磁盤日誌文件中，（不會放入znode節點），然後給leader節點回復一個ACK(確定連接成功，返回的確定字符)","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"java"},"content":[{"type":"text","text":"ACK (Acknowledge character）即是確認字符:\n \n//在數據通信中，接收站發給發送站的一種傳輸類控制字符。表示發來的數據已確認接收無誤。\n \n在TCP/IP協議中，如果接收方成功的接收到數據，那麼會回覆一個ACK數據。通常ACK信號有自己固定的格式,長度大小,由接收方回覆給發送方。\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"5⃣️當leader角色的機器，拿到半數以上的follower角色機器返回的ACK之後,代表已經準備成功，leader角色會推送一個commit消息，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"其他follower就會將數據從磁盤文件寫入，自身進行znode節點中存儲起來，提交之後，數據同步完成，也就可以讀取數據了","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"因爲這個過程，分了兩階段提交，又稱爲2PC事務，用於解決分佈式事務的方法","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"5.Znode的數據模型：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1.zookeeper的數據模型是，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"基於純內存的樹形結構: Znode","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"text"},"content":[{"type":"text","text":" znode分類：\n\n​ 持久節點：客戶端與zk斷開連接，節點依舊存在\n​ 臨時節點：客戶端與zk斷開連接，節點消失\n​ 順序節點： 對於創建節點的時候，自增加全局遞增的序號；\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"zk中有一個基於分佈式鎖的要求環境，curator框架，我們開展臨時節點來實現的，加鎖的時候創建一個順序節點","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2.zk節點的用途","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"zookeeperk做元數據管理的時候： 肯定是持久節點但是一般做分佈式鎖，分佈式協調和通知，都是臨時節點，如果斷開，臨時節點消失，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3 .znode的節點的組成部分：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"6.zookeeper啓動集羣到數據同步過程","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"leader選舉","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"集羣啓動，開始進行leader選舉，半數機器認可當前機器作爲leader，各個follower開始進行數據同步，完成就可以退出恢復模式，然後可以的對外提供服務了","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"宕機重新選舉leader","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3臺機器，允許不超過一半的機器宕機","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2臺機器， 兩個機器都同意某臺機器作爲leader ，這個就可以選舉出leader;","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1臺機器是沒法自己選舉自己的；因爲1臺機器，小於半數，所有不能啓動集羣環境了；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"數據同步","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"leader選舉出來後，其他機器都是follower--開始數據同步","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"強制剩下的follower 數據與自己leader一致；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"數據同步完成之後，就會進去， 消息廣播模式；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"消息寫入： leader寫入，採用2PC模式過半寫機制，給follower進行同步","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"將自己數據更新到，znode數據節點中","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"宕機修復","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"leader宕機，或者follower宕機，只要存活的機器超過一半，那就可以重新選舉leader選舉leader，要求有一半以上的支持 ，其他跟follower數據同步，消息廣播模式","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"zk從恢復模式到消息廣播模式開始同步數據","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/48/485438bc2ab7e8a222f318a8a503103e.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"7.談談在zookeeper集羣下數據一致性的理解？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 在數據同步的過程中，leader 將提議proposal放入隊列（先進先出），然後開始同步到follower，當過半的follower返回ACK(確認字符)之後，leader直接推送一個commit用於提交，follower同步數據；（這裏我們注意，不是全部的follower返回結果）","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"zk的數據同步不是強一致性，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當follower將磁盤日誌文件中的數據，提交到znode之後，數據纔可以被讀取到，最終數據會一致的，但是","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"zk官方給的一個回覆是，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"順序一致性","attrs":{}},{"type":"text","text":"，會根據zxid，以及提議proposal的順序保證","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"因爲leader一定會保證所有的proposal同步到follower上，是按照順序，最終實現順序一致性","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"zk也可以支持強一致行，但是需要手動調節zk的sync()操作","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"8.ZAB協議下，數據不一致的情況有哪些？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"情況1：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當leader推送一個commit提交數據，剛自己提交了，但是他還沒有吧commit提交給follower的時候，就已經掛掉了？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"簡介： 當客戶端發送一個寫操作的請求，leader已經收到半數以上follower發來的ack，他自己本地已經將數據寫入znode中，leader自己commit成功，但是在沒有給別的follower發出commit之前就已經掛了， 客戶端在收到，leader已經commit數據之後，就默認已經將數據更新完成，但是我們新請求，查詢數據follower機器的時候，發現沒有，與之間leader返回的不一樣；（導致了數據不一致）","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這時，follower中的數據和剛剛宕機的leader機器上的數據肯定是不一致的，接下來zk會怎麼做呢？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在具體的時間中，zk集羣中的follower角色發現，老大leader無法回覆，處於失聯，宕機狀態，他們就說我們再選一個leader，然後就再重新選一個leader01（新的），那之前已經掛掉的leader，就順勢變成了follower，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"情況2：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果客戶端在，請求寫數據操作的leader機器上，然後leader，發送一個proposal提議，但是還沒發出去，就掛了；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"導致本地磁盤日誌文件中存在一個proposal；但是其他的follower中沒有這個數據；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當集羣奔潰之後，開展恢復模式，ZAB協議的關鍵核心就顯示出來了，根據","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"情況1的解決思路：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"衆多的follower,開始半數選舉機制，選出新的leader之後，發現本地磁盤上有沒有提交的proposal，然後查看別的follower也存在這樣 的情況，這裏的新leader(是之前未接收到commit消息的follower),然後我們開始發送commit 給其他follower，將數據寫入znode節點中 解決客戶端讀取數據不一致的情況；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"情況2的解決過程：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"根據上述的情況，已經宕機的leader存在一個proposal的提議，但是其他的follow沒有收到，所以在恢復模式之後，新的leader被選擇出來，開展寫操作，但是發優先去查詢一下本地磁盤中是否有之前遺留的proposal提議，查詢到一個follower(之前宕機的發送的一個proposal的leader)磁盤數據與其他的不一致，然後現在的新leader將會同步數據給其他的follower，之前宕機的leader存在一個的提議（proposal提議）會被捨棄掉；","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"9.奔潰恢復時選舉出來的新leader如何跟其他follower進行同步數據的過程","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我們還是來模擬一個場景，5臺機器的zookeeper集羣化部署","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"其中四臺是follower角色用於讀取數據和同步數據的，剩下的一臺就是leader專門進行寫操作的機器","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"背景：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 當一個leader發出一個proposal提議後，其他三臺的follower機器都收到了，唯獨剩下的一個沒收到proposal，但是現在已經有三臺j follower機器已經返回了ack的確認字符， 所以leader判斷已經都通知成功，（半數選舉），然後自身開始commit，沒等到給剩下的follower 角色進行提交就宕機了；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"分析一下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 目前leader已經commit數據到自己的znode節點上了，但是已經掛了，但是其他四臺follow機器，三臺已經將proposal放到自己本地的磁盤日誌文件中，剩餘一臺啥都沒有（之前半數返回ack已經夠了沒有收到proposal提議），萬一其他三臺機器恢復機制時，都選舉最後一臺機器，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"作爲新的leader,那這條數據就永久丟失了","attrs":{}},{"type":"text","text":"，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"先告訴大家一個答案，就是這中事情不會發生，zookeeper集羣不會選擇一個沒有數據的角色選舉成leader","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"主要是要了解一個","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"zxid","attrs":{}},{"type":"text","text":"的概念：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Zxid： 可以理解爲一個事物的自增ID，要是對提議改變了，他就自增，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"要是這個follow沒有收到提議，也就zxid相應的比收到返回ack的follower 節點的機器肯定是大的；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"選舉leader的要求是，在follower中，收到事務zxid最大的，也就是相當於創建時間最早的，作爲新的leader","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"10.zxid","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"znode節點的狀態信息中包含czxid, 那麼什麼是zxid呢?","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ZooKeeper狀態的每一次改變, 都對應着一個遞增的Transaction id, 該id稱爲zxid. 由於zxid的遞增性質, 如果zxid1小於zxid2, 那麼zxid1肯定先於zxid2發生.","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"創建任意節點, 或者更新任意節點的數據, 或者刪除任意節點, 都會導致Zookeeper狀態發生改變, 從而導致zxid的值增加.","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ ZooKeeper中ZXID是一個長度64位的數字，其中低32位是按照數字遞增，即每次客戶端發起一個proposal，低32位的數字簡單加1。高32位是leader週期的epoch編號。也就是","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"leader的一個版本號","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"比如說後面epoch編號爲8， 也就說當前的leader版本是8 ，掛掉之後重新選舉就自增1，epoch的編號就是9","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所以睡要是想之前的","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"情況2，發生","attrs":{}},{"type":"text","text":"就知道是爲什麼會出來丟掉的數據，因爲會找leader的epocn最大的一個版本作爲可以提交的領導者","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"重新發號命令；","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"11.Observer節點（角色）的作用：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"需要配置文件，只會單純的接受和同步數據，observer不會參與過半寫機制，不參與選舉","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"只是被動的接受數據，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果讀的請求需求很大，我們可以添加幾個Observer,","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"優點：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 可一扛下很大的併發量，","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 不會影響過半寫的機制，","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 不會特別影響性能","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"原因：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"要是都是follower節點的話，當leader要發proposal提議，要等待過半的機器返回ack，然後同步數據，也是過半的，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"選舉的時候也是過半的，這樣會很耗費時間以及網絡資源，但是隻要讀取的數據，不參與選舉的observer就很大程度上解決了這個痛點","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"12.zookeeper適用於實際部署場景：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​ 適用於小集羣的部署，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"讀多寫少","attrs":{}},{"type":"text","text":"的場景；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"奇數臺機器，半數選舉，","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"過半寫+磁盤日誌文件====》commit+znode 節點","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"主要是寫入壓力過大leader","attrs":{}}]}]}