把Redis當作隊列來用，真的合適嗎？

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"大家好，我是 Kaito。","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":"我經常聽到很多人討論，關於「把 Redis 當作隊列來用是否合適」的問題。","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":"有些人表示贊成，他們認爲 Redis 很輕量，用作隊列很方便。","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":"也些人則反對，認爲 Redis 會「丟」數據，最好還是用「專業」的隊列中間件更穩妥。","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":"究竟哪種方案更好呢？","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":"這篇文章，我就和你聊一聊把 Redis 當作隊列，究竟是否合適這個問題。","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":"我會從簡單到複雜，一步步帶你梳理其中的細節，把這個問題真正的講清楚。","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":"看完這篇文章後，我希望你對這個問題你會有全新的認識。","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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"在文章的最後，我還會告訴你關於「技術選型」的思路，文章有點長，希望你可以耐心讀完。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/c1/c10790efd66d6ccb755025d68b8bd790.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"從最簡單的開始：List 隊列","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"首先，我們先從最簡單的場景開始講起。","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":"如果你的業務需求足夠簡單，想把 Redis 當作隊列來使用，肯定最先想到的就是使用 List 這個數據類型。","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 List 底層的實現就是一個「鏈表」，在頭部和尾部操作元素，時間複雜度都是 O(1)，這意味着它非常符合消息隊列的模型。","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":"如果把 List 當作隊列，你可以這麼來用。","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":"生產者使用 LPUSH 發佈消息：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"127.0.0.1:6379> LPUSH queue msg1\n(integer) 1\n127.0.0.1:6379> LPUSH queue msg2\n(integer) 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":"消費者這一側，使用 RPOP 拉取消息：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"127.0.0.1:6379> RPOP queue\n\"msg1\"\n127.0.0.1:6379> RPOP queue\n\"msg2\"","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":"這個模型非常簡單，也很容易理解。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/29/294d065a895631616dadc37b0bc4ab2b.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"但這裏有個小問題，當隊列中已經沒有消息了，消費者在執行 RPOP 時，會返回 NULL。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"127.0.0.1:6379> RPOP queue\n(nil) // 沒消息了","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":"而我們在編寫消費者邏輯時，一般是一個「死循環」，這個邏輯需要不斷地從隊列中拉取消息進行處理，僞代碼一般會這麼寫：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"python"},"content":[{"type":"text","text":"while true:\n msg = redis.rpop(\"queue\")\n // 沒有消息，繼續循環\n if msg == null:\n continue\n // 處理消息\n handle(msg)","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":"如果此時隊列爲空，那消費者依舊會頻繁拉取消息，這會造成「CPU 空轉」，不僅浪費 CPU 資源，還會對 Redis 造成壓力。","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":"怎麼解決這個問題呢？","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":"也很簡單，當隊列爲空時，我們可以「休眠」一會，再去嘗試拉取消息。代碼可以修改成這樣：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"python"},"content":[{"type":"text","text":"while true:\n msg = redis.rpop(\"queue\")\n // 沒有消息，休眠2s\n if msg == null:\n sleep(2)\n continue\n // 處理消息 \n handle(msg)","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":"這就解決了 CPU 空轉問題。","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":"這個問題雖然解決了，但又帶來另外一個問題：當消費者在休眠等待時，有新消息來了，那消費者處理新消息就會存在「延遲」。","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":"假設設置的休眠時間是 2s，那新消息最多存在 2s 的延遲。","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":"要想縮短這個延遲，只能減小休眠的時間。但休眠時間越小，又有可能引發 CPU 空轉問題。","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":"魚和熊掌不可兼得。","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":"那如何做，既能及時處理新消息，還能避免 CPU 空轉呢？","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":"Redis 是否存在這樣一種機制：如果隊列爲空，消費者在拉取消息時就「阻塞等待」，一旦有新消息過來，就通知我的消費者立即處理新消息呢？","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":"幸運的是，Redis 確實提供了「阻塞式」拉取消息的命令：BRPOP / BLPOP，這裏的 B 指的是阻塞（Block）。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/5d/5de07e3ca6748a57b808ad176406c572.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"現在，你可以這樣來拉取消息了：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"python"},"content":[{"type":"text","text":"while true:\n // 沒消息阻塞等待，0表示不設置超時時間\n msg = redis.brpop(\"queue\", 0)\n if msg == null:\n continue\n // 處理消息\n handle(msg)","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":"使用 BRPOP 這種阻塞式方式拉取消息時，還支持傳入一個「超時時間」，如果設置爲 0，則表示不設置超時，直到有新消息才返回，否則會在指定的超時時間後返回 NULL。","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":"這個方案不錯，既兼顧了效率，還避免了 CPU 空轉問題，一舉兩得。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"注意：如果設置的超時時間太長，這個連接太久沒有活躍過，可能會被 Redis Server 判定爲無效連接，之後 Redis Server 會強制把這個客戶端踢下線。所以，採用這種方案，客戶端要有重連機制。","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":"解決了消息處理不及時的問題，你可以再思考一下，這種隊列模型，有什麼缺點？","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":"我們一起來分析一下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"不支持重複消費","attrs":{}},{"type":"text","text":"：消費者拉取消息後，這條消息就從 List 中刪除了，無法被其它消費者再次消費，即不支持多個消費者消費同一批數據","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"消息丟失","attrs":{}},{"type":"text","text":"：消費者拉取到消息後，如果發生異常宕機，那這條消息就丟失了","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":"第一個問題是功能上的，使用 List 做消息隊列，它僅僅支持最簡單的，一組生產者對應一組消費者，不能滿足多組生產者和消費者的業務場景。","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":"第二個問題就比較棘手了，因爲從 List 中 POP 一條消息出來後，這條消息就會立即從鏈表中刪除了。也就是說，無論消費者是否處理成功，這條消息都沒辦法再次消費了。","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":"這也意味着，如果消費者在處理消息時異常宕機，那這條消息就相當於丟失了。","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":"針對這 2 個問題怎麼解決呢？我們一個個來看。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"發佈/訂閱模型：Pub/Sub","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"從名字就能看出來，這個模塊是 Redis 專門是針對「發佈/訂閱」這種隊列模型設計的。","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":"它正好可以解決前面提到的第一個問題：重複消費。","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":"即多組生產者、消費者的場景，我們來看它是如何做的。","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":"Redis 提供了 PUBLISH / SUBSCRIBE 命令，來完成發佈、訂閱的操作。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/96/9617d977ee7075c3b20eee6532a878e5.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"假設你想開啓 2 個消費者，同時消費同一批數據，就可以按照以下方式來實現。","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":"首先，使用 SUBSCRIBE 命令，啓動 2 個消費者，並「訂閱」同一個隊列。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// 2個消費者 都訂閱一個隊列\n127.0.0.1:6379> SUBSCRIBE queue\nReading messages... (press Ctrl-C to quit)\n1) \"subscribe\"\n2) \"queue\"\n3) (integer) 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":"此時，2 個消費者都會被阻塞住，等待新消息的到來。","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":"之後，再啓動一個生產者，發佈一條消息。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"127.0.0.1:6379> PUBLISH queue msg1\n(integer) 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":"這時，2 個消費者就會解除阻塞，收到生產者發來的新消息。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"127.0.0.1:6379> SUBSCRIBE queue\n// 收到新消息\n1) \"message\"\n2) \"queue\"\n3) \"msg1\"","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":"看到了麼，使用 Pub/Sub 這種方案，既支持阻塞式拉取消息，還很好地滿足了多組消費者，消費同一批數據的業務需求。","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":"除此之外，Pub/Sub 還提供了「匹配訂閱」模式，允許消費者根據一定規則，訂閱「多個」自己感興趣的隊列。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// 訂閱符合規則的隊列\n127.0.0.1:6379> PSUBSCRIBE queue.*\nReading messages... (press Ctrl-C to quit)\n1) \"psubscribe\"\n2) \"queue.*\"\n3) (integer) 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":"這裏的消費者，訂閱了 queue.* 相關的隊列消息。","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":"之後，生產者分別向 queue.p1 和 queue.p2 發佈消息。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"127.0.0.1:6379> PUBLISH queue.p1 msg1\n(integer) 1\n127.0.0.1:6379> PUBLISH queue.p2 msg2\n(integer) 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":"這時再看消費者，它就可以接收到這 2 個生產者的消息了。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"127.0.0.1:6379> PSUBSCRIBE queue.*\nReading messages... (press Ctrl-C to quit)\n...\n// 來自queue.p1的消息\n1) \"pmessage\"\n2) \"queue.*\"\n3) \"queue.p1\"\n4) \"msg1\"\n\n// 來自queue.p2的消息\n1) \"pmessage\"\n2) \"queue.*\"\n3) \"queue.p2\"\n4) \"msg2\"","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/cf/cf0aa1483deebd5a3ef28c408a8adb1e.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"我們可以看到，Pub/Sub 最大的優勢就是，支持多組生產者、消費者處理消息。","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":"講完了它的優點，那它有什麼缺點呢？","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":"其實，Pub/Sub 最大問題是：","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"丟數據","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":"如果發生以下場景，就有可能導致數據丟失：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"消費者下線","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"Redis 宕機","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"消息堆積","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":"究竟是怎麼回事？","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":"這其實與 Pub/Sub 的實現方式有很大關係。","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":"Pub/Sub 在實現時非常簡單，它沒有基於任何數據類型，也沒有做任何的數據存儲，它只是單純地爲生產者、消費者建立「數據轉發通道」，把符合規則的數據，從一端轉發到另一端。","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":"一個完整的發佈、訂閱消息處理流程是這樣的：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"消費者訂閱指定隊列，Redis 就會記錄一個映射關係：隊列->消費者","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"生產者向這個隊列發佈消息，那 Redis 就從映射關係中找出對應的消費者，把消息轉發給它","attrs":{}}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/a3/a3e030071be0440cde1f874988055d5f.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"看到了麼，整個過程中，沒有任何的數據存儲，一切都是實時轉發的。","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":"這種設計方案，就導致了上面提到的那些問題。","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":"例如，如果一個消費者異常掛掉了，它再重新上線後，只能接收新的消息，在下線期間生產者發佈的消息，因爲找不到消費者，都會被丟棄掉。","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":"如果所有消費者都下線了，那生產者發佈的消息，因爲找不到任何一個消費者，也會全部「丟棄」。","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，當你在使用 Pub/Sub 時，一定要注意：","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"消費者必須先訂閱隊列，生產者才能發佈消息，否則消息會丟失。","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":"這也是前面講例子時，我們讓消費者先訂閱隊列，之後才讓生產者發佈消息的原因。","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":"另外，因爲 Pub/Sub 沒有基於任何數據類型實現，所以它也不具備「數據持久化」的能力。","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":"也就是說，Pub/Sub 的相關操作，不會寫入到 RDB 和 AOF 中，當 Redis 宕機重啓，Pub/Sub 的數據也會全部丟失。","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":"最後，我們來看 Pub/Sub 在處理「消息積壓」時，爲什麼也會丟數據？","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":"當消費者的速度，跟不上生產者時，就會導致數據積壓的情況發生。","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":"如果採用 List 當作隊列，消息積壓時，會導致這個鏈表很長，最直接的影響就是，Redis 內存會持續增長，直到消費者把所有數據都從鏈表中取出。","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":"但 Pub/Sub 的處理方式卻不一樣，當消息積壓時，有可能會導致","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"消費失敗和消息丟失","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":"這是怎麼回事？","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":"還是回到 Pub/Sub 的實現細節上來說。","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":"每個消費者訂閱一個隊列時，Redis 都會在 Server 上給這個消費者在分配一個「緩衝區」，這個緩衝區其實就是一塊內存。","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":"當生產者發佈消息時，Redis 先把消息寫到對應消費者的緩衝區中。","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":"之後，消費者不斷地從緩衝區讀取消息，處理消息。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/37/3725adceae0959149b1dd1d8b0bad00c.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"但是，問題就出在這個緩衝區上。","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":"因爲這個緩衝區其實是有「上限」的（可配置），如果消費者拉取消息很慢，就會造成生產者發佈到緩衝區的消息開始積壓，緩衝區內存持續增長。","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":"如果超過了緩衝區配置的上限，此時，Redis 就會「強制」把這個消費者踢下線。","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":"這時消費者就會消費失敗，也會丟失數據。","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":"如果你有看過 Redis 的配置文件，可以看到這個緩衝區的默認配置：client-output-buffer-limit pubsub 32mb 8mb 60。","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":"它的參數含義如下：","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":"32mb：緩衝區一旦超過 32MB，Redis 直接強制把消費者踢下線","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"8mb + 60：緩衝區超過 8MB，並且持續 60 秒，Redis 也會把消費者踢下線","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":"Pub/Sub 的這一點特點，是與 List 作隊列差異比較大的。","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","text":"從這裏你應該可以看出，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"List 其實是屬於「拉」模型，而 Pub/Sub 其實屬於「推」模型","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":"List 中的數據可以一直積壓在內存中，消費者什麼時候來「拉」都可以。","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":"但 Pub/Sub 是把消息先「推」到消費者在 Redis Server 上的緩衝區中，然後等消費者再來取。","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":"當生產、消費速度不匹配時，就會導致緩衝區的內存開始膨脹，Redis 爲了控制緩衝區的上限，所以就有了上面講到的，強制把消費者踢下線的機制。","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":"好了，現在我們總結一下 Pub/Sub 的優缺點：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"支持發佈 / 訂閱，支持多組生產者、消費者處理消息","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"消費者下線，數據會丟失","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"不支持數據持久化，Redis 宕機，數據也會丟失","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","text":"消息堆積，緩衝區溢出，消費者會被強制踢下線，數據也會丟失","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":"有沒有發現，除了第一個是優點之外，剩下的都是缺點。","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":"所以，很多人看到 Pub/Sub 的特點後，覺得這個功能很「雞肋」。","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":"也正是以上原因，Pub/Sub 在實際的應用場景中用得並不多。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"目前只有哨兵集羣和 Redis 實例通信時，採用了 Pub/Sub 的方案，因爲哨兵正好符合即時通訊的業務場景。","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":"我們再來看一下，Pub/Sub 有沒有解決，消息處理時異常宕機，無法再次消費的問題呢？","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":"其實也不行，Pub/Sub 從緩衝區取走數據之後，數據就從 Redis 緩衝區刪除了，消費者發生異常，自然也無法再次重新消費。","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":"好，現在我們重新梳理一下，我們在使用消息隊列時的需求。","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":"當我們在使用一個消息隊列時，希望它的功能如下：","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":"支持阻塞等待拉取消息","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"支持發佈 / 訂閱模式","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"消費失敗，可重新消費，消息不丟失","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"實例宕機，消息不丟失，數據可持久化","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"消息可堆積","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":"Redis 除了 List 和 Pub/Sub 之外，還有符合這些要求的數據類型嗎？","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":"其實，Redis 的作者也看到了以上這些問題，也一直在朝着這些方向努力着。","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":"Redis 作者在開發 Redis 期間，還另外開發了一個開源項目 disque。","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":"這個項目的定位，就是一個基於內存的分佈式消息隊列中間件。","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":"但由於種種原因，這個項目一直不溫不火。","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":"終於，在 Redis 5.0 版本，作者把 disque 功能移植到了 Redis 中，並給它定義了一個新的數據類型：","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"Stream","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":"下面我們就來看看，它能符合上面提到的這些要求嗎？","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"趨於成熟的隊列：Stream","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我們來看 Stream 是如何解決上面這些問題的。","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":"我們依舊從簡單到複雜，依次來看 Stream 在做消息隊列時，是如何處理的？","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":"首先，Stream 通過 XADD 和 XREAD 完成最簡單的生產、消費模型：","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":"XADD：發佈消息","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"XREAD：讀取消息","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":"生產者發佈 2 條消息：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// *表示讓Redis自動生成消息ID\n127.0.0.1:6379> XADD queue * name zhangsan\n\"1618469123380-0\"\n127.0.0.1:6379> XADD queue * name lisi\n\"1618469127777-0\"","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":"使用 XADD 命令發佈消息，其中的「*」表示讓 Redis 自動生成唯一的消息 ID。","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":"這個消息 ID 的格式是「時間戳-自增序號」。","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":"消費者拉取消息：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// 從開頭讀取5條消息，0-0表示從開頭讀取\n127.0.0.1:6379> XREAD COUNT 5 STREAMS queue 0-0\n1) 1) \"queue\"\n 2) 1) 1) \"1618469123380-0\"\n 2) 1) \"name\"\n 2) \"zhangsan\"\n 2) 1) \"1618469127777-0\"\n 2) 1) \"name\"\n 2) \"lisi\"","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":"如果想繼續拉取消息，需要傳入上一條消息的 ID：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"127.0.0.1:6379> XREAD COUNT 5 STREAMS queue 1618469127777-0\n(nil)","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":"沒有消息，Redis 會返回 NULL。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/0c/0ce3c0db96015e85a93fcda0b227e7f7.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"以上就是 Stream 最簡單的生產、消費。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這裏不再重點介紹 Stream 命令的各種參數，我在例子中演示時，凡是大寫的單詞都是「固定」參數，凡是小寫的單詞，都是可以自己定義的，例如隊列名、消息長度等等，下面的例子規則也是一樣，爲了方便你理解，這裏有必要提醒一下。","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":"下面我們來看，針對前面提到的消息隊列要求，Stream 都是如何解決的？","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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"1) Stream 是否支持「阻塞式」拉取消息？","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":"可以的，在讀取消息時，只需要增加 BLOCK 參數即可。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// BLOCK 0 表示阻塞等待，不設置超時時間\n127.0.0.1:6379> XREAD COUNT 5 BLOCK 0 STREAMS queue 1618469127777-0","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，消費者就會阻塞等待，直到生產者發佈新的消息纔會返回。","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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"2) Stream 是否支持發佈 / 訂閱模式？","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":"也沒問題，Stream 通過以下命令完成發佈訂閱：","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":"XGROUP：創建消費者組","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"XREADGROUP：在指定消費組下，開啓消費者拉取消息","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":"下面我們來看具體如何做？","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":"首先，生產者依舊發佈 2 條消息：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"127.0.0.1:6379> XADD queue * name zhangsan\n\"1618470740565-0\"\n127.0.0.1:6379> XADD queue * name lisi\n\"1618470743793-0\"","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":"之後，我們想要開啓 2 組消費者處理同一批數據，就需要創建 2 個消費者組：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// 創建消費者組1，0-0表示從頭拉取消息\n127.0.0.1:6379> XGROUP CREATE queue group1 0-0\nOK\n// 創建消費者組2，0-0表示從頭拉取消息\n127.0.0.1:6379> XGROUP CREATE queue group2 0-0\nOK","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":"消費者組創建好之後，我們可以給每個「消費者組」下面掛一個「消費者」，讓它們分別處理同一批數據。","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":"第一個消費組開始消費：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// group1的consumer開始消費，>表示拉取最新數據\n127.0.0.1:6379> XREADGROUP GROUP group1 consumer COUNT 5 STREAMS queue >\n1) 1) \"queue\"\n 2) 1) 1) \"1618470740565-0\"\n 2) 1) \"name\"\n 2) \"zhangsan\"\n 2) 1) \"1618470743793-0\"\n 2) 1) \"name\"\n 2) \"lisi\"","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":"同樣地，第二個消費組開始消費：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// group2的consumer開始消費，>表示拉取最新數據\n127.0.0.1:6379> XREADGROUP GROUP group2 consumer COUNT 5 STREAMS queue >\n1) 1) \"queue\"\n 2) 1) 1) \"1618470740565-0\"\n 2) 1) \"name\"\n 2) \"zhangsan\"\n 2) 1) \"1618470743793-0\"\n 2) 1) \"name\"\n 2) \"lisi\"","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":"我們可以看到，這 2 組消費者，都可以獲取同一批數據進行處理了。","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":"這樣一來，就達到了多組消費者「訂閱」消費的目的。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/a2/a2f7f5867e033b31aba9dfb6f4d67fbd.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"3) 消息處理時異常，Stream 能否保證消息不丟失，重新消費？","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":"除了上面拉取消息時用到了消息 ID，這裏爲了保證重新消費，也要用到這個消息 ID。","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":"當一組消費者處理完消息後，需要執行 XACK 命令告知 Redis，這時 Redis 就會把這條消息標記爲「處理完成」。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// group1下的 1618472043089-0 消息已處理完成\n127.0.0.1:6379> XACK queue group1 1618472043089-0","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/7c/7c25d66c322ca6c8887d9e959f8cd864.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"如果消費者異常宕機，肯定不會發送 XACK，那麼 Redis 就會依舊保留這條消息。","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":"待這組消費者重新上線後，Redis 就會把之前沒有處理成功的數據，重新發給這個消費者。這樣一來，即使消費者異常，也不會丟失數據了。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// 消費者重新上線，0-0表示重新拉取未ACK的消息\n127.0.0.1:6379> XREADGROUP GROUP group1 consumer1 COUNT 5 STREAMS queue 0-0\n// 之前沒消費成功的數據，依舊可以重新消費\n1) 1) \"queue\"\n 2) 1) 1) \"1618472043089-0\"\n 2) 1) \"name\"\n 2) \"zhangsan\"\n 2) 1) \"1618472045158-0\"\n 2) 1) \"name\"\n 2) \"lisi\"","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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"4) Stream 數據會寫入到 RDB 和 AOF 做持久化嗎？","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":"Stream 是新增加的數據類型，它與其它數據類型一樣，每個寫操作，也都會寫入到 RDB 和 AOF 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":"我們只需要配置好持久化策略，這樣的話，就算 Redis 宕機重啓，Stream 中的數據也可以從 RDB 或 AOF 中恢復回來。","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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"5) 消息堆積時，Stream 是怎麼處理的？","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":"其實，當消息隊列發生消息堆積時，一般只有 2 個解決方案：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"生產者限流：避免消費者處理不及時，導致持續積壓","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"丟棄消息：中間件丟棄舊消息，只保留固定長度的新消息","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":"而 Redis 在實現 Stream 時，採用了第 2 個方案。","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":"在發佈消息時，你可以指定隊列的最大長度，防止隊列積壓導致內存爆炸。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"// 隊列長度最大10000\n127.0.0.1:6379> XADD queue MAXLEN 10000 * name zhangsan\n\"1618473015018-0\"","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":"當隊列長度超過上限後，舊消息會被刪除，只保留固定長度的新消息。","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":"這麼來看，Stream 在消息積壓時，如果指定了最大長度，還是有可能丟失消息的。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"除了以上介紹到的命令，Stream 還支持查看消息長度（XLEN）、查看消費者狀態（XINFO）等命令，使用也比較簡單，你可以查詢官方文檔瞭解一下，這裏就不過多介紹了。","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":"好了，通過以上介紹，我們可以看到，Redis 的 Stream 幾乎覆蓋到了消息隊列的各種場景，是不是覺得很完美？","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":"既然它的功能這麼強大，這是不是意味着，Redis 真的可以作爲專業的消息隊列中間件來使用呢？","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":"但是還「差一點」，就算 Redis 能做到以上這些，也只是「趨近於」專業的消息隊列。","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":"原因在於 Redis 本身的一些問題，如果把其定位成消息隊列，還是有些欠缺的。","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":"到這裏，就不得不把 Redis 與專業的隊列中間件做對比了。","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":"下面我們就來看一下，Redis 在作隊列時，到底還有哪些欠缺？","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"與專業的消息隊列對比","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"其實，一個專業的消息隊列，必須要做到兩大塊：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"消息不丟","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"消息可堆積","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":"前面我們討論的重點，很大篇幅圍繞的是第一點展開的。","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":"這裏我們換個角度，從一個消息隊列的「使用模型」來分析一下，怎麼做，才能保證數據不丟？","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":"使用一個消息隊列，其實就分爲三大塊：","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"生產者、隊列中間件、消費者","attrs":{}},{"type":"text","text":"。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/50/50add583d6d0258c4316e68446b93619.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"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":"消息是否會發生丟失，其重點也就在於以下 3 個環節：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"生產者會不會丟消息？","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"消費者會不會丟消息？","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"隊列中間件會不會丟消息？","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","marks":[{"type":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"1) 生產者會不會丟消息？","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":"當生產者在發佈消息時，可能發生以下異常情況：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"消息沒發出去：網絡故障或其它問題導致發佈失敗，中間件直接返回失敗","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"不確定是否發佈成功：網絡問題導致發佈超時，可能數據已發送成功，但讀取響應結果超時了","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":"如果是情況 1，消息根本沒發出去，那麼重新發一次就好了。","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":"如果是情況 2，生產者沒辦法知道消息到底有沒有發成功？所以，爲了避免消息丟失，它也只能繼續重試，直到發佈成功爲止。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"生產者一般會設定一個最大重試次數，超過上限依舊失敗，需要記錄日誌報警處理。","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":"也就是說，生產者爲了避免消息丟失，只能採用失敗重試的方式來處理。","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":"但發現沒有？這也意味着消息可能會重複發送。","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":"是的，在使用消息隊列時，要保證消息不丟，寧可重發，也不能丟棄。","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":"那消費者這邊，就需要多做一些邏輯了。","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":"對於敏感業務，當消費者收到重複數據數據時，要設計冪等邏輯，保證業務的正確性。","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":"從這個角度來看，生產者會不會丟消息，取決於生產者對於異常情況的處理是否合理。","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，無論是 Redis 還是專業的隊列中間件，生產者在這一點上都是可以保證消息不丟的。","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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"2) 消費者會不會丟消息？","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":"這種情況就是我們前面提到的，消費者拿到消息後，還沒處理完成，就異常宕機了，那消費者還能否重新消費失敗的消息？","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":"要解決這個問題，消費者在處理完消息後，必須「告知」隊列中間件，隊列中間件纔會把標記已處理，否則仍舊把這些數據發給消費者。","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":"這種方案需要消費者和中間件互相配合，才能保證消費者這一側的消息不丟。","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":"無論是 Redis 的 Stream，還是專業的隊列中間件，例如 RabbitMQ、Kafka，其實都是這麼做的。","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":"所以，從這個角度來看，Redis 也是合格的。","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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"3) 隊列中間件會不會丟消息？","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":"前面 2 個問題都比較好處理，只要客戶端和服務端配合好，就能保證生產端、消費端都不丟消息。","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":"但是，如果隊列中間件本身就不可靠呢？","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":"畢竟生產者和消費這都依賴它，如果它不可靠，那麼生產者和消費者無論怎麼做，都無法保證數據不丟。","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":"在這個方面，Redis 其實沒有達到要求。","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":"Redis 在以下 2 個場景下，都會導致數據丟失。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"AOF 持久化配置爲每秒寫盤，但這個寫盤過程是異步的，Redis 宕機時會存在數據丟失的可能","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"主從複製也是異步的，主從切換時，也存在丟失數據的可能（從庫還未同步完成主庫發來的數據，就被提成主庫）","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":"基於以上原因我們可以看到，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"Redis 本身的無法保證嚴格的數據完整性","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":"所以，如果把 Redis 當做消息隊列，在這方面是有可能導致數據丟失的。","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":"再來看那些專業的消息隊列中間件是如何解決這個問題的？","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":"像 RabbitMQ 或 Kafka 這類專業的隊列中間件，在使用時，一般是部署一個集羣，生產者在發佈消息時，隊列中間件通常會寫「多個節點」，以此保證消息的完整性。這樣一來，即便其中一個節點掛了，也能保證集羣的數據不丟失。","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":"也正因爲如此，RabbitMQ、Kafka在設計時也更復雜。畢竟，它們是專門針對隊列場景設計的。","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":"但 Redis 的定位則不同，它的定位更多是當作緩存來用，它們兩者在這個方面肯定是存在差異的。","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":"最後，我們來看消息積壓怎麼辦？","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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"4) 消息積壓怎麼辦？","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 Redis 的數據都存儲在內存中，這就意味着一旦發生消息積壓，則會導致 Redis 的內存持續增長，如果超過機器內存上限，就會面臨被 OOM 的風險。","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，Redis 的 Stream 提供了可以指定隊列最大長度的功能，就是爲了避免這種情況發生。","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":"但 Kafka、RabbitMQ 這類消息隊列就不一樣了，它們的數據都會存儲在磁盤上，磁盤的成本要比內存小得多，當消息積壓時，無非就是多佔用一些磁盤空間，相比於內存，在面對積壓時也會更加「坦然」。","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":"綜上，我們可以看到，把 Redis 當作隊列來使用時，始終面臨的 2 個問題：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"Redis 本身可能會丟數據","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"面對消息積壓，Redis 內存資源緊張","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":"到這裏，Redis 是否可以用作隊列，我想這個答案你應該會比較清晰了。","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":"如果你的業務場景足夠簡單，對於數據丟失不敏感，而且消息積壓概率比較小的情況下，把 Redis 當作隊列是完全可以的。","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":"而且，Redis 相比於 Kafka、RabbitMQ，部署和運維也更加輕量。","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":"如果你的業務場景對於數據丟失非常敏感，而且寫入量非常大，消息積壓時會佔用很多的機器資源，那麼我建議你使用專業的消息隊列中間件。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"總結","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"好了，總結一下。這篇文章我們從「Redis 能否用作隊列」這個角度出發，介紹了 List、Pub/Sub、Stream 在做隊列的使用方式，以及它們各自的優劣。","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":"之後又把 Redis 和專業的消息隊列中間件做對比，發現 Redis 的不足之處。","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":"最後，我們得出 Redis 做隊列的合適場景。","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":"這裏我也列了一個表格，總結了它們各自的優缺點。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/e0/e0e17142b8eb2317bf8847a2b31b52fc.jpeg","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":1},"content":[{"type":"text","text":"後記","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"最後，我想和你再聊一聊關於「","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"技術方案選型","attrs":{}},{"type":"text","text":"」的問題。","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":"你應該也看到了，這篇文章雖然始於 Redis，但並不止於 Redis。","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":"我們在分析 Redis 細節時，一直在提出問題，然後尋找更好的解決方案，在文章最後，又聊到一個專業的消息隊列應該怎麼做。","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":"其實，我們在討論技術選型時，就是一個關於如何取捨的問題。","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":"而這裏我想傳達給你的信息是，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"在面對技術選型時，不要不經過思考就覺得哪個方案好，哪個方案不好","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":"你需要根據具體場景具體分析，這裏我把這個分析過程分爲 2 個層面：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":1,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"業務功能角度","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"技術資源角度","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":"這篇文章所講到的內容，都是以業務功能角度出發做決策的。","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":"但這裏的第二點，從技術資源角度出發，其實也很重要。","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":"技術資源的角度是說，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"你所處的公司環境、技術資源能否匹配這些技術方案","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":"這個怎麼解釋呢？","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":"簡單來講，就是你所在的公司、團隊，是否有匹配的資源能 hold 住這些技術方案。","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":"我們都知道 Kafka、RabbitMQ 是非常專業的消息中間件，但它們的部署和運維，相比於 Redis 來說，也會更復雜一些。","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":"如果你在一個大公司，公司本身就有優秀的運維團隊，那麼使用這些中間件肯定沒問題，因爲有足夠優秀的人能 hold 住這些中間件，公司也會投入人力和時間在這個方向上。","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":"但如果你是在一個初創公司，業務正處在快速發展期，暫時沒有能 hold 住這些中間件的團隊和人，如果貿然使用這些組件，當發生故障時，排查問題也會變得很困難，甚至會阻礙業務的發展。","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":"而這種情形下，如果公司的技術人員對於 Redis 都很熟，綜合評估來看，Redis 也基本可以滿足業務 90% 的需求，那當下選擇 Redis 未必不是一個好的決策。","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，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"做技術選型不只是技術問題，還與人、團隊、管理、組織結構有關","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":"也正是因爲這些原因，當你在和別人討論技術選型問題時，你會發現每個公司的做法都不相同。","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":"畢竟每個公司所處的環境和文化不一樣，做出的決策當然就會各有差異。","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":"如果你不瞭解這其中的邏輯，那在做技術選型時，只會趨於表面現象，無法深入到問題根源。","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":"而一旦你理解了這個邏輯，那麼你在看待這個問題時，不僅對於技術會有更加深刻認識，對技術資源和人的把握，也會更加清晰。","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":"希望你以後在做技術選型時，能夠把這些因素也考慮在內，這對你的技術成長之路也是非常有幫助的。","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":"如果你還想看更多優質原創文章，歡迎關注我的公衆號「水滴與銀彈」。","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":"color","attrs":{"color":"#FF827B","name":"pink"}},{"type":"strong","attrs":{}}],"text":"我是 Kaito，是一個對於技術有思考的資深後端程序員，在我的文章中，我不僅會告訴你一個技術點是什麼，還會告訴你爲什麼這麼做？我還會嘗試把這些思考過程，提煉成通用的方法論，讓你可以應用在其它領域中，做到舉一反三。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"微信搜索關注「水滴與銀彈」公衆號，第一時間獲取優質技術乾貨。7 年資深後端研發，給你呈現不一樣的技術視角。","attrs":{}}]}],"attrs":{}}]}