源碼級別理解 Redis 持久化

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}},{"type":"strong","attrs":{}}],"text":"❝ ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"文章首發於公衆號“蘑菇睡不着”，歡迎來訪~ ❞","attrs":{}}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"前言","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"  大家都知道 Redis 是一個內存數據庫，數據都存儲在內存中，這也是 Redis 非常快的原因之一。雖然速度提上來了，但是如果數據一直放在內存中，是非常容易丟失的。比如 服務器關閉或宕機了，內存中的數據就木有了。爲了解決這一問題，Redis 提供了 持久化 機制。分別是 ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「RDB」","attrs":{}},{"type":"text","text":" 以及 ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「AOF」","attrs":{}},{"type":"text","text":" 持久化。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"RDB","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"什麼是 RDB 持久化?","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"RDB 持久化可以在指定的時間間隔內生成數據集的時間點快照(point-in-time snapshot)。","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/0f/0f8ee88669ba784ef0293673d0aa80f4.png","alt":"","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":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"RDB 的優點？","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"RDB 是一種表示某個即時點的 Redis 數據的緊湊文件。RDB 文件適用於備份。例如，你可能想要每小時歸檔最近24小時的 RDB 文件，每天保存近30天的 RDB 快照。這允許你很容易的恢復不同版本的數據集以容災。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"RDB 非常適合於災難恢復，作爲一個緊湊的單一文件，可以被傳輸到遠程的數據中心。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"RDB 最大化了 Redis 的性能。因爲 Redis 父進程持久化時唯一需要做的是啓動(fork)一個子進程，由子進程完成所有剩餘的工作。父進程實例不需要執行像磁盤IO這樣的操作。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"RDB 在重啓保存了大數據集的實例比 AOF 快。","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"RDB 的缺點？","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"當你需要在Redis停止工作(例如停電)時最小化數據丟失，RDB可能不太好。你可以配置不同的保存點(save point)來保存RDB文件(例如，至少5分鐘和對數據集100次寫之後，但是你可以有多個保存點)。然而，你通常每隔5分鐘或更久創建一個RDB快照，所以一旦Redis因爲任何原因沒有正確關閉而停止工作，你就得做好最近幾分鐘數據丟失的準備了。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"RDB需要經常調用fork()子進程來持久化到磁盤。如果數據集很大的話，fork()比較耗時，結果就是，當數據集非常大並且CPU性能不夠強大的話，Redis會停止服務客戶端幾毫秒甚至一秒。AOF也需要fork()，但是你可以調整多久頻率重寫日誌而不會有損(trade-off)持久性(durability)。","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"RDB 文件的創建與載入","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"  有個兩個 Redis 命令可以用於生成 RDB 文件，一個是 ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「SAVE」","attrs":{}},{"type":"text","text":"，另一個是 ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「BGSAVE」","attrs":{}},{"type":"text","text":"。  ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「SAVE」","attrs":{}},{"type":"text","text":" 命令會阻塞 Redis 服務器進程，直到 RDB 文件創建完畢爲止，在服務器進程阻塞期間，服務器不能處理任何命令請求。","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"> SAVE     // 一直等到 RDB 文件創建完畢\nOK\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"和 SAVE 命令直接阻塞服務器進程不同的是，BGSAVE 命令會派生出一個子進程，然後由子進程負責創建 RDB 文件，服務器進程(父進程)繼續處理命令進程。","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「執行fork的時候操作系統（類Unix操作系統）會使用寫時複製（copy-on-write）策略，即fork函數發生的一刻父子進程共享同一內存數據，當父進程要更改其中某片數據時（如執行一個寫命令 ），操作系統會將該片數據複製一份以保證子進程的數據不受影響，所以新的RDB文件存儲的是執行fork一刻的內存數據。」","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"> BGSAVE  // 派生子進程，並由子進程創建 RDB 文件\nBackground saving started\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"生成 RDB 文件由兩種方式：","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「一種是手動，就是上邊介紹的用命令的方式；另一種是自動的方式。」","attrs":{}},{"type":"text","text":"接下來詳細介紹一下自動生成 RDB 文件的流程。Redis 允許用戶通過設置服務器配置的 ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「save」","attrs":{}},{"type":"text","text":" 選項，讓服務器每隔一段時間自動執行一次 BGSAVE 命令。用戶可以通過在 redis.conf 配置文件中的 SNAPSHOTTING 下 save 選項設置多個保存條件，但只要其中任意一個條件被滿足，服務器就會執行 BGSAEVE 命令。如，以下配置：save 900 1save 300 10save 60 10000上邊三個配置的含義是：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"服務器在 900 秒內，對數據庫進行了至少 1 次修改。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"服務器在 300 秒內，對數據庫進行了至少 10 次修改。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"服務器在 60 秒內，對數據庫進行了至少 10000 次修改。","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"如果沒有手動去配置 save 選項，那麼服務器會爲 save 選項配置默認參數：save 900 1save 300 10save 60 10000接着，服務器就會根據 save 選項的配置，去設置服務器狀態 redisServer 結構的 saveparams 屬性：","attrs":{}}]},{"type":"codeblock","attrs":{"lang":"cpp"},"content":[{"type":"text","text":"struct redisServer{\n\n  // ...\n  \n  // 記錄了保存條件的數組\n  struct saveparams *saveparams;\n  \n  // ...\n};\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"saveparams 屬性是一個數組，數組中的每一個元素都是一個 saveparam 結構，每個 saveparam 結構都保存了一個 save 選項設置的保存條件：","attrs":{}}]},{"type":"codeblock","attrs":{"lang":"cpp"},"content":[{"type":"text","text":"struct saveparam {\n\n  // 秒數\n  time_t seconds;\n  \n  // 修改數\n  int changes;\n};\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"除了 saveparams 數組之外，服務器狀態還維持着一個 dirty 計數器，以及一個 lastsave 屬性；","attrs":{}}]},{"type":"codeblock","attrs":{"lang":"cpp"},"content":[{"type":"text","text":"struct redisServer {\n    // ...\n    \n    // 修改計數器\n    long long dirty;\n    \n    // 上一次執行保存時間\n     time_t lastsave;\n     \n     // ...\n}\n","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"dirty 計數器記錄距離上一次成功執行 SAVE 或 BGSAVE 命令之後，服務器對數據庫狀態(服務器中的所有數據庫)進行了多少次修改(包括寫入、刪除、更新等操作)。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"lastsave 屬性是一個 UNIX 時間戳，記錄了服務器上一次執行 SAVE 或 BGSAVE 命令的時間。","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"檢查條件是否滿足觸發 RDB","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"Redis 的服務器週期性操作函數 serverCron 默認每隔 100 毫秒執行一次，該函數用於對正在運行的服務器進行維護，它的其中一項工作就是檢查 save 選項所設置的保存條件是否已經滿足，如果滿足的話就執行 BGSAVE 命令。Redis serverCron 源碼解析如下：","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/44/446e4e4b221984030695e5a8137c055f.png","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"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":"#2b2b2b","name":"user"}}],"text":"程序會遍歷並檢查 saveparams 數組中的所有保存條件，只要有任意一個條件被滿足，服務器就會執行 BGSAVE 命令。下面是 rdbSaveBackground 的源碼流程：","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/c3/c34d14687753cc2042e6420a9f77a6ce.png","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"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":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"RDB 文件結構","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"下圖展示了一個完整 RDB 文件所包含的各個部分。","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/ea/eae116936ace09485566ca5e0cbdb44f.png","alt":"","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},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"redis 文件的最開頭是 ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「REDIS」","attrs":{}},{"type":"text","text":" 部分，這個部分的長度是 5 字節，保存着 “REDIS” 五個字符。通過這五個字符，程序可以在載入文件時，快速檢查所載入的文件是否時 RDB 文件。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「db_version」","attrs":{}},{"type":"text","text":" 長度爲 4 字節，他的值時一個字符串表示的整數，這個整數記錄了 RDB 文件的版本號，比如 “0006” 就代表 RDB 文件的版本爲第六版。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「database」","attrs":{}},{"type":"text","text":" 部分包含着零個或任意多個數據庫，以及各個數據庫中的鍵值對數據：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"如果服務器的數據庫狀態爲空(所有數據庫都是空的)，那麼這個部分也爲空，長度爲 0 字節。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"如果服務器的數據庫狀態爲非空(有至少一個數據庫非空)，那麼這個部分也爲非空，根據數據庫所保存鍵值對的數量、類型和內容不同，這個部分的長度也會有所不同。","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「EOF」","attrs":{}},{"type":"text","text":" 常量的長度爲 1 字節，這個常量標誌着 RDB 文件正文內容的結束，當讀入程序遇到這個值後，他知道所有數據庫的所有鍵值對已經載入完畢了。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「check_sum」","attrs":{}},{"type":"text","text":" 是一個 8 字節長的無符號整數，保存着一個校驗和，這個校驗和時程序通過對 REDIS、db_version、database、EOF 四個部分的內容進行計算得出的。服務器在載入 RDB 文件時，會將載入數據所計算出的校驗和與 check_sum 所記錄的校驗和進行對比，以此來檢查 RDB 是否有出錯或者損壞的情況。舉個例子：下圖是一個 0 號數據庫和 3 號數據庫的 RDB 文件。第一個就是 “REDIS” 表示是一個 RDB 文件，之後的 “0006” 表示這是第六版的 REDIS 文件，然後是兩個數據庫，之後就是 EOF 結束標識符，最後就是 check_sum。","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/61/612d76eb9de49a81d1ad3af45b5ffe5e.png","alt":"","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":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"AOF 持久化","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"什麼是 AOF 持久化","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"AOF持久化方式記錄每次對服務器寫的操作,當服務器重啓的時候會重新執行這些命令來恢復原始的數據,AOF命令以redis協議追加保存每次寫的操作到文件末尾.Redis還能對AOF文件進行後臺重寫,使得AOF文件的體積不至於過大.","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/25/258888ba9a9ea1f0112c6e93999e8480.png","alt":"","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":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"AOF 的優點？","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"使用AOF 會讓你的Redis更加耐久: 你可以使用不同的fsync策略：無fsync,每秒fsync,每次寫的時候fsync.使用默認的每秒fsync策略,Redis的性能依然很好(fsync是由後臺線程進行處理的,主線程會盡力處理客戶端請求),一旦出現故障，你最多丟失1秒的數據.","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"AOF文件是一個只進行追加的日誌文件,所以不需要寫入seek,即使由於某些原因(磁盤空間已滿，寫的過程中宕機等等)未執行完整的寫入命令,你也也可使用redis-check-aof工具修復這些問題.","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"Redis 可以在 AOF 文件體積變得過大時，自動地在後臺對 AOF 進行重寫： 重寫後的新 AOF 文件包含了恢復當前數據集所需的最小命令集合。 整個重寫操作是絕對安全的，因爲 Redis 在創建新 AOF 文件的過程中，會繼續將命令追加到現有的 AOF 文件裏面，即使重寫過程中發生停機，現有的 AOF 文件也不會丟失。 而一旦新 AOF 文件創建完畢，Redis 就會從舊 AOF 文件切換到新 AOF 文件，並開始對新 AOF 文件進行追加操作。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"AOF 文件有序地保存了對數據庫執行的所有寫入操作， 這些寫入操作以 Redis 協議的格式保存， 因此 AOF 文件的內容非常容易被人讀懂， 對文件進行分析（parse）也很輕鬆。 導出（export） AOF 文件也非常簡單： 舉個例子， 如果你不小心執行了 FLUSHALL 命令， 但只要 AOF 文件未被重寫， 那麼只要停止服務器， 移除 AOF 文件末尾的 FLUSHALL 命令， 並重啓 Redis ， 就可以將數據集恢復到 FLUSHALL 執行之前的狀態。","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"AOF 的缺點？","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"對於相同的數據集來說，AOF 文件的體積通常要大於 RDB 文件的體積。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"根據所使用的 fsync 策略，AOF 的速度可能會慢於 RDB 。 在一般情況下， 每秒 fsync 的性能依然非常高， 而關閉 fsync 可以讓 AOF 的速度和 RDB 一樣快， 即使在高負荷之下也是如此。 不過在處理巨大的寫入載入時，RDB 可以提供更有保證的最大延遲時間（latency）。","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"AOF持久化的實現","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"AOF持久化功能的實現可以分爲命令追加（append）、文件寫入、文件同步（sync）三個步驟。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"命令追加","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"當 AOF 持久化功能處於打開狀態時，服務器在執行完一個寫命令之後，會以協議格式將被執行的寫命令追加到服務器狀態的 aof_buf 緩衝區的末尾。","attrs":{}}]},{"type":"codeblock","attrs":{"lang":"cpp"},"content":[{"type":"text","text":"struct redisServer {\n  // ...\n  // AOF 緩衝區  \n  sds aof_buf;\n  \n  // ..\n};\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"如果客戶端向服務器發送以下命令：","attrs":{}}]},{"type":"codeblock","attrs":{"lang":"cpp"},"content":[{"type":"text","text":"> set KEY VALUE\nOK\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"那麼服務器在執行這個 set 命令之後，會將以下協議內容追加到 aof_buf 緩衝區的末尾；","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"*3\\r\\n$3\\r\\nSET\\r\\n$3\\r\\nKEY\\r\\n$5\\r\\nVALUE\\r\\n\n","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"AOF 文件的寫入與同步","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"  Redis的服務器進程就是一個事件循環（loop），這個循環中的文件事件負責接收客戶端的命令請求，以及向客戶端發送命令回覆，而時間事件則負責執行像 serverCron 函數這樣需要定時運行的函數。  因爲服務器在處理文件事件時可能會執行寫命令，使得一些內容被追加到aof_buf緩衝區裏面，所以在服務器每次結束一個事件循環之前，它都會調用 ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「flushAppendOnlyFile」","attrs":{}},{"type":"text","text":" 函數，考慮是否需要將aof_buf緩衝區中的內容寫入和保存到AOF文件裏面，這個過程可以用以下僞代碼表示：","attrs":{}}]},{"type":"codeblock","attrs":{"lang":"cpp"},"content":[{"type":"text","text":"def eventLoop():\n  while True:\n  \n  #處理文件事件，接收命令請求以及發送命令回覆\n  #處理命令請求時可能會有新內容被追加到 aof_buf緩衝區中\n  processFileEvents()\n  \n  #處理時間事件\n  processTimeEvents()\n  \n  #考慮是否要將 aof_buf中的內容寫入和保存到 AOF文件裏面\n  flushAppendOnlyFile()\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"flushAppendOnlyFile函數的行爲由服務器配置的 ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「appendfsync」","attrs":{}},{"type":"text","text":" 選項的值來決定，各個不同值產生的行爲如下表所示。","attrs":{}}]},{"type":"embedcomp","attrs":{"type":"table","data":{"content":"

appendfsync 選項的值	flushAppendOnlyFile 函數的行爲
always	將 aof_buf 緩衝區中的所有內容寫入並同步到 AOF 文件
everysec	將 aof_buf 緩衝區中的所有內容寫入到 AOF 文件，如果上次同步 AOF 文件的時間距離現在超過一秒鐘，那麼再次對 AOF 文件進行同步，並且這個同步操作是由一個線程專門負責執行的
no	將 aof_buf 緩衝區中的所有內容寫入到 AOF 文件，但並不對 AOF 文件進行同步，何時同步由操作系統來決定

"}}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"如果用戶沒有主動爲appendfsync選項設置值，那麼appendfsync選項的默認值爲everysec。寫到這裏有的小夥伴可能會對上面說的寫入和同步含義弄混，這裏說一下：","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「寫入：將 aof_buf 中的數據寫入到 AOF 文件中。」「同步：調用 fsync 以及 fdatasync 函數，將 AOF 文件中的數據保存到磁盤中。」","attrs":{}},{"type":"text","text":"通俗地講就是，你要往一個文件寫東西，寫的過程就是寫入，而同步則是將文件保存，數據落到磁盤上。大家之前看文章的時候是不是大多都說 AOF 最多丟失一秒鐘的數據，那是因爲 redis AOF 默認是 everysec 策略，這個策略每秒執行一次，所以 AOF 持久化最多丟失一秒鐘的數據。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"AOF 文件的載入與數據還原","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"因爲AOF文件裏面包含了重建數據庫狀態所需的所有寫命令，所以服務器只要讀入並重新執行一遍AOF文件裏面保存的寫命令，就可以還原服務器關閉之前的數據庫狀態。 Redis讀取AOF文件並還原數據庫狀態的詳細步驟如下：","attrs":{}}]},{"type":"numberedlist","attrs":{"start":null,"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":"color","attrs":{"color":"#595959","name":"user"}}],"text":"創建一個不帶網絡連接的僞客戶端（fake client）：因爲Redis的命令只能在客戶端上 下文中執行，而載入AOF文件時所使用的命令直接來源於AOF文件而不是網絡連接，所以服 務器使用了一個沒有網絡連接的僞客戶端來執行AOF文件保存的寫命令，僞客戶端執行命令 的效果和帶網絡連接的客戶端執行命令的效果完全一樣。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"從AOF文件中分析並讀取出一條寫命令。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"使用僞客戶端執行被讀出的寫命令。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"一直執行步驟2和步驟3，直到AOF文件中的所有寫命令都被處理完畢爲止。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":5,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":6,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"當完成以上步驟之後，AOF文件所保存的數據庫狀態就會被完整地還原出來，整個過程 如下圖所示。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":7,"align":null,"origin":null}}]}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/54/5492a19465f8f3c4743b4a4e1554c24d.png","alt":"","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null}}]}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"AOF 重寫","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"因爲AOF持久化是通過保存被執行的寫命令來記錄數據庫狀態的，所以隨着服務器運行 時間的流逝，AOF文件中的內容會越來越多，文件的體積也會越來越大，如果不加以控制的 話，體積過大的AOF文件很可能對Redis服務器、甚至整個宿主計算機造成影響，並且AOF文 件的體積越大，使用AOF文件來進行數據還原所需的時間就越多。如 客戶端執行了以下命令是：","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"> rpush list \"A\" \"B\"\nOK\n> rpush list \"C\"\nOK\n> rpush list \"D\"\nOK\n> rpush list \"E\" \"F\"\nOK\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"  那麼光是爲了記錄這個list鍵的狀態，AOF文件就需要保存四條命令。  對於實際的應用程度來說，寫命令執行的次數和頻率會比上面的簡單示例要高得多，所 以造成的問題也會嚴重得多。 爲了解決AOF文件體積膨脹的問題，Redis提供了AOF文件重寫（rewrite）功能。通過該 功能，Redis服務器可以創建一個新的AOF文件來替代現有的AOF文件，新舊兩個AOF文件所 保存的數據庫狀態相同，但新AOF文件不會包含任何浪費空間的冗餘命令，所以新AOF文件 的體積通常會比舊AOF文件的體積要小得多。 在接下來的內容中，我們將介紹AOF文件重寫的實現原理，以及BGREWEITEAOF命令 的實現原理。  雖然Redis將生成新AOF文件替換舊AOF文件的功能命名爲“AOF文件重寫”，但實際上， AOF文件重寫並不需要對現有的AOF文件進行任何讀取、分析或者寫入操作，這個功能是通 過讀取服務器當前的數據庫狀態來實現的。  就像上面的情況，服務器完全可以將這六條命令合併成一條。","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"> rpush list \"A\" \"B\" \"C\" \"D\" \"E\" \"F\"\n","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"  除了上面列舉的列表鍵之外，其他所有類型的鍵都可以用同樣的方法去減少 AOF文件中的命令數量。首先從數據庫中讀取鍵現在的值，然後用一條命令去記錄鍵值對，代替之前記錄這個鍵值對的多條命令，這就是AOF重寫功能的實現原理。  在實際中，爲了避免在執行命令時造成客戶端輸入緩衝區溢出，重寫程序在處理列表、 哈希表、集合、有序集合這四種可能會帶有多個元素的鍵時，會先檢查鍵所包含的元素數 量，如果元素的數量超過了redis.h/REDIS_AOF_REWRITE_ITEMS_PER_CMD常量的值，那 麼重寫程序將使用多條命令來記錄鍵的值，而不單單使用一條命令。 在目前版本中，REDIS_AOF_REWRITE_ITEMS_PER_CMD常量的值爲64，這也就是 說，如果一個集合鍵包含了超過64個元素，那麼重寫程序會用多條SADD命令來記錄這個集 合，並且每條命令設置的元素數量也爲64個。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"AOF 後臺重寫","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"  AOF 重寫會執行大量的寫操作，這樣會影響主線程，所以redis AOF 重寫放到了子進程去執行。這樣可以達到兩個目的：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"子進程進行AOF重寫期間，服務器進程（父進程）可以繼續處理命令請求。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"子進程帶有服務器進程的數據副本，使用子進程而不是線程，可以在避免使用鎖的情況 下，保證數據的安全性。","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"但是有一個問題，當子進程重寫數據時，主進程依然在處理新的數據，這也就會造成數據不一致情況。","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「爲了解決這種數據不一致問題，Redis服務器設置了一個AOF重寫緩衝區，這個緩衝區在 服務器創建子進程之後開始使用，當Redis服務器執行完一個寫命令之後，它會同時將這個寫 命令發送給AOF緩衝區和AOF重寫緩衝區」","attrs":{}},{"type":"text","text":"，如下圖：","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/1e/1e642dc58c2ba628c734ce17d5ee80ac.png","alt":"","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":"#2b2b2b","name":"user"}}],"text":"這也就是說，在子進程執行AOF重寫期間，服務器進程需要執行以下三個工作：","attrs":{}}]},{"type":"numberedlist","attrs":{"start":null,"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":"color","attrs":{"color":"#595959","name":"user"}}],"text":"執行客戶端發來的命令。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"將執行後的寫命令追加到AOF緩衝區。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"將執行後的寫命令追加到AOF重寫緩衝區。","attrs":{}}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"這樣一來可以保證：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"AOF緩衝區的內容會定期被寫入和同步到AOF文件，對現有AOF文件的處理工作會如常 進行。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"從創建子進程開始，服務器執行的所有寫命令都會被記錄到AOF重寫緩衝區裏面。","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"當子進程完成AOF重寫工作之後，它會向父進程發送一個信號，父進程在接到該信號之 後，會調用一個信號處理函數，並執行以下工作：","attrs":{}}]},{"type":"numberedlist","attrs":{"start":null,"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":"color","attrs":{"color":"#595959","name":"user"}}],"text":"將AOF重寫緩衝區中的所有內容寫入到新AOF文件中，這時新AOF文件所保存的數 據庫狀態將和服務器當前的數據庫狀態一致。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#595959","name":"user"}}],"text":"對新的AOF文件進行改名，原子地（atomic）覆蓋現有的AOF文件，完成新舊兩個 AOF文件的替換。","attrs":{}}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"這個信號處理函數執行完畢之後，父進程就可以繼續像往常一樣接受命令請求了。在整個AOF後臺重寫過程中，只有信號處理函數執行時會對服務器進程（父進程）造成 阻塞，在其他時候，AOF後臺重寫都不會阻塞父進程，這將AOF重寫對服務器性能造成的影 響降到了最低。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#40B8FA","name":"user"}}],"text":"Redis 混合持久化","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"Redis 還可以同時使用 AOF 持久化和 RDB 持久化。 在這種情況下， 當 Redis 重啓時， 它會優先使用 AOF 文件來還原數據集， 因爲 AOF 文件保存的數據集通常比 RDB 文件所保存的數據集更完整。但是 AOF 恢復比較慢，Redis 4.0 推出了","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「混合持久化」","attrs":{}},{"type":"text","text":"。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"混合持久化： 將 rdb 文件的內容和增量的 AOF 日誌文件存在一起。這裏的 AOF 日誌不再是全量的日誌，而是 自持久化開始到持久化結束 的這段時間發生的增量 AOF 日誌，通常這部分 AOF 日誌很小。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"於是在 Redis 重啓的時候，可以先加載 ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「RDB」","attrs":{}},{"type":"text","text":" 的內容，然後再重放增量 AOF 日誌就可以完全替代之前的 AOF 全量文件重放，重啓效率因此大幅得到提升。","attrs":{}}]},{"type":"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":"#2b2b2b","name":"user"}}],"text":"覺得文章不錯的話，小夥伴們麻煩點個贊、關個注、轉個發一下唄～你的支持就是我寫文章的動力。","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","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"更多精彩的文章請關注公衆號“","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#3594F7","name":"user"}},{"type":"strong","attrs":{}}],"text":"「蘑菇睡不着」","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","marks":[{"type":"color","attrs":{"color":"#2b2b2b","name":"user"}}],"text":"你越主動就會越主動，我們下期見～","attrs":{}}]}]}