Redis爲什麼變慢了？一文講透如何排查Redis性能問題 | 萬字長文

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Redis 作爲優秀的內存數據庫，其擁有非常高的性能，單個實例的 OPS 能夠達到 10W 左右。但也正因此如此，當我們在使用 Redis 時，如果發現操作延遲變大的情況，就會與我們的預期不符。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"你也許或多或少地，也遇到過以下這些場景：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在 Redis 上執行同樣的命令，爲什麼有時響應很快，有時卻很慢？","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"爲什麼 Redis 執行 SET、DEL 命令耗時也很久？","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"爲什麼我的 Redis 突然慢了一波，之後又恢復正常了？","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"爲什麼我的 Redis 穩定運行了很久，突然從某個時間點開始變慢了？","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"...","attrs":{}}]}],"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":"如果你並不清楚 Redis 內部的實現原理，那麼在排查這種延遲問題時就會一頭霧水。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你也遇到了以上情況，那麼，這篇文章將會給你一個「全面」的問題排查思路，並且針對這些導致變慢的場景，我還會給你一個高效的解決方案。","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","marks":[{"type":"strong","attrs":{}}],"text":"在正文開始之前，我需要提醒你的是，這篇文章很長，涵蓋的 Redis 知識點也非常廣，全篇文章接近 2W 字，如果此時你的閱讀環境不適合專注閱讀，我建議你先收藏此文章，然後在合適的時間專注閱讀這篇文章。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你能耐心且認真地讀完這篇文章，我可以保證，你對 Redis 的性能調優將會有非常大的收穫。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你準備好了，那就跟着我的思路開始吧！","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/ea/ea3e484f08ebe9dde9b117a994d218e8.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"Redis真的變慢了嗎？","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":"首先，在開始之前，你需要弄清楚 Redis 是否真的變慢了？","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":"如果你發現你的業務服務 API 響應延遲變長，首先你需要先排查服務內部，究竟是哪個環節拖慢了整個服務。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"比較高效的做法是，在服務內部集成","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"鏈路追蹤","attrs":{}},{"type":"text","text":"，也就是在服務訪問外部依賴的出入口，記錄下每次請求外部依賴的響應延時。","attrs":{}}]},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/a3/a33803fc385125d42f4b2b65d341fa31.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},"content":[{"type":"text","text":"如果你發現確實是操作 Redis 的這條鏈路耗時變長了，那麼此刻你需要把焦點關注在業務服務到 Redis 這條鏈路上。","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":"從你的業務服務到 Redis 這條鏈路變慢的原因可能也有 2 個：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"業務服務器到 Redis 服務器之間的網絡存在問題，例如網絡線路質量不佳，網絡數據包在傳輸時存在延遲、丟包等情況","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"Redis 本身存在問題，需要進一步排查是什麼原因導致 Redis 變慢","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":"通常來說，第一種情況發生的概率比較小，如果是服務器之間網絡存在問題，那部署在這臺業務服務器上的所有服務都會發生網絡延遲的情況，此時你需要聯繫網絡運維同事，讓其協助解決網絡問題。","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":"我們這篇文章，重點關注的是第二種情況。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"也就是從 Redis 角度來排查，是否存在導致變慢的場景，以及都有哪些因素會導致 Redis 的延遲增加，然後針對性地進行優化。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"排除網絡原因，如何確認你的 Redis 是否真的變慢了？","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":"首先，你需要對 Redis 進行基準性能測試，瞭解你的 Redis 在生產環境服務器上的基準性能。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"什麼是基準性能？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"簡單來講，基準性能就是指 Redis 在一臺負載正常的機器上，其最大的響應延遲和平均響應延遲分別是怎樣的？","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":"爲什麼要測試基準性能？我參考別人提供的響應延遲，判斷自己的 Redis 是否變慢不行嗎？","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":"答案是否定的。","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":"因爲 Redis 在不同的軟硬件環境下，它的性能是各不相同的。","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":"例如，我的機器配置比較低，當延遲爲 2ms 時，我就認爲 Redis 變慢了，但是如果你的硬件配置比較高，那麼在你的運行環境下，可能延遲是 0.5ms 時就可以認爲 Redis 變慢了。","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":"所以，你只有瞭解了你的 Redis 在生產環境服務器上的基準性能，才能進一步評估，當其延遲達到什麼程度時，才認爲 Redis 確實變慢了。","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":"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":"爲了避免業務服務器到 Redis 服務器之間的網絡延遲，你需要直接在 Redis 服務器上測試實例的響應延遲情況。執行以下命令，就可以測試出這個實例 60 秒內的最大響應延遲：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"$ redis-cli -h 127.0.0.1 -p 6379 --intrinsic-latency 60\nMax latency so far: 1 microseconds.\nMax latency so far: 15 microseconds.\nMax latency so far: 17 microseconds.\nMax latency so far: 18 microseconds.\nMax latency so far: 31 microseconds.\nMax latency so far: 32 microseconds.\nMax latency so far: 59 microseconds.\nMax latency so far: 72 microseconds.\n\n1428669267 total runs (avg latency: 0.0420 microseconds / 42.00 nanoseconds per run).\nWorst run took 1429x longer than the average latency.","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"從輸出結果可以看到，這 60 秒內的最大響應延遲爲 72 微秒（0.072毫秒）。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"你還可以使用以下命令，查看一段時間內 Redis 的最小、最大、平均訪問延遲：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"$ redis-cli -h 127.0.0.1 -p 6379 --latency-history -i 1\nmin: 0, max: 1, avg: 0.13 (100 samples) -- 1.01 seconds range\nmin: 0, max: 1, avg: 0.12 (99 samples) -- 1.01 seconds range\nmin: 0, max: 1, avg: 0.13 (99 samples) -- 1.01 seconds range\nmin: 0, max: 1, avg: 0.10 (99 samples) -- 1.01 seconds range\nmin: 0, max: 1, avg: 0.13 (98 samples) -- 1.00 seconds range\nmin: 0, max: 1, avg: 0.08 (99 samples) -- 1.01 seconds range\n...","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以上輸出結果是，每間隔 1 秒，採樣 Redis 的平均操作耗時，其結果分佈在 0.08 ~ 0.13 毫秒之間。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"瞭解了基準性能測試方法，那麼你就可以按照以下幾步，來判斷你的 Redis 是否真的變慢了：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"在相同配置的服務器上，測試一個正常 Redis 實例的基準性能","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"找到你認爲可能變慢的 Redis 實例，測試這個實例的基準性能","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"如果你觀察到，這個實例的運行延遲是正常 Redis 基準性能的 2 倍以上，即可認爲這個 Redis 實例確實變慢了 ","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":"確認是 Redis 變慢了，那如何排查是哪裏發生了問題呢？","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":"下面跟着我的思路，我們從易到難，一步步來分析可能導致 Redis 變慢的因素。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"使用複雜度過高的命令","attrs":{}}]},{"type":"paragraph","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，第一步，你需要去查看一下 Redis 的慢日誌（slowlog）。","attrs":{}}]},{"type":"paragraph","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 提供了慢日誌命令的統計功能，它記錄了有哪些命令在執行時耗時比較久。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"查看 Redis 慢日誌之前，你需要設置慢日誌的閾值。例如，設置慢日誌的閾值爲 5 毫秒，並且保留最近 500 條慢日誌記錄：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":""},"content":[{"type":"text","text":"# 命令執行耗時超過 5 毫秒，記錄慢日誌\nCONFIG SET slowlog-log-slower-than 5000\n# 只保留最近 500 條慢日誌\nCONFIG SET slowlog-max-len 500","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"設置完成之後，所有執行的命令如果操作耗時超過了 5 毫秒，都會被 Redis 記錄下來。","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":"此時，你可以執行以下命令，就可以查詢到最近記錄的慢日誌：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":""},"content":[{"type":"text","text":"127.0.0.1:6379> SLOWLOG get 5\n1) 1) (integer) 32693 # 慢日誌ID\n 2) (integer) 1593763337 # 執行時間戳\n 3) (integer) 5299 # 執行耗時(微秒)\n 4) 1) \"LRANGE\" # 具體執行的命令和參數\n 2) \"user_list:2000\"\n 3) \"0\"\n 4) \"-1\"\n2) 1) (integer) 32692\n 2) (integer) 1593763337\n 3) (integer) 5044\n 4) 1) \"GET\"\n 2) \"user_info:1000\"\n...","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"通過查看慢日誌，我們就可以知道在什麼時間點，執行了哪些命令比較耗時。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你的應用程序執行的 Redis 命令有以下特點，那麼有可能會導致操作延遲變大：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"經常使用 O(N) 以上覆雜度的命令，例如 SORT、SUNION、ZUNIONSTORE 聚合類命令","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"使用 O(N) 複雜度的命令，但 N 的值非常大","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":"第一種情況導致變慢的原因在於，Redis 在操作內存數據時，時間複雜度過高，要花費更多的 CPU 資源。","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":"第二種情況導致變慢的原因在於，Redis 一次需要返回給客戶端的數據過多，更多時間花費在數據協議的組裝和網絡傳輸過程中。","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":"另外，我們還可以從資源使用率層面來分析，如果你的應用程序操作 Redis 的 OPS 不是很大，但 Redis 實例的 ","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"CPU 使用率卻很高","attrs":{}},{"type":"text","text":"，那麼很有可能是使用了複雜度過高的命令導致的。","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":"除此之外，我們都知道，Redis 是單線程處理客戶端請求的，如果你經常使用以上命令，那麼當 Redis 處理客戶端請求時，一旦前面某個命令發生耗時，就會導致後面的請求發生排隊，對於客戶端來說，響應延遲也會變長。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/0c/0ce7e8c0f53edbaf17a2c13d91c66882.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},"content":[{"type":"text","text":"針對這種情況如何解決呢？","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":"答案很簡單，你可以使用以下方法優化你的業務：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"儘量不使用 O(N) 以上覆雜度過高的命令，對於數據的聚合操作，放在客戶端做","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"執行 O(N) 命令，保證 N 儘量的小（推薦 N <= 300），每次獲取儘量少的數據，讓 Redis 可以及時處理返回","attrs":{}}]}],"attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"操作bigkey","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你查詢慢日誌發現，並不是複雜度過高的命令導致的，而都是 SET / DEL 這種簡單命令出現在慢日誌中，那麼你就要懷疑你的實例否寫入了 bigkey。","attrs":{}}]},{"type":"paragraph","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 在寫入數據時，需要爲新的數據分配內存，相對應的，當從 Redis 中刪除數據時，它會釋放對應的內存空間。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果一個 key 寫入的 value 非常大，那麼 Redis 在","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"分配內存時就會比較耗時","attrs":{}},{"type":"text","text":"。同樣的，當刪除這個 key 時，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"釋放內存也會比較耗時","attrs":{}},{"type":"text","text":"，這種類型的 key 我們一般稱之爲 bigkey。","attrs":{}}]},{"type":"paragraph","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，你需要檢查你的業務代碼，是否存在寫入 bigkey 的情況。你需要評估寫入一個 key 的數據大小，儘量避免一個 key 存入過大的數據。","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":"如果已經寫入了 bigkey，那有沒有什麼辦法可以掃描出實例中 bigkey 的分佈情況呢？","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":"答案是可以的。","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":"Redis 提供了掃描 bigkey 的命令，執行以下命令就可以掃描出，一個實例中 bigkey 的分佈情況，輸出結果是以類型維度展示的：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":""},"content":[{"type":"text","text":"$ redis-cli -h 127.0.0.1 -p 6379 --bigkeys -i 0.01\n\n...\n-------- summary -------\n\nSampled 829675 keys in the keyspace!\nTotal key length in bytes is 10059825 (avg len 12.13)\n\nBiggest string found 'key:291880' has 10 bytes\nBiggest list found 'mylist:004' has 40 items\nBiggest set found 'myset:2386' has 38 members\nBiggest hash found 'myhash:3574' has 37 fields\nBiggest zset found 'myzset:2704' has 42 members\n\n36313 strings with 363130 bytes (04.38% of keys, avg size 10.00)\n787393 lists with 896540 items (94.90% of keys, avg size 1.14)\n1994 sets with 40052 members (00.24% of keys, avg size 20.09)\n1990 hashs with 39632 fields (00.24% of keys, avg size 19.92)\n1985 zsets with 39750 members (00.24% of keys, avg size 20.03)","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"從輸出結果我們可以很清晰地看到，每種數據類型所佔用的最大內存 / 擁有最多元素的 key 是哪一個，以及每種數據類型在整個實例中的佔比和平均大小 / 元素數量。","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":"其實，使用這個命令的原理，就是 Redis 在內部執行了 SCAN 命令，遍歷整個實例中所有的 key，然後針對 key 的類型，分別執行 STRLEN、LLEN、HLEN、SCARD、ZCARD 命令，來獲取 String 類型的長度、容器類型（List、Hash、Set、ZSet）的元素個數。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這裏我需要提醒你的是，當執行這個命令時，要注意 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","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"對線上實例進行 bigkey 掃描時，Redis 的 OPS 會突增，爲了降低掃描過程中對 Redis 的影響，最好控制一下掃描的頻率，指定 -i 參數即可，它表示掃描過程中每次掃描後休息的時間間隔，單位是秒","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"掃描結果中，對於容器類型（List、Hash、Set、ZSet）的 key，只能掃描出元素最多的 key。但一個 key 的元素多，不一定表示佔用內存也多，你還需要根據業務情況，進一步評估內存佔用情況","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":"那針對 bigkey 導致延遲的問題，有什麼好的解決方案呢？","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":"這裏有兩點可以優化：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"業務應用盡量避免寫入 bigkey","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"如果你使用的 Redis 是 4.0 以上版本，用 UNLINK 命令替代 DEL，此命令可以把釋放 key 內存的操作，放到後臺線程中去執行，從而降低對 Redis 的影響","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"如果你使用的 Redis 是 6.0 以上版本，可以開啓 lazy-free 機制（lazyfree-lazy-user-del = yes），在執行 DEL 命令時，釋放內存也會放到後臺線程中執行","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":"但即便可以使用方案 2，我也不建議你在實例中存入 bigkey。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這是因爲 bigkey 在很多場景下，依舊會產生性能問題。例如，bigkey 在分片集羣模式下，對於數據的遷移也會有性能影響，以及我後面即將講到的數據過期、數據淘汰、透明大頁，都會受到 bigkey 的影響。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"集中過期","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你發現，平時在操作 Redis 時，並沒有延遲很大的情況發生，但在某個時間點突然出現一波延時，其現象表現爲：","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"變慢的時間點很有規律，例如某個整點，或者每間隔多久就會發生一波延遲。","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":"如果是出現這種情況，那麼你需要排查一下，業務代碼中是否存在設置大量 key 集中過期的情況。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果有大量的 key 在某個固定時間點集中過期，在這個時間點訪問 Redis 時，就有可能導致延時變大。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"爲什麼集中過期會導致 Redis 延遲變大？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這就需要我們瞭解 Redis 的過期策略是怎樣的。","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 的過期數據採用被動過期 + 主動過期兩種策略：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"被動過期：只有當訪問某個 key 時，才判斷這個 key 是否已過期，如果已過期，則從實例中刪除","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"主動過期：Redis 內部維護了一個定時任務，默認每隔 100 毫秒（1秒10次）就會從全局的過期哈希表中隨機取出 20 個 key，然後刪除其中過期的 key，如果過期 key 的比例超過了 25%，則繼續重複此過程，直到過期 key 的比例下降到 25% 以下，或者這次任務的執行耗時超過了 25 毫秒，纔會退出循環","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":"注意，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"這個主動過期 key 的定時任務，是在 Redis 主線程中執行的","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":"也就是說如果在執行主動過期的過程中，出現了需要大量刪除過期 key 的情況，那麼此時應用程序在訪問 Redis 時，必須要等待這個過期任務執行結束，Redis 纔可以服務這個客戶端請求。","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":"此時就會出現，應用訪問 Redis 延時變大。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果此時需要過期刪除的是一個 bigkey，那麼這個耗時會更久。而且，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"這個操作延遲的命令並不會記錄在慢日誌中","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":"因爲慢日誌中","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"只記錄一個命令真正操作內存數據的耗時","attrs":{}},{"type":"text","text":"，而 Redis 主動刪除過期 key 的邏輯，是在命令真正執行之前執行的。","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，此時你會看到，慢日誌中沒有操作耗時的命令，但我們的應用程序卻感知到了延遲變大，其實時間都花費在了刪除過期 key 上，這種情況我們需要尤爲注意。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/25/25dad196bad2e2bb79edc2059acde8f1.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},"content":[{"type":"text","text":"那遇到這種情況，如何分析和排查？","attrs":{}}]},{"type":"paragraph","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，你需要檢查你的業務代碼，是否存在集中過期 key 的邏輯。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一般集中過期使用的是 expireat / pexpireat 命令，你需要在代碼中搜索這個關鍵字。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"排查代碼後，如果確實存在集中過期 key 的邏輯存在，但這種邏輯又是業務所必須的，那此時如何優化，同時又不對 Redis 有性能影響呢？","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":"一般有兩種方案來規避這個問題：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"集中過期 key 增加一個隨機過期時間，把集中過期的時間打散，降低 Redis 清理過期 key 的壓力","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"如果你使用的 Redis 是 4.0 以上版本，可以開啓 lazy-free 機制，當刪除過期 key 時，把釋放內存的操作放到後臺線程中執行，避免阻塞主線程","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":"第一種方案，在設置 key 的過期時間時，增加一個隨機時間，僞代碼可以這麼寫：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":""},"content":[{"type":"text","text":"# 在過期時間點之後的 5 分鐘內隨機過期掉\nredis.expireat(key, expire_time + random(300))","attrs":{}}]},{"type":"paragraph","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，Redis 在處理過期時，不會因爲集中刪除過多的 key 導致壓力過大，從而避免阻塞主線程。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二種方案，Redis 4.0 以上版本，開啓 lazy-free 機制：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"# 釋放過期 key 的內存，放到後臺線程執行\nlazyfree-lazy-expire yes","attrs":{}}]},{"type":"paragraph","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，除了業務層面的優化和修改配置之外，你還可以通過運維手段及時發現這種情況。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"運維層面，你需要把 Redis 的各項運行狀態數據監控起來，在 Redis 上執行 INFO 命令就可以拿到這個實例所有的運行狀態數據。","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":"在這裏我們需要重點關注 expired_keys 這一項，它代表整個實例到目前爲止，累計刪除過期 key 的數量。","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":"你需要把這個指標監控起來，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"當這個指標在很短時間內出現了突增","attrs":{}},{"type":"text","text":"，需要及時報警出來，然後與業務應用報慢的時間點進行對比分析，確認時間是否一致，如果一致，則可以確認確實是因爲集中過期 key 導致的延遲變大。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"實例內存達到上限","attrs":{}}]},{"type":"paragraph","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 Redis 實例設置了內存上限 maxmemory，那麼也有可能導致 Redis 變慢。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當我們把 Redis 當做純緩存使用時，通常會給這個實例設置一個內存上限 maxmemory，然後設置一個數據淘汰策略。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而當實例的內存達到了 maxmemory 後，你可能會發現，在此之後每次寫入新數據，操作延遲變大了。","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":"這是爲什麼？","attrs":{}}]},{"type":"paragraph","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，當 Redis 內存達到 maxmemory 後，每次寫入新的數據之前，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"Redis 必須先從實例中踢出一部分數據，讓整個實例的內存維持在 maxmemory 之下","attrs":{}},{"type":"text","text":"，然後才能把新數據寫進來。","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":"這個踢出舊數據的邏輯也是需要消耗時間的，而具體耗時的長短，要取決於你配置的淘汰策略：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"allkeys-lru：不管 key 是否設置了過期，淘汰最近最少訪問的 key","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"volatile-lru：只淘汰最近最少訪問、並設置了過期時間的 key","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"allkeys-random：不管 key 是否設置了過期，隨機淘汰 key","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"volatile-random：只隨機淘汰設置了過期時間的 key","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"allkeys-ttl：不管 key 是否設置了過期，淘汰即將過期的 key","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"noeviction：不淘汰任何 key，實例內存達到 maxmeory 後，再寫入新數據直接返回錯誤","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"allkeys-lfu：不管 key 是否設置了過期，淘汰訪問頻率最低的 key（4.0+版本支持）","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"volatile-lfu：只淘汰訪問頻率最低、並設置了過期時間 key（4.0+版本支持）","attrs":{}}]}],"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":"具體使用哪種策略，我們需要根據具體的業務場景來配置。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一般最常使用的是 allkeys-lru / volatile-lru 淘汰策略，它們的處理邏輯是，每次從實例中隨機取出一批 key（這個數量可配置），然後淘汰一個最少訪問的 key，之後把剩下的 key 暫存到一個池子中，繼續隨機取一批 key，並與之前池子中的 key 比較，再淘汰一個最少訪問的 key。以此往復，直到實例內存降到 maxmemory 之下。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"需要注意的是，Redis 的淘汰數據的邏輯與刪除過期 key 的一樣，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"也是在命令真正執行之前執行的","attrs":{}},{"type":"text","text":"，也就是說它也會增加我們操作 Redis 的延遲，而且，寫 OPS 越高，延遲也會越明顯。","attrs":{}}]},{"type":"paragraph","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，如果此時你的 Redis 實例中還存儲了 bigkey，那麼","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"在淘汰刪除 bigkey 釋放內存時，也會耗時比較久","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/36/367a42fe57927cee24476747d2f42e43.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},"content":[{"type":"text","text":"看到了麼？bigkey 的危害到處都是，這也是前面我提醒你儘量不存儲 bigkey 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":"針對這種情況，如何解決呢？","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":"我給你 4 個方面的優化建議：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"避免存儲 bigkey，降低釋放內存的耗時","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"淘汰策略改爲隨機淘汰，隨機淘汰比 LRU 要快很多（視業務情況調整）","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"拆分實例，把淘汰 key 的壓力分攤到多個實例上","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","text":"如果使用的是 Redis 4.0 以上版本，開啓 layz-free 機制，把淘汰 key 釋放內存的操作放到後臺線程中執行（配置 lazyfree-lazy-eviction = yes）","attrs":{}}]}],"attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"fork耗時嚴重","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"爲了保證 Redis 數據的安全性，我們可能會開啓後臺定時 RDB 和 AOF rewrite 功能。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但如果你發現，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"操作 Redis 延遲變大，都發生在 Redis 後臺 RDB 和 AOF rewrite 期間","attrs":{}},{"type":"text","text":"，那你就需要排查，在這期間有可能導致變慢的情況。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當 Redis 開啓了後臺 RDB 和 AOF rewrite 後，在執行時，它們都需要主進程創建出一個子進程進行數據的持久化。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"主進程創建子進程，會調用操作系統提供的 fork 函數。","attrs":{}}]},{"type":"paragraph","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 fork 在執行過程中，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"主進程需要拷貝自己的內存頁表給子進程","attrs":{}},{"type":"text","text":"，如果這個實例很大，那麼這個拷貝的過程也會比較耗時。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而且這個 fork 過程會消耗大量的 CPU 資源，在完成 fork 之前，整個 Redis 實例會被阻塞住，無法處理任何客戶端請求。","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":"如果此時你的 CPU 資源本來就很緊張，那麼 fork 的耗時會更長，甚至達到秒級，這會嚴重影響 Redis 的性能。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"那如何確認確實是因爲 fork 耗時導致的 Redis 延遲變大呢？","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":"你可以在 Redis 上執行 INFO 命令，查看 latest","attrs":{}},{"type":"text","marks":[{"type":"italic","attrs":{}}],"text":"fork","attrs":{}},{"type":"text","text":"usec 項，單位微秒。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"# 上一次 fork 耗時，單位微秒\nlatest_fork_usec:59477","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這個時間就是主進程在 fork 子進程期間，整個實例阻塞無法處理客戶端請求的時間。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你發現這個耗時很久，就要警惕起來了，這意味在這期間，你的整個 Redis 實例都處於不可用的狀態。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"除了數據持久化會生成 RDB 之外，當主從節點第一次建立數據同步時，主節點也創建子進程生成 RDB，然後發給從節點進行一次全量同步，所以，這個過程也會對 Redis 產生性能影響。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/53/53e020351d0f7483480735a9e103a9fe.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},"content":[{"type":"text","text":"要想避免這種情況，你可以採取以下方案進行優化：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"控制 Redis 實例的內存：儘量在 10G 以下，執行 fork 的耗時與實例大小有關，實例越大，耗時越久","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"合理配置數據持久化策略：在 slave 節點執行 RDB 備份，推薦在低峯期執行，而對於丟失數據不敏感的業務（例如把 Redis 當做純緩存使用），可以關閉 AOF 和 AOF rewrite","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"Redis 實例不要部署在虛擬機上：fork 的耗時也與系統也有關，虛擬機比物理機耗時更久","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","text":"降低主從庫全量同步的概率：適當調大 repl-backlog-size 參數，避免主從全量同步","attrs":{}}]}],"attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"開啓內存大頁","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"除了上面講到的子進程 RDB 和 AOF rewrite 期間，fork 耗時導致的延時變大之外，這裏還有一個方面也會導致性能問題，這就是操作系統是否開啓了","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"內存大頁機制","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":"什麼是內存大頁？","attrs":{}}]},{"type":"paragraph","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，應用程序向操作系統申請內存時，是按","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"內存頁","attrs":{}},{"type":"text","text":"進行申請的，而常規的內存頁大小是 4KB。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Linux 內核從 2.6.38 開始，支持了","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"內存大頁機制","attrs":{}},{"type":"text","text":"，該機制允許應用程序以 2MB 大小爲單位，向操作系統申請內存。","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":"應用程序每次向操作系統申請的內存單位變大了，但這也意味着申請內存的耗時變長。","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":"這對 Redis 會有什麼影響呢？","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":"當 Redis 在執行後臺 RDB 和 AOF rewrite 時，採用 fork 子進程的方式來處理。但主進程 fork 子進程後，此時的","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"主進程依舊是可以接收寫請求的","attrs":{}},{"type":"text","text":"，而進來的寫請求，會採用 Copy On Write（寫時複製）的方式操作內存數據。","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":"也就是說，主進程一旦有數據需要修改，Redis 並不會直接修改現有內存中的數據，而是","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"先將這塊內存數據拷貝出來，再修改這塊新內存的數據","attrs":{}},{"type":"text","text":"，這就是所謂的「寫時複製」。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"寫時複製你也可以理解成，誰需要發生寫操作，誰就需要先拷貝，再修改。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這樣做的好處是，父進程有任何寫操作，並不會影響子進程的數據持久化（子進程只持久化 fork 這一瞬間整個實例中的所有數據即可，不關心新的數據變更，因爲子進程只需要一份內存快照，然後持久化到磁盤上）。","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":"但是請注意，主進程在拷貝內存數據時，這個階段就涉及到新內存的申請，如果此時操作系統開啓了內存大頁，那麼在此期間，客戶端即便只修改 10B 的數據，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"Redis 在申請內存時也會以 2MB 爲單位向操作系統申請，申請內存的耗時變長，進而導致每個寫請求的延遲增加，影響到 Redis 性能。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"同樣地，如果這個寫請求操作的是一個 bigkey，那主進程在拷貝這個 bigkey 內存塊時，一次申請的內存會更大，時間也會更久。可見，bigkey 在這裏又一次影響到了性能。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/40/4020506b7da20ae3bd4c2e76b4c867ba.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},"content":[{"type":"text","text":"那如何解決這個問題？","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":"很簡單，你只需要關閉內存大頁機制就可以了。","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":"首先，你需要查看 Redis 機器是否開啓了內存大頁：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"$ cat /sys/kernel/mm/transparent_hugepage/enabled\n[always] madvise never","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果輸出選項是 always，就表示目前開啓了內存大頁機制，我們需要關掉它：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"$ echo never > /sys/kernel/mm/transparent_hugepage/enabled","attrs":{}}]},{"type":"paragraph","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，操作系統提供的內存大頁機制，其優勢是，可以在一定程序上降低應用程序申請內存的次數。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但是對於 Redis 這種對性能和延遲極其敏感的數據庫來說，我們希望 Redis 在每次申請內存時，耗時儘量短，所以我不建議你在 Redis 機器上開啓這個機制。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"開啓AOF","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"前面我們分析了 RDB 和 AOF rewrite 對 Redis 性能的影響，主要關注點在 fork 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":"其實，關於數據持久化方面，還有影響 Redis 性能的因素，這次我們重點來看 AOF 數據持久化。","attrs":{}}]},{"type":"paragraph","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 AOF 配置不合理，還是有可能會導致性能問題。","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":"當 Redis 開啓 AOF 後，其工作原理如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"Redis 執行寫命令後，把這個命令寫入到 AOF 文件內存中（write 系統調用）","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"Redis 根據配置的 AOF 刷盤策略，把 AOF 內存數據刷到磁盤上（fsync 系統調用）","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":"爲了保證 AOF 文件數據的安全性，Redis 提供了 3 種刷盤機制：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"appendfsync always：主線程每次執行寫操作後立即刷盤，此方案會佔用比較大的磁盤 IO 資源，但數據安全性最高","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"appendfsync no：主線程每次寫操作只寫內存就返回，內存數據什麼時候刷到磁盤，交由操作系統決定，此方案對性能影響最小，但數據安全性也最低，Redis 宕機時丟失的數據取決於操作系統刷盤時機","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"appendfsync everysec：主線程每次寫操作只寫內存就返回，然後由後臺線程每隔 1 秒執行一次刷盤操作（觸發fsync系統調用），此方案對性能影響相對較小，但當 Redis 宕機時會丟失 1 秒的數據","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":"下面我們依次來分析，這幾個機制對性能的影響。","attrs":{}}]},{"type":"paragraph","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 AOF 配置爲 appendfsync always，那麼 Redis 每處理一次寫操作，都會把這個命令寫入到磁盤中才返回，整個過程都是在主線程執行的，這個過程必然會加重 Redis 寫負擔。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"原因也很簡單，操作磁盤要比操作內存慢幾百倍，採用這個配置會嚴重拖慢 Redis 的性能，因此我不建議你把 AOF 刷盤方式配置爲 always。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我們接着來看 appendfsync no 配置項。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在這種配置下，Redis 每次寫操作只寫內存，什麼時候把內存中的數據刷到磁盤，交給操作系統決定，此方案對 Redis 的性能影響最小，但當 Redis 宕機時，會丟失一部分數據，爲了數據的安全性，一般我們也不採取這種配置。","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":"如果你的 Redis 只用作純緩存，對於數據丟失不敏感，採用配置 appendfsync no 也是可以的。","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":"看到這裏，我猜你肯定和大多數人的想法一樣，選比較折中的方案 appendfsync everysec 就沒問題了吧？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這個方案優勢在於，Redis 主線程寫完內存後就返回，具體的刷盤操作是放到後臺線程中執行的，後臺線程每隔 1 秒把內存中的數據刷到磁盤中。","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":"這種方案既兼顧了性能，又儘可能地保證了數據安全，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":"但是，這裏我要給你潑一盆冷水了，採用這種方案你也要警惕一下，因爲這種方案還是存在導致 Redis 延遲變大的情況發生，甚至會阻塞整個 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":"這是爲什麼？我把 AOF 最耗時的刷盤操作，放到後臺線程中也會影響到 Redis 主線程？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"你試想這樣一種情況：當 Redis 後臺線程在執行 AOF 文件刷盤時，如果此時磁盤的 IO 負載很高，那這個後臺線程在執行刷盤操作（fsync系統調用）時就會被阻塞住。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此時的主線程依舊會接收寫請求，緊接着，主線程又需要把數據寫到文件內存中（write 系統調用），","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"但此時的後臺子線程由於磁盤負載過高，導致 fsync 發生阻塞，遲遲不能返回，那主線程在執行 write 系統調用時，也會被阻塞住","attrs":{}},{"type":"text","text":"，直到後臺線程 fsync 執行完成後，主線程執行 write 才能成功返回。","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","text":"看到了麼？在這個過程中，主線程依舊有阻塞的風險。","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，儘管你的 AOF 配置爲 appendfsync everysec，也不能掉以輕心，要警惕磁盤壓力過大導致的 Redis 有性能問題。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/54/54d128ba902795a6000bc66cb7905729.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},"content":[{"type":"text","text":"那什麼情況下會導致磁盤 IO 負載過大？以及如何解決這個問題呢？","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":"我總結了以下幾種情況，你可以參考進行問題排查：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"子進程正在執行 AOF rewrite，這個過程會佔用大量的磁盤 IO 資源","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"有其他應用程序在執行大量的寫文件操作，也會佔用磁盤 IO 資源","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":"對於情況1，說白了就是，Redis 的 AOF 後臺子線程刷盤操作，撞上了子進程 AOF rewrite！","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這怎麼辦？難道要關閉 AOF rewrite 纔行？","attrs":{}}]},{"type":"paragraph","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 提供了一個配置項，當子進程在 AOF rewrite 期間，可以讓後臺子線程不執行刷盤（不觸發 fsync 系統調用）操作。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這相當於在 AOF rewrite 期間，臨時把 appendfsync 設置爲了 none，配置如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"# AOF rewrite 期間，AOF 後臺子線程不進行刷盤操作\n# 相當於在這期間，臨時把 appendfsync 設置爲了 none\nno-appendfsync-on-rewrite yes","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，開啓這個配置項，在 AOF rewrite 期間，如果實例發生宕機，那麼此時會丟失更多的數據，性能和數據安全性，你需要權衡後進行選擇。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果佔用磁盤資源的是其他應用程序，那就比較簡單了，你需要定位到是哪個應用程序在大量寫磁盤，然後把這個應用程序遷移到其他機器上執行就好了，避免對 Redis 產生影響。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，如果你對 Redis 的性能和數據安全都有很高的要求，那麼我建議從","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"硬件層面","attrs":{}},{"type":"text","text":"來優化，更換爲 SSD 磁盤，提高磁盤的 IO 能力，保證 AOF 期間有充足的磁盤資源可以使用。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"綁定CPU","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"很多時候，我們在部署服務時，爲了提高服務性能，降低應用程序在多個 CPU 核心之間的上下文切換帶來的性能損耗，通常採用的方案是進程綁定 CPU 的方式提高性能。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但在部署 Redis 時，如果你需要綁定 CPU 來提高其性能，我建議你仔細斟酌後再做操作。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"爲什麼？","attrs":{}}]},{"type":"paragraph","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 在綁定 CPU 時，是有很多考究的，如果你不瞭解 Redis 的運行原理，隨意綁定 CPU 不僅不會提高性能，甚至有可能會帶來相反的效果。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我們都知道，一般現代的服務器會有多個 CPU，而每個 CPU 又包含多個物理核心，每個物理核心又分爲多個邏輯核心，每個物理核下的邏輯核共用 L1/L2 Cache。","attrs":{}}]},{"type":"paragraph","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 Server 除了主線程服務客戶端請求之外，還會創建子進程、子線程。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"其中子進程用於數據持久化，而子線程用於執行一些比較耗時操作，例如異步釋放 fd、異步 AOF 刷盤、異步 lazy-free 等等。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你把 Redis 進程只綁定了一個 CPU 邏輯核心上，那麼當 Redis 在進行數據持久化時，fork 出的子進程會繼承父進程的 CPU 使用偏好。","attrs":{}}]},{"type":"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":"而此時的子進程會消耗大量的 CPU 資源進行數據持久化（把實例數據全部掃描出來需要耗費CPU），這就會導致子進程會與主進程發生 CPU 爭搶，進而影響到主進程服務客戶端請求，訪問延遲變大。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這就是 Redis 綁定 CPU 帶來的性能問題。","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":"那如何解決這個問題呢？","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":"如果你確實想要綁定 CPU，可以優化的方案是，不要讓 Redis 進程只綁定在一個 CPU 邏輯核上，而是綁定在多個邏輯核心上，而且，綁定的多個邏輯核心最好是同一個物理核心，這樣它們還可以共用 L1/L2 Cache。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，即便我們把 Redis 綁定在多個邏輯核心上，也只能在一定程度上緩解主線程、子進程、後臺線程在 CPU 資源上的競爭。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"因爲這些子進程、子線程還是會在這多個邏輯核心上進行切換，存在性能損耗。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如何再進一步優化？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"可能你已經想到了，我們是否可以讓主線程、子進程、後臺線程，分別綁定在固定的 CPU 核心上，不讓它們來回切換，這樣一來，他們各自使用的 CPU 資源互不影響。","attrs":{}}]},{"type":"paragraph","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，這個方案 Redis 官方已經想到了。","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":"Redis 在 6.0 版本已經推出了這個功能，我們可以通過以下配置，對主線程、後臺線程、後臺 RDB 進程、AOF rewrite 進程，綁定固定的 CPU 邏輯核心：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"# Redis Server 和 IO 線程綁定到 CPU核心 0,2,4,6\nserver_cpulist 0-7:2\n\n# 後臺子線程綁定到 CPU核心 1,3\nbio_cpulist 1,3\n\n# 後臺 AOF rewrite 進程綁定到 CPU 核心 8,9,10,11\naof_rewrite_cpulist 8-11\n\n# 後臺 RDB 進程綁定到 CPU 核心 1,10,11\n# bgsave_cpulist 1,10-1","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":"如果你使用的正好是 Redis 6.0 版本，就可以通過以上配置，來進一步提高 Redis 性能。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這裏我需要提醒你的是，一般來說，Redis 的性能已經足夠優秀，除非你對 Redis 的性能有更加嚴苛的要求，否則不建議你綁定 CPU。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"從上面的分析你也能看出，綁定 CPU 需要你對計算機體系結構有非常清晰的瞭解，否則謹慎操作。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我們繼續分析還有什麼場景會導致 Redis 變慢。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"使用Swap","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你發現 Redis 突然變得非常慢，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"每次的操作耗時都達到了幾百毫秒甚至秒級","attrs":{}},{"type":"text","text":"，那此時你就需要檢查 Redis 是否使用到了 Swap，在這種情況下 Redis 基本上已經無法提供高性能的服務了。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"什麼是 Swap？爲什麼使用 Swap 會導致 Redis 的性能下降？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你對操作系統有些瞭解，就會知道操作系統爲了緩解內存不足對應用程序的影響，允許把一部分內存中的數據換到磁盤上，以達到應用程序對內存使用的緩衝，這些內存數據被換到磁盤上的區域，就是 Swap。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"問題就在於，當內存中的數據被換到磁盤上後，Redis 再訪問這些數據時，就需要從磁盤上讀取，訪問磁盤的速度要比訪問內存慢幾百倍！","attrs":{}}]},{"type":"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":"尤其是針對 Redis 這種對性能要求極高、性能極其敏感的數據庫來說，這個操作延時是無法接受的。","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":"此時，你需要檢查 Redis 機器的內存使用情況，確認是否存在使用了 Swap。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"你可以通過以下方式來查看 Redis 進程是否使用到了 Swap：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"# 先找到 Redis 的進程 ID\n$ ps -aux | grep redis-server\n\n# 查看 Redis Swap 使用情況\n$ cat /proc/$pid/smaps | egrep '^(Swap|Size)'","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"輸出結果如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"Size: 1256 kB\nSwap: 0 kB\nSize: 4 kB\nSwap: 0 kB\nSize: 132 kB\nSwap: 0 kB\nSize: 63488 kB\nSwap: 0 kB\nSize: 132 kB\nSwap: 0 kB\nSize: 65404 kB\nSwap: 0 kB\nSize: 1921024 kB\nSwap: 0 kB\n...","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這個結果會列出 Redis 進程的內存使用情況。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"每一行 Size 表示 Redis 所用的一塊內存大小，Size 下面的 Swap 就表示這塊 Size 大小的內存，有多少數據已經被換到磁盤上了，如果這兩個值相等，說明這塊內存的數據都已經完全被換到磁盤上了。","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":"如果只是少量數據被換到磁盤上，例如每一塊 Swap 佔對應 Size 的比例很小，那影響並不是很大。","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"如果是幾百兆甚至上 GB 的內存被換到了磁盤上","attrs":{}},{"type":"text","text":"，那麼你就需要警惕了，這種情況 Redis 的性能肯定會急劇下降。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此時的解決方案是：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"增加機器的內存，讓 Redis 有足夠的內存可以使用","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"整理內存空間，釋放出足夠的內存供 Redis 使用，然後釋放 Redis 的 Swap，讓 Redis 重新使用內存","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":"釋放 Redis 的 Swap 過程通常要重啓實例，爲了避免重啓實例對業務的影響，一般會先進行主從切換，然後釋放舊主節點的 Swap，重啓舊主節點實例，待從庫數據同步完成後，再進行主從切換即可。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"可見，當 Redis 使用到 Swap 後，此時的 Redis 性能基本已達不到高性能的要求（你可以理解爲武功被廢），所以你也需要提前預防這種情況。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"預防的辦法就是，你需要對 Redis 機器的內存和 Swap 使用情況進行監控，在內存不足或使用到 Swap 時報警出來，及時處理。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"碎片整理","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":"Redis 的數據都存儲在內存中，當我們的應用程序頻繁修改 Redis 中的數據時，就有可能會導致 Redis 產生內存碎片。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"內存碎片會降低 Redis 的內存使用率，我們可以通過執行 INFO 命令，得到這個實例的內存碎片率：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"# Memory\nused_memory:5709194824\nused_memory_human:5.32G\nused_memory_rss:8264855552\nused_memory_rss_human:7.70G\n...\nmem_fragmentation_ratio:1.45","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這個內存碎片率是怎麼計算的？","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":"很簡單，mem","attrs":{}},{"type":"text","marks":[{"type":"italic","attrs":{}}],"text":"fragmentation","attrs":{}},{"type":"text","text":"ratio = used","attrs":{}},{"type":"text","marks":[{"type":"italic","attrs":{}}],"text":"memory","attrs":{}},{"type":"text","text":"rss / used_memory。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"其中 used","attrs":{}},{"type":"text","marks":[{"type":"italic","attrs":{}}],"text":"memory 表示 Redis 存儲數據的內存大小，而 used","attrs":{}},{"type":"text","text":"memory_rss 表示操作系統實際分配給 Redis 進程的大小。","attrs":{}}]},{"type":"paragraph","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 mem","attrs":{}},{"type":"text","marks":[{"type":"italic","attrs":{}}],"text":"fragmentation","attrs":{}},{"type":"text","text":"ratio > 1.5，說明內存碎片率已經超過了 50%，這時我們就需要採取一些措施來降低內存碎片了。","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":"解決的方案一般如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":"1","normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"如果你使用的是 Redis 4.0 以下版本，只能通過重啓實例來解決","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"如果你使用的是 Redis 4.0 版本，它正好提供了自動碎片整理的功能，可以通過配置開啓碎片自動整理","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","marks":[{"type":"strong","attrs":{}}],"text":"但是，開啓內存碎片整理，它也有可能會導致 Redis 性能下降。","attrs":{}}]},{"type":"paragraph","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 的碎片整理工作是也在","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"主線程","attrs":{}},{"type":"text","text":"中執行的，當其進行碎片整理時，必然會消耗 CPU 資源，產生更多的耗時，從而影響到客戶端的請求。","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":"所以，當你需要開啓這個功能時，最好提前測試評估它對 Redis 的影響。","attrs":{}}]},{"type":"paragraph","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 碎片整理的參數配置如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":"shell"},"content":[{"type":"text","text":"# 開啓自動內存碎片整理（總開關）\nactivedefrag yes\n\n# 內存使用 100MB 以下，不進行碎片整理\nactive-defrag-ignore-bytes 100mb\n\n# 內存碎片率超過 10%，開始碎片整理\nactive-defrag-threshold-lower 10\n# 內存碎片率超過 100%，盡最大努力碎片整理\nactive-defrag-threshold-upper 100\n\n# 內存碎片整理佔用 CPU 資源最小百分比\nactive-defrag-cycle-min 1\n# 內存碎片整理佔用 CPU 資源最大百分比\nactive-defrag-cycle-max 25\n\n# 碎片整理期間，對於 List/Set/Hash/ZSet 類型元素一次 Scan 的數量\nactive-defrag-max-scan-fields 1000","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"你需要結合 Redis 機器的負載情況，以及應用程序可接受的延遲範圍進行評估，合理調整碎片整理的參數，儘可能降低碎片整理期間對 Redis 的影響。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"網絡帶寬過載","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果以上產生性能問題的場景，你都規避掉了，而且 Redis 也穩定運行了很長時間，但在某個時間點之後開始，操作 Redis 突然開始變慢了，而且一直持續下去，這種情況又是什麼原因導致？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此時你需要排查一下 Redis 機器的網絡帶寬是否過載，是否存在某個實例把整個機器的網路帶寬佔滿的情況。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"網絡帶寬過載的情況下，服務器在 TCP 層和網絡層就會出現數據包發送延遲、丟包等情況。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Redis 的高性能，除了操作內存之外，就在於網絡 IO 了，如果網絡 IO 存在瓶頸，那麼也會嚴重影響 Redis 的性能。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果確實出現這種情況，你需要及時確認佔滿網絡帶寬 Redis 實例，如果屬於正常的業務訪問，那就需要及時擴容或遷移實例了，避免因爲這個實例流量過大，影響這個機器的其他實例。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"運維層面，你需要對 Redis 機器的各項指標增加監控，包括網絡流量，在網絡流量達到一定閾值時提前報警，及時確認和擴容。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"其他原因","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":"好了，以上這些方面就是如何排查 Redis 延遲問題的思路和路徑。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"除了以上這些，還有一些比較小的點，你也需要注意一下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"1) 頻繁短連接","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"你的業務應用，應該使用長連接操作 Redis，避免頻繁的短連接。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"頻繁的短連接會導致 Redis 大量時間耗費在連接的建立和釋放上，TCP 的三次握手和四次揮手同樣也會增加訪問延遲。","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","marks":[{"type":"strong","attrs":{}}],"text":"2) 運維監控","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"前面我也提到了，要想提前預知 Redis 變慢的情況發生，必不可少的就是做好完善的監控。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"監控其實就是對採集 Redis 的各項運行時指標，通常的做法是監控程序定時採集 Redis 的 INFO 信息，然後根據 INFO 信息中的狀態數據做數據展示和報警。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這裏我需要提醒你的是，在寫一些監控腳本，或使用開源的監控組件時，也不能掉以輕心。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在寫監控腳本訪問 Redis 時，儘量採用長連接的方式採集狀態信息，避免頻繁短連接。同時，你還要注意控制訪問 Redis 的頻率，避免影響到業務請求。","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":"在使用一些開源的監控組件時，最好了解一下這些組件的實現原理，以及正確配置這些組件，防止出現監控組件發生 Bug，導致短時大量操作 Redis，影響 Redis 性能的情況發生。","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":"我們當時就發生過，DBA 在使用一些開源組件時，因爲配置和使用問題，導致監控程序頻繁地與 Redis 建立和斷開連接，導致 Redis 響應變慢。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"3）其它程序爭搶資源","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":"最後需要提醒你的是，你的 Redis 機器最好專項專用，只用來部署 Redis 實例，不要部署其他應用程序，儘量給 Redis 提供一個相對「安靜」的環境，避免其它程序佔用 CPU、內存、磁盤資源，導致分配給 Redis 的資源不足而受到影響。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"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}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"好了，以上就是我總結的在使用 Redis 過程中，常見的可能導致延遲、甚至阻塞的問題場景，以及如何快速定位和分析這些問題，並且針對性地提供瞭解決方案。","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":"這裏我也彙總成了思維導圖，方便你在排查 Redis 性能問題時，快速地去分析和定位。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/f3/f3ce2d49fd3f146a4ea12511a9fbff48.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},"content":[{"type":"text","text":"這裏再簡單總結一下，Redis 的性能問題，既涉及到了業務開發人員的使用方面，也涉及到了 DBA 的運維方面。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"作爲業務開發人員，我們需要了解 Redis 的基本原理，例如各個命令執行的時間複雜度、數據過期策略、數據淘汰策略等，從而更合理地使用 Redis 命令，並且結合業務場景進行優化。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"作爲 DBA 和運維人員，需要了解 Redis 運行機制，例如數據持久化、內存碎片整理、進程綁核配置。除此之外，還需要了解操作系統相關知識，例如寫時複製、內存大頁、Swap 機制等等。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"同時，DBA 在部署 Redis 時，需要提前對進行容量規劃，預留足夠的機器資源，還要對 Redis 機器和實例做好完善的監控，這樣才能儘可能地保證 Redis 的穩定運行。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"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}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你能耐心地看到這裏，想必你肯定已經對 Redis 的性能調優有了很大的收穫。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"你應該也發現了，Redis 的性能問題，涉及到的知識點非常廣，幾乎涵蓋了 CPU、內存、網絡、甚至磁盤的方方面面，同時，你還需要了解計算機的體系結構，以及操作系統的各種機制。","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":"從資源使用角度來看，包含的知識點如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"CPU 相關：使用複雜度過高命令、數據的持久化，都與耗費過多的 CPU 資源有關","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"內存相關：bigkey 內存的申請和釋放、數據過期、數據淘汰、碎片整理、內存大頁、內存寫時複製都與內存息息相關","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"磁盤相關：數據持久化、AOF 刷盤策略，也會受到磁盤的影響","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"網絡相關：短連接、實例流量過載、網絡流量過載，也會降低 Redis 性能","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"計算機系統：CPU 結構、內存分配，都屬於最基礎的計算機系統知識","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"操作系統：寫時複製、內存大頁、Swap、CPU 綁定，都屬於操作系統層面的知識","attrs":{}}]}],"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":"沒想到吧？Redis 爲了把性能做到極致，涉及到了這麼多項優化。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果這篇文章內容，你能吸收 90% 以上，說明你對 Redis 原理、計算機基礎、操作系統都已經有了較爲深刻的理解。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你能吸收 50% 左右，那你可以好好梳理一下，哪些方面是自己的知識盲區，這樣可以針對性地去學習。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你吸收的只在 30% 以下，那麼你可以先從 Redis 的基本原理出發，先了解 Redis 的各種機制，進而思考 Redis 爲了提高性能，爲什麼使用這些機制？這些機制又是利用了計算機和操作系統的哪些特性去做的？進而一步步地去擴充你的知識體系，這是一個非常高效的學習路徑。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"由於篇幅限制，關於 Redis 的很多細節無法全部展開，其實，這篇文章提到的每一個導致 Redis 性能問題的場景，如果展開來講，都可以寫出一篇文章出來。","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":"例如，關於 Redis 進程綁定 CPU，以及操作系統使用 Swap，其實這些還涉及到了非一致性內存訪問 NUMA 架構的影響，其中也有很多細節沒有展開來講。","attrs":{}}]},{"type":"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":"這篇文章主要站在一個宏觀層面進行性能分析，如果你想了解更多細節，歡迎關注我的公衆號「水滴與銀彈」，我會在後續的文章中，帶你深度剖析 Redis 的運行原理，以及產生性能問題的更多細節。","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":"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，用簡單的方式把技術講清楚。","attrs":{}}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}