[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"blog-strapi-article:zh:websocket-proxy-concurrent-connection-capacity":3,"blog-article:\u002Fblog\u002Fwebsocket-proxy-concurrent-connection-capacity":858,"blog-related:\u002Fblog\u002Fwebsocket-proxy-concurrent-connection-capacity":859,"blog-translations:\u002Fblog\u002Fwebsocket-proxy-concurrent-connection-capacity":860},{"available":4,"article":5,"relatedArticles":208,"translations":725},true,{"source":6,"id":7,"documentId":8,"translationKey":8,"locale":9,"slug":10,"path":11,"_path":11,"title":12,"description":13,"excerpt":13,"image":14,"publishedAt":15,"updatedAt":16,"author":17,"tags":18,"featured":19,"sortOrder":20,"readingTime":21,"bodyType":22,"bodyBlocks":23,"bodyHtml":207},"strapi",3973,"visywgnknoweo7zhs046zun6","zh","websocket-proxy-concurrent-connection-capacity","\u002Fblog\u002Fwebsocket-proxy-concurrent-connection-capacity","如何评估代理连接能够稳定承载多少 WebSocket 长连接","评估 WebSocket 代理链路容量时，应说明并发数、观察时长和断连指标的统计口径。本文介绍如何区分稳态长连接容量与建连速率，并依据客户端、代理和上游证据定位瓶颈。","https:\u002F\u002Fcms.quanqiudaili.com\u002Fuploads\u002Flarge_Web_Socket_cover_a59c46cc77.webp","2026-11-05T03:00:49.091Z","2026-11-05T03:00:49.078Z","Global Proxies Team",[],false,0,3,"blocks",[24,30,36,40,44,48,52,56,60,77,81,100,114,118,122,126,130,134,138,142,164,168,172,176,180,184,188,192,196],{"type":25,"children":26},"paragraph",[27],{"text":28,"type":29},"WebSocket 长连接容量需要同时说明并发数、持续时长和可接受的断连指标。报告“支持多少并发”而不说明观察窗口和连接状态，无法表示稳定性。测试应逐档加压并观察各层资源占用，识别先达到上限的是客户端、代理还是上游。","text",{"type":31,"level":32,"children":33},"heading",2,[34],{"text":35,"type":29},"WebSocket 容量必须结合并发、时长和断连率判断",{"type":25,"children":37},[38],{"text":39,"type":29},"三个参数必须一起定：并发连接数、保持时长和可接受的断连指标。断连指标还要说明分母，例如观察期内异常断开的会话比例，或每连接小时的断开频率。具体阈值应由业务目标确认。若多个业务共用账号、出口或 IP 池，实际配额范围需查当前产品授权，不能仅从测试结果推断。",{"type":25,"children":41},[42],{"text":43,"type":29},"同时明确测试对象：是客户端到代理的条数，还是包含上游服务的全链路。两者数字可能相差很多，结论里必须写清楚是哪一种。",{"type":31,"level":32,"children":45},[46],{"text":47,"type":29},"阶梯加压如何逐步暴露连接容量瓶颈",{"type":25,"children":49},[50],{"text":51,"type":29},"从基线开始，按固定步长或翻倍向上爬升。每档保持的时间要覆盖代理的空闲回收周期与心跳周期，否则看到的“稳定”只是还没走到回收时刻。档与档之间留冷却，让连接自然释放，再进入下一档。",{"type":25,"children":53},[54],{"text":55,"type":29},"一次打满不可取：所有客户端同时建连会形成握手风暴，断连集中的原因往往是被这种瞬时压力放大，而不是稳态容量不足。",{"type":25,"children":57},[58],{"text":59,"type":29},"阶梯参数建议固定下来，每次复测不改。",{"type":61,"format":62,"children":63},"list","unordered",[64,69,73],{"type":65,"children":66},"list-item",[67],{"text":68,"type":29},"参数：并发连接 ｜ 起点：由业务目标确定 ｜ 步长：按预先约定的固定步长 ｜ 每档持续：分开记录预热、稳态观察与正式保持 ｜ 判据：按事先约定的断连指标判定",{"type":65,"children":70},[71],{"text":72,"type":29},"参数：新建速率 ｜ 起点：单独设计建连\u002F重连实验 ｜ 步长：与稳态并发测试分开变化 ｜ 每档持续：记录各档建连持续时间 ｜ 判据：按建连成功率和拒绝证据判定",{"type":65,"children":74},[75],{"text":76,"type":29},"参数：保持与冷却 ｜ 起点：按业务场景确定 ｜ 步长：正式实验中保持一致 ｜ 每档持续：记录是否主动制造空闲、档后资源回落情况 ｜ 判据：确认上一档连接已释放或记录残留后再升档",{"type":31,"level":32,"children":78},[79],{"text":80,"type":29},"客户端、代理和上游都可能成为容量瓶颈",{"type":25,"children":82},[83,85,91,93,98],{"text":84,"type":29},"如果容量测试通过 HTTP 代理建立隧道，HTTP 代理的类型与企业接入方式可参考",{"text":86,"type":29,"marks":87},"HTTP 代理指南",[88],{"href":89,"type":90},"https:\u002F\u002Fwww.quanqiudaili.com\u002Fblog\u002Fhttp-proxy-enterprise-guide-v2","link",{"text":92,"type":29},"。具体并发配额及其作用范围，应以当前后台和授权说明为准；多出口结果不能在未确认账号或 IP 池是否共享额度时直接相加。测试指标的样本和稳定性口径可参照",{"text":94,"type":29,"marks":95},"代理 IP 可用率、延迟和稳定性测试",[96],{"href":97,"type":90},"https:\u002F\u002Fwww.quanqiudaili.com\u002Fblog\u002Fproxy-ip-availability-latency-stability-test",{"text":99,"type":29},"。",{"type":61,"format":62,"children":101},[102,106,110],{"type":65,"children":103},[104],{"text":105,"type":29},"层：客户端 ｜ 典型上限：文件描述符上限、本地端口耗尽、内存 ｜ 撞线时的表现：建连失败、无法创建连接 ｜ 该看什么：进程句柄与错误类型",{"type":65,"children":107},[108],{"text":109,"type":29},"层：代理侧 ｜ 典型上限：配额和新建速率限制（以产品说明为准） ｜ 撞线时的表现：新连接被拒、已有连接异常 ｜ 该看什么：代理侧连接 ID、拒绝事件与配额记录",{"type":65,"children":111},[112],{"text":113,"type":29},"层：上游服务 ｜ 典型上限：连接数上限、单用户并发限制 ｜ 撞线时的表现：连接建立后很快被断开 ｜ 该看什么：上游日志与关闭码",{"type":25,"children":115},[116],{"text":117,"type":29},"断连集中在哪一档、哪项资源先达到限制，可以帮助缩小排查范围，但阶段本身只是线索。应按连接 ID 对照客户端错误、代理侧拒绝事件和上游日志；握手失败也可能来自客户端配置、网络或上游鉴权，不能仅凭阶段断定来源。",{"type":25,"children":119},[120],{"text":121,"type":29},"单档结果至少记录目标并发、实际在线连接数、消息负载、心跳与重连行为、建连速率和资源占用。分别定义建连成功率（成功建立数\u002F尝试建立数）、异常断开会话比例（观察期内异常断开的会话数\u002F纳入观察的会话数）或每连接小时断开频率，不能将这些指标统称为“断连率”。",{"type":31,"level":32,"children":123},[124],{"text":125,"type":29},"重连风暴为何会放大代理侧压力",{"type":25,"children":127},[128],{"text":129,"type":29},"大量连接同时断开后，客户端的重连策略可能形成新的建连峰值。容量实验应记录该策略的退避、随机抖动和恢复时间；若目标是测稳态长连接容量，应固定 ramp-up 速率并另行测试建连或重连速率，避免同时改变两个变量。",{"type":25,"children":131},[132],{"text":133,"type":29},"否则测出来的不是容量上限，而是重连策略的上限，两者不能混为一谈。",{"type":31,"level":32,"children":135},[136],{"text":137,"type":29},"容量结论必须说明测试条件和实际瓶颈",{"type":25,"children":139},[140],{"text":141,"type":29},"结论只适用于记录的客户端版本、网络条件、代理配置和测试负载。报告并发档位、实际在线数、保持时长、指标分母、异常断开数、瓶颈证据及重连配置；账号、出口和 IP 池的配额范围应从当前后台或授权说明确认，不替服务商推定并发能力。",{"type":61,"format":62,"children":143},[144,148,152,156,160],{"type":65,"children":145},[146],{"text":147,"type":29},"记录项：档位与并发数 ｜ 填写内容：目标并发、实际建连\u002F在线数、固定 ramp-up 速率",{"type":65,"children":149},[150],{"text":151,"type":29},"记录项：保持与冷却 ｜ 填写内容：预热、稳态观察、正式保持时间、连接资源回落情况",{"type":65,"children":153},[154],{"text":155,"type":29},"记录项：断连统计 ｜ 填写内容：异常断开会话数、分母与所用频率口径",{"type":65,"children":157},[158],{"text":159,"type":29},"记录项：重连表现 ｜ 填写内容：重连次数、退避是否生效、恢复时长",{"type":65,"children":161},[162],{"text":163,"type":29},"记录项：瓶颈判断 ｜ 填写内容：客户端、代理侧、上游，依据哪条证据",{"type":31,"level":32,"children":165},[166],{"text":167,"type":29},"FAQ",{"type":31,"level":21,"children":169},[170],{"text":171,"type":29},"连接都能建立，为什么很快就断？",{"type":25,"children":173},[174],{"text":175,"type":29},"可能原因包括客户端资源限制、代理或上游策略、心跳与空闲回收不匹配。连接阶段和关闭信息只能提供线索，应结合客户端、代理与上游对应日志或可控测试记录核实。",{"type":31,"level":21,"children":177},[178],{"text":179,"type":29},"容量测试要不要换多个出口 IP？",{"type":25,"children":181},[182],{"text":183,"type":29},"是否需要多个出口取决于测试目标。若评估单出口能力，应分别测量；若评估多出口整体能力，应先确认是否共用账号、池或速率配额，再说明聚合方法。更换出口时固定其他条件，并记录每次实际分配结果。",{"type":31,"level":21,"children":185},[186],{"text":187,"type":29},"如何判断是代理上限还是上游上限？",{"type":25,"children":189},[190],{"text":191,"type":29},"按连接 ID 对齐客户端错误、代理侧响应来源和上游记录。握手阶段失败只是线索，连接建立后的断开也可能来自代理或客户端；HTTP 代理认证、代理连接和上游关闭应分别按可用证据判断。",{"type":25,"children":193},[194],{"text":195,"type":29},"---",{"type":25,"children":197},[198,200,205],{"text":199,"type":29},"容量测试结果应注明连接数、保持时长、断连指标及分母、资源负载和瓶颈证据。接入参数可查阅",{"text":201,"type":29,"marks":202},"代理设置帮助",[203],{"href":204,"type":90},"https:\u002F\u002Fhelp.quanqiudaili.com\u002Fdocs\u002Fproxy-Settings",{"text":206,"type":29},"。原并发规划和故障排查专题链接目前返回 404，暂不使用。","",[209,400,555],{"source":6,"id":210,"documentId":211,"translationKey":211,"locale":9,"slug":212,"path":213,"_path":213,"title":214,"description":215,"excerpt":215,"image":216,"publishedAt":217,"updatedAt":218,"author":17,"tags":219,"featured":19,"sortOrder":20,"readingTime":32,"bodyType":22,"bodyBlocks":220,"bodyHtml":207},3977,"t6hh3bpjb00uvfiyfm3u1wb3","proxy-application-failure-injection-rollback-drill","\u002Fblog\u002Fproxy-application-failure-injection-rollback-drill","如何为使用代理的应用设计故障注入和回滚演练","代理故障演练用于验证应用在连接、认证或出口异常时的降级与恢复表现。本文说明如何依据业务目标设定通过条件、控制演练影响范围，并检查回滚后的服务状态。","https:\u002F\u002Fcms.quanqiudaili.com\u002Fuploads\u002Flarge_cover_dbecee53d1.webp","2026-11-05T03:14:30.990Z","2026-11-05T03:14:30.969Z",[],[221,225,229,233,237,241,257,279,283,287,291,295,299,303,307,311,315,319,323,327,331,335,361,364,368,372,376,380,384,388,391],{"type":25,"children":222},[223],{"text":224,"type":29},"代理故障演练应验证应用在连接失败、认证异常或出口变化时的处理是否符合约定，并在故障解除后恢复。开始前由业务或 SLO 明确可接受的错误率、降级范围、数据损失目标和恢复时限；文中的固定数值只能作为待确认的示例。",{"type":31,"level":32,"children":226},[227],{"text":228,"type":29},"故障注入的通过标准应包含哪些指标",{"type":25,"children":230},[231],{"text":232,"type":29},"通过标准至少包含三项：降级级别（功能降级、只读、排队、拒绝）、可接受的错误率、从故障到恢复的时长。三项写清楚，演练才有“通过”与“不通过”的区分。",{"type":25,"children":234},[235],{"text":236,"type":29},"没有标准时，演练结束只能记录“发生了什么”，无法判断应用的行为是否可接受。",{"type":31,"level":32,"children":238},[239],{"text":240,"type":29},"一次只注入一个依赖层，故障结果才容易归因",{"type":25,"children":242},[243,245,248,250,255],{"text":244,"type":29},"如果注入项涉及代理协议或客户端接入配置，可按项目实际采用的协议参考",{"text":86,"type":29,"marks":246},[247],{"href":89,"type":90},{"text":249,"type":29},"或",{"text":251,"type":29,"marks":252},"SOCKS5 代理指南",[253],{"href":254,"type":90},"https:\u002F\u002Fwww.quanqiudaili.com\u002Fblog\u002Fsocks5-proxy-guide-v2",{"text":256,"type":29},"，再单独注入对应配置故障。两种协议分别选择，不需同时添加。",{"type":61,"format":62,"children":258},[259,263,267,271,275],{"type":65,"children":260},[261],{"text":262,"type":29},"注入层：DNS 解析 ｜ 模拟方式：指向无效解析服务 ｜ 客户端表现：解析失败或超时 ｜ 期望的降级行为：切备用解析或返回缓存结果",{"type":65,"children":264},[265],{"text":266,"type":29},"注入层：建连 ｜ 模拟方式：在隔离测试环境阻断到代理的连接 ｜ 客户端表现：按客户端记录超时或连接失败 ｜ 期望的降级行为：按已确认的重试和故障切换策略处理",{"type":65,"children":268},[269],{"text":270,"type":29},"注入层：代理认证 ｜ 模拟方式：使用失效凭据 ｜ 客户端表现：407 ｜ 期望的降级行为：提示链路异常，不重试业务凭据",{"type":65,"children":272},[273],{"text":274,"type":29},"注入层：上游服务 ｜ 模拟方式：返回 5xx 或延迟 ｜ 客户端表现：502、504 或读超时 ｜ 期望的降级行为：熔断、排队或返回降级内容",{"type":65,"children":276},[277],{"text":278,"type":29},"注入层：代理资源不可用（测试桩） ｜ 模拟方式：由自建测试桩模拟资源不足 ｜ 客户端表现：使用测试桩约定的响应 ｜ 期望的降级行为：按测试桩契约验证处理",{"type":25,"children":280},[281],{"text":282,"type":29},"每层单独注入，才便于判断降级策略在哪一层失效。测试桩模拟的状态码应与真实产品行为分开记录；真实 429 的来源、错误体和含义依代理、网关或目标服务契约而定。应结合响应来源、配额范围和分配事件判断，不凭 429 自动换池或出口，也不消耗真实配额来制造资源耗尽。",{"type":31,"level":32,"children":284},[285],{"text":286,"type":29},"注入时长和流量范围如何控制故障影响",{"type":25,"children":288},[289],{"text":290,"type":29},"单次注入只改一个变量，限定测试环境、流量范围、账号和时间窗，并准备停止与回滚方式。生产演练只有在对象、流量、影响范围和停止条件均获明确授权后才能安排；仅限制为配置变更不能替代授权。",{"type":25,"children":292},[293],{"text":294,"type":29},"生产环境演练除回滚能力和相关方通知外，还要明确获准影响的服务、流量和时间窗，以及监控、停止和恢复条件。未满足既定授权与范围要求时，应在隔离测试环境演练。",{"type":31,"level":32,"children":296},[297],{"text":298,"type":29},"告警、自动降级和数据补偿分别验证什么",{"type":31,"level":21,"children":300},[301],{"text":302,"type":29},"告警触发时间能否满足预期",{"type":25,"children":304},[305],{"text":306,"type":29},"记录从注入到告警触发的时间差；未触发本身也是一个结论，说明监控阈值或覆盖面有缺口。只记录“告警响了”没有意义，要记录它比预期晚了多久。",{"type":31,"level":21,"children":308},[309],{"text":310,"type":29},"自动降级是否在故障后及时生效",{"type":25,"children":312},[313],{"text":314,"type":29},"区分自动降级与人工介入：自动降级看是否在预期时间内生效，人工介入则要记录从发现到处置的耗时。两者混在一起，会把“人救回来的”误当成系统能力。",{"type":31,"level":21,"children":316},[317],{"text":318,"type":29},"故障期间产生的数据能否补偿",{"type":25,"children":320},[321],{"text":322,"type":29},"按业务契约检查请求是否幂等、是否可查询执行结果，以及允许的数据损失或补偿范围。数据不可重放不一定意味着演练失败；应依据事先批准的 RPO、只读\u002F拒绝\u002F补偿目标判定。",{"type":25,"children":324},[325],{"text":326,"type":29},"三件事只测其中两件，演练的价值会大打折扣。",{"type":31,"level":32,"children":328},[329],{"text":330,"type":29},"回滚后为什么还要重跑基线",{"type":25,"children":332},[333],{"text":334,"type":29},"配置是否回滚、连接是否恢复、数据是否核对一致、恢复耗时是否在标准内。回滚完成后必须重跑一遍基线，确认不是“看起来恢复”。",{"type":61,"format":62,"children":336},[337,341,345,349,353,357],{"type":65,"children":338},[339],{"text":340,"type":29},"检查项：配置回滚 ｜ 通过判据：配置与应用版本回到基线 ｜ 记录内容：回滚时间、操作人",{"type":65,"children":342},[343],{"text":344,"type":29},"检查项：连接恢复 ｜ 通过判据：建连成功、心跳正常 ｜ 记录内容：恢复耗时、失败次数",{"type":65,"children":346},[347],{"text":348,"type":29},"检查项：错误率恢复 ｜ 通过判据：按事先约定的 SLO、样本和观察窗口判断 ｜ 记录内容：错误率曲线、失败样本、恢复耗时",{"type":65,"children":350},[351],{"text":352,"type":29},"检查项：告警关闭 ｜ 通过判据：相关告警解除、无新增告警 ｜ 记录内容：告警时间线",{"type":65,"children":354},[355],{"text":356,"type":29},"检查项：数据一致 ｜ 通过判据：无重复、无缺失 ｜ 记录内容：核对方式与差异数",{"type":65,"children":358},[359],{"text":360,"type":29},"检查项：基线复跑 ｜ 通过判据：按事先约定的请求集和容差判断 ｜ 记录内容：通过率、异常项与证据",{"type":31,"level":32,"children":362},[363],{"text":167,"type":29},{"type":31,"level":21,"children":365},[366],{"text":367,"type":29},"故障注入和生产演练的边界在哪？",{"type":25,"children":369},[370],{"text":371,"type":29},"默认在自有应用和隔离测试环境中注入故障。生产演练必须具备明确授权，限定服务对象、流量、影响范围、时间窗和停止\u002F回滚条件；只在配置层操作并不自动代表安全或获准。",{"type":31,"level":21,"children":373},[374],{"text":375,"type":29},"为什么回滚后还要重跑基线？",{"type":25,"children":377},[378],{"text":379,"type":29},"回滚只证明配置已恢复，不证明连接池、缓存、队列和业务状态都已恢复。应按演练前约定的请求集、错误率容差和数据一致性目标复测，并保留未通过的样本；通过标准不应临时设为固定的“全部成功”。",{"type":31,"level":21,"children":381},[382],{"text":383,"type":29},"演练频率怎么定？",{"type":25,"children":385},[386],{"text":387,"type":29},"按依赖变更频率定：换了代理出口、升级了客户端、调整了超时或重试策略之后都应重跑。稳定期可以按季度做一次全量演练。",{"type":25,"children":389},[390],{"text":195,"type":29},{"type":25,"children":392},[393,395,398],{"text":394,"type":29},"一次完整演练应能说明故障是否被发现、应用是否按标准降级，以及回滚后服务和数据是否恢复。客户端代理配置可参阅",{"text":201,"type":29,"marks":396},[397],{"href":204,"type":90},{"text":399,"type":29},"。原 407 排障和故障排查中心链接当前返回 404，不保留失效目标。",{"source":6,"id":401,"documentId":402,"translationKey":402,"locale":9,"slug":403,"path":404,"_path":404,"title":405,"description":406,"excerpt":406,"image":407,"publishedAt":408,"updatedAt":409,"author":17,"tags":410,"featured":19,"sortOrder":20,"readingTime":32,"bodyType":22,"bodyBlocks":411,"bodyHtml":207},3976,"t89caecwwq76wbohgzryvs5k","websocket-heartbeat-idle-timeout-testing","\u002Fblog\u002Fwebsocket-heartbeat-idle-timeout-testing","如何为实时应用验证 WebSocket 心跳和空闲超时阈值","WebSocket 空闲连接可能受客户端心跳、代理或 NAT 回收、服务端探测及应用策略影响。本文说明如何按客户端能力观测这些行为，并依据实际连接事件验证心跳和故障检测阈值。","https:\u002F\u002Fcms.quanqiudaili.com\u002Fuploads\u002Flarge_Web_Socket_cover_8eac762518.webp","2026-11-05T03:06:49.468Z","2026-11-05T03:06:49.444Z",[],[412,416,420,434,438,442,446,450,454,458,462,466,470,474,478,482,486,490,494,516,519,523,527,531,535,539,543,546],{"type":25,"children":413},[414],{"text":415,"type":29},"WebSocket 空闲连接断开，往往是客户端心跳、代理或 NAT 空闲回收、服务端 Ping 周期中的某个阈值先到。心跳间隔需要结合业务可接受的无流量时长和中间层回收时间设定，再通过长时间观测检查误断率。",{"type":31,"level":32,"children":417},[418],{"text":419,"type":29},"客户端、代理和服务端的空闲超时如何相互影响",{"type":61,"format":62,"children":421},[422,426,430],{"type":65,"children":423},[424],{"text":425,"type":29},"层：客户端探测 ｜ 常见行为：业务消息心跳或客户端库支持的协议探测 ｜ 如何观测：客户端事件、消息和时间戳 ｜ 谁决定断开：浏览器 JS 通常不能直接发送\u002F接收 WebSocket 协议 Ping\u002FPong",{"type":65,"children":427},[428],{"text":429,"type":29},"层：代理或 NAT ｜ 常见行为：空闲连接超过阈值后回收 ｜ 如何观测：断开时刻与最后流量时间 ｜ 谁决定断开：代理侧策略",{"type":65,"children":431},[432],{"text":433,"type":29},"层：服务端 ping ｜ 常见行为：服务端定时探测并清理 ｜ 如何观测：服务端连接日志与关闭码 ｜ 谁决定断开：服务端配置",{"type":25,"children":435},[436],{"text":437,"type":29},"某一层配置了空闲回收或连接失败时，连接可能断开；不同实现重置计时的流量类型也不同。要测实际阈值，应在受控环境分别暂停业务流量、应用心跳和协议探测，并记录代理或 NAT 策略中已知和未知的部分。不能把一次长时间保持成功当成所有层阈值已测出。",{"type":31,"level":32,"children":439},[440],{"text":441,"type":29},"心跳间隔应如何匹配最短空闲回收时间",{"type":25,"children":443},[444],{"text":445,"type":29},"如果某层的空闲回收阈值已知，业务心跳间隔通常需要短于该阈值，并留出时钟抖动和传输延迟余量。具体计算取决于哪些字节会重置该层计时、允许丢失次数和客户端调度精度，需通过实际环境验证。",{"type":25,"children":447},[448],{"text":449,"type":29},"在“每次心跳都会重置空闲计时、允许连续丢失固定次数、发送间隔稳定”的假设下，可以用最短已知回收时间估算初始间隔，再预留抖动余量。该估算不是通用阈值。心跳发送间隔、等待 Pong 的时间和连续丢失后的故障检测 deadline 是不同参数，应分别定义；不要用客户端读超时代替故障检测期限。",{"type":25,"children":451},[452],{"text":453,"type":29},"服务端协议 Ping、自动 Pong 和应用层业务心跳验证的是不同层级，不一定重复。应先明确要验证连接存活、应用授权还是业务消息处理，再决定保留哪些探测，并记录每种探测的方向、发送端和可观测证据。收到 Pong 只说明对应协议端点作出响应，不足以证明业务健康。",{"type":31,"level":32,"children":455},[456],{"text":457,"type":29},"长时间观测才能覆盖完整的空闲回收周期",{"type":25,"children":459},[460],{"text":461,"type":29},"先建立受控基线并记录全部流量。分别暂停业务消息、应用心跳和可控的协议探测，在获准且可配置的层单独改变空闲策略，以多档无流量时长重复观察并夹逼阈值；对无法控制的代理\u002FNAT层标记为未知。恢复原配置后，再测试正常保持和预期故障检测时间。记录本地关闭事件、实际收到的 Close 帧、两端日志和重连耗时；单纯持续心跳可能不断重置计时，不能用一个观察周期推断各层阈值。",{"type":25,"children":463},[464],{"text":465,"type":29},"只有一轮、只有一个网络条件，得到的是单次经验，不是阈值。",{"type":31,"level":32,"children":467},[468],{"text":469,"type":29},"关闭码与 Ping\u002FPong 能说明什么",{"type":31,"level":21,"children":471},[472],{"text":473,"type":29},"常见关闭码分别说明哪类 WebSocket 断开",{"type":25,"children":475},[476],{"text":477,"type":29},"1006 是本地用于表示异常关闭且未收到正常 Close 帧的保留码，不是线上 Close 帧，也不指明根因。1001 表示端点声明离开，但不保证重连或恢复成功。自定义码的含义应查应用协议与服务端定义，并结合本地关闭事件、实际 Close 帧及两端日志判断。",{"type":31,"level":21,"children":479},[480],{"text":481,"type":29},"Ping\u002FPong 往返情况如何反映连接层状态",{"type":25,"children":483},[484],{"text":485,"type":29},"只有能观测协议 Ping\u002FPong 的客户端或抓取点，才可将它们作为协议层证据；浏览器脚本通常不能直接读取这些控制帧。按方向和时间记录 Ping\u002FPong 与业务心跳。Pong 到达只证明对应协议端点响应，不代表应用订阅、授权或业务处理正常；Pong 缺失也需结合连接、服务端和代理事件排查。",{"type":31,"level":32,"children":487},[488],{"text":489,"type":29},"心跳过于频繁会增加多少流量成本",{"type":25,"children":491},[492],{"text":493,"type":29},"估算心跳流量时，应计入双向报文、WebSocket 帧、TCP\u002FTLS 开销和实际计费口径。简单的“连接数 × 频率 × 单次报文大小”只能作为负载粗估，不等于供应商账单；应注明假设并用实际计量复核。",{"type":61,"format":62,"children":495},[496,500,504,508,512],{"type":65,"children":497},[498],{"text":499,"type":29},"记录项：探测配置 ｜ 填写内容：应用心跳、服务端协议 Ping、客户端能力与各自超时判定",{"type":65,"children":501},[502],{"text":503,"type":29},"记录项：观测条件 ｜ 填写内容：网络类型、时长、并发连接数",{"type":65,"children":505},[506],{"text":507,"type":29},"记录项：断开情况 ｜ 填写内容：断开时刻、关闭码、最后流量时间",{"type":65,"children":509},[510],{"text":511,"type":29},"记录项：恢复情况 ｜ 填写内容：重连耗时、是否重试成功",{"type":65,"children":513},[514],{"text":515,"type":29},"记录项：成本估算 ｜ 填写内容：心跳频率、单次报文大小、日流量估算",{"type":31,"level":32,"children":517},[518],{"text":167,"type":29},{"type":31,"level":21,"children":520},[521],{"text":522,"type":29},"心跳间隔设多少合适？",{"type":25,"children":524},[525],{"text":526,"type":29},"没有统一值。先确认哪些流量会重置相关空闲计时，再根据已知回收阈值、客户端调度误差和业务允许的故障检测时间设定初值，并在目标客户端、后台运行和弱网条件下测误断率。任何比例公式都应标明假设，并通过目标环境验证。",{"type":31,"level":21,"children":528},[529],{"text":530,"type":29},"为什么空闲一段时间后连接可能会断？",{"type":25,"children":532},[533],{"text":534,"type":29},"如果客户端、代理、NAT、服务端或应用配置了空闲回收，达到相应条件时连接可能被关闭；也可能因网络故障、进程退出或授权变化而断开。心跳只能影响会被该层计入的流量计时，不能保证连接始终存活。",{"type":31,"level":21,"children":536},[537],{"text":538,"type":29},"双重心跳有什么问题？",{"type":25,"children":540},[541],{"text":542,"type":29},"是否需要两种心跳取决于探测目标：协议 Ping\u002FPong 可用于连接层探测，应用消息心跳可验证业务处理。两者可能同时有用，也可能造成额外流量；应按层级分别配置与观测，不能一概删除其中一种。",{"type":25,"children":544},[545],{"text":195,"type":29},{"type":25,"children":547},[548,550,553],{"text":549,"type":29},"连接意外断开时，可对照关闭码、最后一次心跳时间和空闲回收阈值。比较连接可用率、延迟与稳定性时，可参考",{"text":94,"type":29,"marks":551},[552],{"href":97,"type":90},{"text":554,"type":29},"。原故障排查中心链接当前返回 404，不保留该失效目标。",{"source":6,"id":556,"documentId":557,"translationKey":557,"locale":9,"slug":558,"path":559,"_path":559,"title":560,"description":561,"excerpt":561,"image":562,"publishedAt":563,"updatedAt":564,"author":17,"tags":565,"featured":19,"sortOrder":20,"readingTime":21,"bodyType":22,"bodyBlocks":566,"bodyHtml":207},3975,"jddv5ul0y7azfjmzeejqz05u","proxy-api-request-timeout-budget-testing","\u002Fblog\u002Fproxy-api-request-timeout-budget-testing","如何为经代理的 API 请求拆分和验证超时预算","经代理的 API 请求需要区分客户端实际支持的连接、读取和端到端超时。本文说明如何按客户端版本记录超时覆盖范围、安排总 deadline 与重试，并用阶段日志定位延迟来源。","https:\u002F\u002Fcms.quanqiudaili.com\u002Fuploads\u002Flarge_API_cover_fa6f8d6718.webp","2026-11-05T03:01:23.422Z","2026-11-05T03:01:23.393Z",[],[567,571,575,583,609,613,617,621,625,629,633,637,641,645,649,653,657,661,665,669,691,694,698,702,706,710,714,718,721],{"type":25,"children":568},[569],{"text":570,"type":29},"代理 API 请求可能在 DNS 解析、建立连接、TLS 握手、等待首字节或读取响应时耗尽时间。将总超时拆分到这些阶段，并计入代理认证与重试开销，可以根据失败阶段定位延迟来源，也能避免单次请求无限占用资源。",{"type":31,"level":32,"children":572},[573],{"text":574,"type":29},"API 请求总超时由哪五段组成",{"type":25,"children":576},[577,579,582],{"text":578,"type":29},"以下阶段用于分析一次请求的耗时，并不代表客户端都能为每段单独配置超时。应选定客户端及版本，核对其连接池等待、DNS、TCP、代理协商、TLS、写入、读取和端到端 deadline 参数实际覆盖哪些阶段。HTTP 代理的接入背景可参考",{"text":86,"type":29,"marks":580},[581],{"href":89,"type":90},{"text":99,"type":29},{"type":61,"format":62,"children":584},[585,589,593,597,601,605],{"type":65,"children":586},[587],{"text":588,"type":29},"耗时阶段：连接池等待 ｜ 可观测字段：排队\u002F获取连接耗时 ｜ 客户端参数需核对的范围：是否有独立 pool-acquire 超时 ｜ 证据边界：按客户端库诊断信息记录",{"type":65,"children":590},[591],{"text":592,"type":29},"耗时阶段：DNS 与 TCP ｜ 可观测字段：解析、到代理的 TCP 建连时间 ｜ 客户端参数需核对的范围：是否由 connect timeout 覆盖，是否能分开配置 ｜ 证据边界：代理模式下客户端通常连接代理；目标解析位置依协议和实现确认",{"type":65,"children":594},[595],{"text":596,"type":29},"耗时阶段：代理协商与 TLS ｜ 可观测字段：CONNECT\u002F认证结果、TLS 握手时间 ｜ 客户端参数需核对的范围：客户端是否暴露独立阶段超时 ｜ 证据边界：TLS 隧道内内容对普通代理不可见",{"type":65,"children":598},[599],{"text":600,"type":29},"耗时阶段：请求写入 ｜ 可观测字段：写入耗时、发送字节 ｜ 客户端参数需核对的范围：是否有 write timeout ｜ 证据边界：按客户端实际参数与传输事件记录",{"type":65,"children":602},[603],{"text":604,"type":29},"耗时阶段：响应等待与读取 ｜ 可观测字段：首字节时间、读取间隔与响应体耗时 ｜ 客户端参数需核对的范围：read timeout 是阶段总时长还是无数据间隔 ｜ 证据边界：结合响应头、首字节和读取中断事件说明",{"type":65,"children":606},[607],{"text":608,"type":29},"耗时阶段：端到端调用 ｜ 可观测字段：从逻辑调用开始到返回\u002F取消 ｜ 客户端参数需核对的范围：是否有单一总 deadline ｜ 证据边界：应覆盖排队、解析、认证、传输、重试和退避",{"type":25,"children":610},[611],{"text":612,"type":29},"阶段日志有助于缩小延迟范围，但只有客户端确实支持相应参数时才能单独设定阶段超时。read timeout 可能表示连续无数据的等待上限，而非整个响应读取时长；应按库文档说明，并验证超时或取消后连接与响应资源是否释放。",{"type":31,"level":32,"children":614},[615],{"text":616,"type":29},"为什么代理请求超时应参考 p95 和 p99",{"type":25,"children":618},[619],{"text":620,"type":29},"平均值可能掩盖长尾。预算可以参考端到端耗时分布、目标失败率和资源约束，并报告样本量与采样窗口；各阶段 p95\u002Fp99 可辅助分配预算，但不能简单相加后当作端到端同分位数保证。",{"type":25,"children":622},[623],{"text":624,"type":29},"阶段预算应结合实测分布、客户端参数语义和业务失败容忍度设定，并定期用新样本复核。超时请求属于截尾样本，应记录失败与超时情况，不能只用成功请求估算分布。不同代理产品或出口类型的耗时差异需要在相同客户端、目标、并发和时段条件下实测，不能预设动态出口必然慢于静态出口。",{"type":31,"level":32,"children":626},[627],{"text":628,"type":29},"代理连接和重试会怎样增加请求总耗时",{"type":25,"children":630},[631],{"text":632,"type":29},"一次逻辑调用的总耗时包括排队、每次尝试的解析与传输、认证处理、重试间隔和退避。应区分“总尝试次数”和“额外重试次数”：例如最多重试三次通常意味着初次尝试加三次重试，而不是总共三次。记录每次实际耗时和退避时间，避免表格与说明使用不同口径。",{"type":25,"children":634},[635],{"text":636,"type":29},"端到端 deadline 应从逻辑调用开始只创建一次，并包含连接池等待、每次尝试、退避和取消时间。每个阶段的可用时限取阶段上限与剩余总预算中的较小值；不得在重试时重置总 deadline。认证失败和参数错误通常不应盲目重试，是否重试需按客户端和服务契约确认。",{"type":31,"level":32,"children":638},[639],{"text":640,"type":29},"错误类型可以帮助定位超时发生的阶段",{"type":25,"children":642},[643],{"text":644,"type":29},"超时名称只能说明客户端报告的等待阶段，不能单独确定责任方。read timeout 可能发生在等待响应头\u002F首字节之前，也可能发生在读取响应过程中；407 是代理认证响应，不是连接超时。遇到 502\u002F504 应记录实际返回组件。通过客户端阶段事件、代理日志和源站记录核实候选原因。",{"type":25,"children":646},[647],{"text":648,"type":29},"把错误类型与阶段对照起来，排查就不用从零开始猜。建议在客户端日志里同时记录阶段与耗时字段，而不是只记一个总耗时。",{"type":31,"level":32,"children":650},[651],{"text":652,"type":29},"端到端验证如何检查预算并触发降级",{"type":25,"children":654},[655],{"text":656,"type":29},"若客户端不支持每段独立超时，就记录它能配置的参数及未覆盖阶段，并使用一次性的端到端 deadline 控制逻辑调用。总 deadline 应与上游调用方的时限协调，预留处理和返回结果的时间；超时或取消后还要检查请求是否继续占用连接或资源。",{"type":25,"children":658},[659],{"text":660,"type":29},"部分可用场景下，可以先返回部分结果并标注缺失，而不是整体失败。",{"type":25,"children":662},[663],{"text":664,"type":29},"校验的对象是总预算，不是降级逻辑：构造慢响应与断流场景，确认总时限在预期内触发，并记录实际触发的是哪一段。",{"type":25,"children":666},[667],{"text":668,"type":29},"以下数值仅用于说明如何记录阶段预算，并不代表通用参数或某个客户端一定支持这些独立设置。实际字段应按选定客户端及版本替换；无法独立配置的阶段标为“由其他参数覆盖”或“不可单独设置”。",{"type":61,"format":62,"children":670},[671,675,679,683,687],{"type":65,"children":672},[673],{"text":674,"type":29},"参数\u002F阶段：连接池等待 ｜ 示例记录值：由客户端配置确认 ｜ 计时语义：获取连接前的排队时间 ｜ 客户端版本与覆盖范围：填写实际参数或“不支持独立设置”",{"type":65,"children":676},[677],{"text":678,"type":29},"参数\u002F阶段：connect timeout ｜ 示例记录值：由客户端配置确认 ｜ 计时语义：TCP\u002F代理连接阶段，具体范围查版本文档 ｜ 客户端版本与覆盖范围：记录是否包含 DNS 或代理协商",{"type":65,"children":680},[681],{"text":682,"type":29},"参数\u002F阶段：read timeout ｜ 示例记录值：由客户端配置确认 ｜ 计时语义：可能是无数据等待间隔，不一定是整个响应期限 ｜ 客户端版本与覆盖范围：记录首字节前与读取中的行为",{"type":65,"children":684},[685],{"text":686,"type":29},"参数\u002F阶段：总 deadline ｜ 示例记录值：由业务目标确认 ｜ 计时语义：从逻辑调用发起到完成\u002F取消，仅创建一次 ｜ 客户端版本与覆盖范围：包含全部尝试和退避",{"type":65,"children":688},[689],{"text":690,"type":29},"参数\u002F阶段：重试 ｜ 示例记录值：明确初次尝试 + 额外重试次数 ｜ 计时语义：每次记录耗时与退避 ｜ 客户端版本与覆盖范围：受同一个总 deadline 限制",{"type":31,"level":32,"children":692},[693],{"text":167,"type":29},{"type":31,"level":21,"children":695},[696],{"text":697,"type":29},"为什么平均耗时不能用来定预算？",{"type":25,"children":699},[700],{"text":701,"type":29},"平均值被少数快请求拉低，掩盖了长尾。超时设计关心的是最慢的那部分请求，必须看 p95 与 p99，并说明采样窗口。",{"type":31,"level":21,"children":703},[704],{"text":705,"type":29},"代理超时和上游超时怎么区分？",{"type":25,"children":707},[708],{"text":709,"type":29},"407 表示代理认证挑战，不代表代理连接超时。应核对客户端实际连接阶段和响应来源，再对齐代理及目标服务日志；没有首字节可能涉及上游处理、代理转发、TLS\u002F协议协商或中间层排队，不能仅凭等待阶段归因。",{"type":31,"level":21,"children":711},[712],{"text":713,"type":29},"重试次数该怎么算进预算？",{"type":25,"children":715},[716],{"text":717,"type":29},"先明确总尝试次数等于初次尝试加额外重试次数。每次尝试和退避都消耗同一个端到端 deadline 的剩余时间；阶段预算取自身上限与剩余 deadline 的较小值。若客户端没有总 deadline，只能将多次尝试的放大作为该配置的风险示例，不能和有总 deadline 的策略混用。",{"type":25,"children":719},[720],{"text":195,"type":29},{"type":25,"children":722},[723],{"text":724,"type":29},"记录各阶段耗时后，可以判断瓶颈出现在解析、代理连接还是上游响应。原 TLS\u002F超时排障和重试策略链接当前返回 404；可先依据本文所列客户端版本文档核对具体参数，待站内存在有效专题页后再补充内链。",[726],{"source":6,"id":7,"documentId":8,"translationKey":8,"locale":9,"slug":10,"path":11,"_path":11,"title":12,"description":13,"excerpt":13,"image":14,"publishedAt":15,"updatedAt":16,"author":17,"tags":727,"featured":19,"sortOrder":20,"readingTime":21,"bodyType":22,"bodyBlocks":728,"bodyHtml":207},[],[729,732,735,738,741,744,747,750,753,764,767,778,789,792,795,798,801,804,807,810,827,830,833,836,839,842,845,848,851],{"type":25,"children":730},[731],{"text":28,"type":29},{"type":31,"level":32,"children":733},[734],{"text":35,"type":29},{"type":25,"children":736},[737],{"text":39,"type":29},{"type":25,"children":739},[740],{"text":43,"type":29},{"type":31,"level":32,"children":742},[743],{"text":47,"type":29},{"type":25,"children":745},[746],{"text":51,"type":29},{"type":25,"children":748},[749],{"text":55,"type":29},{"type":25,"children":751},[752],{"text":59,"type":29},{"type":61,"format":62,"children":754},[755,758,761],{"type":65,"children":756},[757],{"text":68,"type":29},{"type":65,"children":759},[760],{"text":72,"type":29},{"type":65,"children":762},[763],{"text":76,"type":29},{"type":31,"level":32,"children":765},[766],{"text":80,"type":29},{"type":25,"children":768},[769,770,773,774,777],{"text":84,"type":29},{"text":86,"type":29,"marks":771},[772],{"href":89,"type":90},{"text":92,"type":29},{"text":94,"type":29,"marks":775},[776],{"href":97,"type":90},{"text":99,"type":29},{"type":61,"format":62,"children":779},[780,783,786],{"type":65,"children":781},[782],{"text":105,"type":29},{"type":65,"children":784},[785],{"text":109,"type":29},{"type":65,"children":787},[788],{"text":113,"type":29},{"type":25,"children":790},[791],{"text":117,"type":29},{"type":25,"children":793},[794],{"text":121,"type":29},{"type":31,"level":32,"children":796},[797],{"text":125,"type":29},{"type":25,"children":799},[800],{"text":129,"type":29},{"type":25,"children":802},[803],{"text":133,"type":29},{"type":31,"level":32,"children":805},[806],{"text":137,"type":29},{"type":25,"children":808},[809],{"text":141,"type":29},{"type":61,"format":62,"children":811},[812,815,818,821,824],{"type":65,"children":813},[814],{"text":147,"type":29},{"type":65,"children":816},[817],{"text":151,"type":29},{"type":65,"children":819},[820],{"text":155,"type":29},{"type":65,"children":822},[823],{"text":159,"type":29},{"type":65,"children":825},[826],{"text":163,"type":29},{"type":31,"level":32,"children":828},[829],{"text":167,"type":29},{"type":31,"level":21,"children":831},[832],{"text":171,"type":29},{"type":25,"children":834},[835],{"text":175,"type":29},{"type":31,"level":21,"children":837},[838],{"text":179,"type":29},{"type":25,"children":840},[841],{"text":183,"type":29},{"type":31,"level":21,"children":843},[844],{"text":187,"type":29},{"type":25,"children":846},[847],{"text":191,"type":29},{"type":25,"children":849},[850],{"text":195,"type":29},{"type":25,"children":852},[853,854,857],{"text":199,"type":29},{"text":201,"type":29,"marks":855},[856],{"href":204,"type":90},{"text":206,"type":29},null,[],[]]