[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"blog-strapi-article:zh:crawler-403-forbidden-proxy-header-login":3,"blog-article:\u002Fblog\u002Fcrawler-403-forbidden-proxy-header-login":884,"blog-related:\u002Fblog\u002Fcrawler-403-forbidden-proxy-header-login":885,"blog-translations:\u002Fblog\u002Fcrawler-403-forbidden-proxy-header-login":886},{"available":4,"article":5,"relatedArticles":154,"translations":787},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":153},"strapi",2443,"uioqeglpz95wflzh5o0bxbuh","zh","crawler-403-forbidden-proxy-header-login","\u002Fblog\u002Fcrawler-403-forbidden-proxy-header-login","爬虫 403 Forbidden 先查什么？代理、Header 和登录态","爬虫遇到 403 Forbidden 时，按代理认证、出口 IP、Header、Cookie、登录态和请求频率逐层排查，避免盲目换代理 IP。","https:\u002F\u002Fcms.quanqiudaili.com\u002Fuploads\u002Flarge_403_forbidden_header_cover_e9cd957eaf.webp","2026-09-02T02:55:55.361Z","2026-09-02T02:55:55.340Z","Global Proxies Team",[],false,0,3,"blocks",[24,30,36,40,45,49,53,57,61,65,69,73,77,81,85,89,93,97,101,105,109,113,117,121,125,129,133,137,141,145,149],{"type":25,"children":26},"paragraph",[27],{"text":28,"type":29},"爬虫遇到 403 Forbidden 时，不要先批量更换代理 IP。应先确认代理认证是否返回 407，再核对出口 IP、Header、Cookie、登录态和请求频率；只有这些条件稳定后，才判断是否与目标站规则有关。","text",{"type":31,"level":32,"children":33},"heading",2,[34],{"text":35,"type":29},"403、407、429 和超时的判断分支",{"type":25,"children":37},[38],{"text":39,"type":29},"如果需要进一步了解错误码组合排查，可参考 [爬虫 403、429 和验证码排障方法](https:\u002F\u002Fwww.quanqiudaili.com\u002Fblog\u002Fcrawler-403-429-captcha-proxy-ip-troubleshooting)。",{"type":41,"children":42},"code",[43],{"text":44,"type":29},"| 现象 | 第一检查项 | 下一步 |\n| --- | --- | --- |\n| 407 | 代理认证、端口、账号和密码 | 修正认证后重试；仅在动态住宅代理 IP 的 API 提取场景检查白名单 |\n| 403 | Header、Cookie、登录态和目标站规则 | 对比直连与代理请求 |\n| 429 | 请求频率、重试策略和轮换节奏 | 降低频率并设置退避 |\n| 超时 | 本地网络、端口、代理有效期和目标站响应 | 更换协议或测试其他出口 |\n| 浏览器成功、程序失败 | Header、TLS、JavaScript 和 Cookie | 对齐浏览器请求特征 |",{"type":31,"level":32,"children":46},[47],{"text":48,"type":29},"采集目标和访问方式：公开页、登录态、API、渲染页面",{"type":25,"children":50},[51],{"text":52,"type":29},"先确认目标站允许访问的范围。对获授权的公开页面，建立直连基线，再对比代理请求的状态码、响应头、Header、Cookie、登录状态、重定向链和请求频率；登录态、API 或渲染页面则还要遵守站点授权与接口规则。",{"type":31,"level":32,"children":54},[55],{"text":56,"type":29},"浏览器成功、程序失败怎么办",{"type":25,"children":58},[59],{"text":60,"type":29},"浏览器成功而程序失败时，先导出浏览器请求作为对照，逐项比较 User-Agent、Accept、Referer、Cookie、跳转处理和 TLS 特征。只有出口不同而其他请求条件一致时，才适合继续判断代理资源是否影响结果。",{"type":31,"level":32,"children":62},[63],{"text":64,"type":29},"程序请求被拒时，Header、Cookie 和 TLS 怎么逐项核对",{"type":25,"children":66},[67],{"text":68,"type":29},"请求实现、参数和频率也会影响结果；目标站要求登录或接口权限不足时，应先按站点规则处理。",{"type":31,"level":32,"children":70},[71],{"text":72,"type":29},"最小请求与日志示例（仅限获授权的公开页面）",{"type":25,"children":74},[75],{"text":76,"type":29},"下面示例用于核对请求是否确实经过代理、Header 和 Cookie 是否按预期发送，不用于绕过访问控制：",{"type":25,"children":78},[79],{"text":80,"type":29},"```bash\ncurl -i --max-time 15 \\\n-x http:\u002F\u002FUSER:PASSWORD@proxy.example:PORT \\\n-H 'User-Agent: Mozilla\u002F5.0 (compatible; AuthorizedMonitor\u002F1.0)' \\\n-H 'Accept: text\u002Fhtml' \\\n-H 'Cookie: session=REDACTED' \\\n'https:\u002F\u002Fpublic.example\u002Frobots.txt'\n```",{"type":25,"children":82},[83],{"text":84,"type":29},"伪代码核对顺序：`response = request(url, proxy, headers, cookies, timeout=15); log(status, headers[\"location\"], elapsed_ms, proxy_exit, retry_count)`。Cookie、账号标识和认证信息写入日志前必须脱敏。",{"type":25,"children":86},[87],{"text":88,"type":29},"脱敏日志样例：`2026-09-01T14:20:00+08:00 | GET \u002Frobots.txt | proxy_exit=US-NY | status=403 | location=- | elapsed_ms=842 | ua=AuthorizedMonitor\u002F1.0 | cookie=session=REDACTED | retry=0 | action=stop_and_review`",{"type":31,"level":32,"children":90},[91],{"text":92,"type":29},"403 排查时，动态、静态和机房代理怎么取舍",{"type":41,"children":94},[95],{"text":96,"type":29},"| 代理 IP 方案 | 适合任务 | 测试重点 | 不应直接假设 |\n| --- | --- | --- | --- |\n| 动态住宅代理 IP | 多地区、短会话、公开采样或大量出口 | 地区识别、轮换节奏、结果复现 | 轮换越快效果越好 |\n| 静态住宅代理 IP | 长期账号、连续会话、同地区复核 | 出口稳定、地区一致、验证码 | 仍可能触发验证 |\n| 机房代理 IP | 住宅属性非硬要求、重速度或成本的访问测试 | 目标站兼容、带宽、延迟 | 天然不适合所有任务 |",{"type":25,"children":98},[99],{"text":100,"type":29},"公开页面需要多出口轮换时可以评估动态住宅代理 IP；登录态采集或跨页流程更需要稳定会话。选型前先确认目标站是否允许访问、请求是否需要粘性会话，以及失败是否真的随出口变化。",{"type":31,"level":32,"children":102},[103],{"text":104,"type":29},"403、429 出现后，重试和日志应该怎么设",{"type":25,"children":106},[107],{"text":108,"type":29},"1. 用同一 URL 建立一次直连请求基线，保存状态码、响应头和响应片段。\n2. 加入代理后复用相同请求；出现 407 时先修正认证，不进入目标站排查。\n3. 认证成功但返回 403 时，对照浏览器检查 Header、Cookie、登录态和跳转链。\n4. 返回 429 时降低频率，并按 `Retry-After` 或指数退避设置重试间隔。\n5. 为每次失败记录出口、状态码、耗时、重试次数和最终结果，超过上限后停止自动重试。",{"type":31,"level":32,"children":110},[111],{"text":112,"type":29},"排查完成后如何选择代理方案",{"type":25,"children":114},[115],{"text":116,"type":29},"完成直连与代理对照后，再根据失败类型选择方案：需要稳定登录会话时可查看 [静态住宅代理 IP](https:\u002F\u002Fwww.quanqiudaili.com\u002Fproduct\u002Fstatic-residential-standard)；地区匹配要求更高时再了解 [原生静态住宅代理 IP](https:\u002F\u002Fwww.quanqiudaili.com\u002Fproduct\u002Fstatic-residential-native)。错误码较复杂时，参考 [403、429 和验证码排查指南](https:\u002F\u002Fwww.quanqiudaili.com\u002Fblog\u002Fcrawler-403-429-captcha-proxy-ip-troubleshooting)，并把状态码、请求头、Cookie、登录态和重试日志一起提交。",{"type":31,"level":32,"children":118},[119],{"text":120,"type":29},"FAQ",{"type":31,"level":21,"children":122},[123],{"text":124,"type":29},"爬虫代理 IP 适合哪些场景？",{"type":25,"children":126},[127],{"text":128,"type":29},"代理 IP 适合用于验证错误是否随出口变化，以及为允许访问的公开页面提供多出口请求。它不能替代登录权限、接口授权或目标站允许的访问方式。",{"type":31,"level":21,"children":130},[131],{"text":132,"type":29},"爬虫代理 IP 应该选动态还是静态？",{"type":25,"children":134},[135],{"text":136,"type":29},"无登录态的短请求可以先测试动态住宅代理 IP；需要 Cookie 连续、跨页操作或登录态的采集任务，应优先评估可保持会话的方案。",{"type":31,"level":21,"children":138},[139],{"text":140,"type":29},"配置后如何确认代理已经生效？",{"type":25,"children":142},[143],{"text":144,"type":29},"用实际请求确认代理生效：日志中应看到代理出口，并且目标站请求确实经过代理。仅查看浏览器 IP 检测页，不能证明程序请求也使用了相同配置。",{"type":31,"level":21,"children":146},[147],{"text":148,"type":29},"出现验证码\u002F打不开\u002F地区不准时先排查什么？",{"type":25,"children":150},[151],{"text":152,"type":29},"验证码增多时先降低频率并检查 Cookie 与会话；完全打不开时先查认证、协议和端口。地区不准时再核对出口定位，不要把三类问题同时归因于代理质量。","",[155,347,555],{"source":6,"id":156,"documentId":157,"translationKey":157,"locale":9,"slug":158,"path":159,"_path":159,"title":160,"description":161,"excerpt":161,"image":162,"publishedAt":163,"updatedAt":164,"author":17,"tags":165,"featured":19,"sortOrder":20,"readingTime":21,"bodyType":22,"bodyBlocks":166,"bodyHtml":153},2485,"cjleh5o0nn3oys8ogpozjjym","price-residential-proxy-settings","\u002Fblog\u002Fprice-residential-proxy-settings","价格监控用住宅代理 IP 怎么设置轮换和粘性？","按业务路径 P95 设粘性时长、限流先降频、切国家同步换 Cookie，避免把页面参数差异误判为出口问题","https:\u002F\u002Fcms.quanqiudaili.com\u002Fuploads\u002Flarge_E4_BB_B7_E6_A0_BC_E7_9_B_91_E6_8_E_A7_E7_94_A8_E4_BD_8_F_E5_AE_85_E4_BB_A3_E7_90_86_E6_80_8_E_E4_B9_88_E8_AE_BE_E7_BD_AE_E8_BD_AE_E6_8_D_A2_E5_92_8_C_E7_B2_98_E6_80_A7_cover_0b9068f488.webp","2026-09-03T03:46:36.846Z","2026-09-03T03:46:36.834Z",[],[167,171,175,179,183,187,191,195,199,203,207,211,215,219,223,227,231,235,239,243,247,276,280,284,288,292,296,300,304,307,311,315,319,323,327,331,335,339,343],{"type":25,"children":168},[169],{"text":170,"type":29},"价格监控不是“请求越多越该频繁换 IP”。公开列表页或彼此独立的商品详情可以按请求或任务批次轮换；从商品详情进入购物车、配送地区、税费和结算展示的连续流程，应在同一短会话内保持代理出口、Cookie、地区和货币一致。粘性时长以完成一条业务路径的实际 P95 耗时加安全余量为起点，不使用一个适用于所有网站的通用分钟数。",{"type":31,"level":32,"children":172},[173],{"text":174,"type":29},"按列表、详情和购物车阶段设置轮换策略",{"type":41,"children":176},[177],{"text":178,"type":29},"| 页面阶段 | 推荐策略 | 原因 |\n| --- | --- | --- |\n| 搜索或分类列表 | 按请求或小批次轮换 | 页面之间通常相互独立 |\n| 单个商品详情 | 同一次解析保持同一出口 | 避免主页面与异步价格接口地区不一致 |\n| 购物车 | 保持短期粘性 | Cookie、库存和地区开始形成连续状态 |\n| 配送、税费、结算展示 | 整条路径保持同一会话 | 中途切换可能重置地区、货币或购物车 |\n| 下一商品或下一地区任务 | 结束旧会话后再轮换 | 防止状态串到新样本 |",{"type":25,"children":180},[181],{"text":182,"type":29},"一个任务如果只采公开列表价，可以把“商品×地区×时间窗口”作为独立采样单位；如果需要最终展示价，则应把“详情→购物车→配送地区→结算展示”视为一个不可拆的会话。",{"type":25,"children":184},[185],{"text":186,"type":29},"代理协议、主机、端口和认证字段可先按[代理设置帮助](https:\u002F\u002Fhelp.quanqiudaili.com\u002Fdocs\u002Fproxy-Settings)完成基础核验。",{"type":31,"level":32,"children":188},[189],{"text":190,"type":29},"根据完整业务路径耗时设置粘性时长",{"type":25,"children":192},[193],{"text":194,"type":29},"先在不并发的条件下运行一批完整流程，记录每条路径从打开详情页到拿到最终价格的耗时。计算 P95 后再加安全余量：",{"type":25,"children":196},[197],{"text":198,"type":29},"```text\n建议粘性时长 = 完整业务路径 P95 耗时 × 安全系数\n```",{"type":25,"children":200},[201],{"text":202,"type":29},"例如路径 P95 为 80 秒，安全系数取 1.5，则初始值可设为 120 秒，再通过失败率和会话中断率调整。这个数字只是该站点、该流程和当前网络条件下的起始值，不是产品承诺。动态住宅代理 IP 的实际在线时长可能与设置值存在差异；粘性要求越长，可用代理 IP 数量通常越少。",{"type":31,"level":32,"children":204},[205],{"text":206,"type":29},"用站点、地区和会话生成轮换键",{"type":25,"children":208},[209],{"text":210,"type":29},"建议使用下列绑定关系：",{"type":25,"children":212},[213],{"text":214,"type":29},"```text\nsession_key = 站点 + 国家\u002F城市 + 货币 + 账号\u002F匿名会话 + 任务批次\n```",{"type":25,"children":216},[217],{"text":218,"type":29},"同一个 `session_key` 在完成业务路径前复用同一代理出口和 Cookie。更换国家、城市、货币或任务批次时创建新会话，不要只换代理 IP 却继续复用旧 Cookie。",{"type":25,"children":220},[221],{"text":222,"type":29},"如果目标站允许直接通过 URL、请求参数或官方接口指定地区，应把这些参数与代理 IP 地区保持一致，避免把页面参数造成的差异误判为出口问题。",{"type":31,"level":32,"children":224},[225],{"text":226,"type":29},"失败重试先保持会话，再判断是否换出口",{"type":25,"children":228},[229],{"text":230,"type":29},"第一次网络超时或临时 5xx 可以在同一会话内有限重试，避免购物车和地区状态丢失。出现下列情况时再分支：",{"type":41,"children":232},[233],{"text":234,"type":29},"| 现象 | 是否立即换代理 IP | 处理动作 |\n| --- | --- | --- |\n| 单次超时或临时 5xx | 否 | 同会话退避重试，设置次数上限 |\n| 429 | 否 | 读取限流信息，先降频和延长间隔 |\n| 购物车被清空、地区跳回 | 否 | 查 Cookie、地区参数和会话是否中断 |\n| 只有少数出口重复失败 | 可以评估 | 保持其他变量不变，更换同地区出口复测 |\n| 账号或权限提示 | 否 | 处理账号或平台提示 |\n| 粘性会话提前中断 | 先记录 | 缩短单条路径、减少并发或重新评估方案 |",{"type":25,"children":236},[237],{"text":238,"type":29},"不要在一次失败后同时更换代理 IP、Cookie、UA、地区和货币，否则无法知道是哪项导致结果变化。",{"type":31,"level":32,"children":240},[241],{"text":242,"type":29},"价格结果必须记录哪些字段",{"type":25,"children":244},[245],{"text":246,"type":29},"至少记录：",{"type":248,"format":249,"children":250},"list","unordered",[251,256,260,264,268,272],{"type":252,"children":253},"list-item",[254],{"text":255,"type":29},"站点、商品 ID、页面 URL 和页面阶段；",{"type":252,"children":257},[258],{"text":259,"type":29},"国家、城市、邮编或配送地区；",{"type":252,"children":261},[262],{"text":263,"type":29},"页面语言、货币和登录状态；",{"type":252,"children":265},[266],{"text":267,"type":29},"会话 ID、脱敏后的代理标识与出口 IP；",{"type":252,"children":269},[270],{"text":271,"type":29},"抓取时间、原价、促销价、会员价、税费、运费和最终展示价；",{"type":252,"children":273},[274],{"text":275,"type":29},"状态码、验证码、空结果或重定向情况。",{"type":25,"children":277},[278],{"text":279,"type":29},"如果只保存一个“价格”字段，后续无法判断差异来自促销、税费、配送、会员状态还是地区定价。",{"type":25,"children":281},[282],{"text":283,"type":29},"例如一次采样可以记录“列表页发现商品→详情页读取原价→购物车确认配送费”三个时间点，并为同一条路径使用同一个会话 ID。后续价格变化时，才能判断是页面阶段不同，还是地区和会话真的发生了变化。",{"type":31,"level":32,"children":285},[286],{"text":287,"type":29},"通过重复采样验证价格差异是否真实",{"type":25,"children":289},[290],{"text":291,"type":29},"同一商品、同一地区至少在多个时间窗口重复采样，并设置一个基准市场。比较时保持页面阶段、登录状态、货币和 Cookie 策略一致。",{"type":41,"children":293},[294],{"text":295,"type":29},"| 结果 | 更可能的原因 | 下一步 |\n| --- | --- | --- |\n| 同出口、同会话连续跳价 | 促销、缓存、A\u002FB 测试或库存变化 | 增加重复样本，保存页面阶段和时间 |\n| 换国家后语言和货币变化 | 地区策略可能生效 | 再核对配送与税费字段 |\n| 出口地区正确但价格不变 | 目标站可能不按该地区定价 | 检查 Cookie、账号和页面参数 |\n| 只有购物车阶段失败 | 会话连续性或地区参数问题 | 保持短期粘性并逐步复测 |",{"type":31,"level":32,"children":297},[298],{"text":299,"type":29},"如何选择代理 IP 类型",{"type":25,"children":301},[302],{"text":303,"type":29},"多地区公开价格采样通常先评估动态住宅代理 IP；需要较长的登录态、购物车或后台操作时，应先看任务是否更适合静态住宅代理 IP。选择前重点确认地区覆盖、目标站识别、粘性能力、并发、计费方式和失败重试成本，不把代理 IP 当成价格一致性的保证。",{"type":31,"level":32,"children":305},[306],{"text":120,"type":29},{"type":31,"level":21,"children":308},[309],{"text":310,"type":29},"粘性时间越长越好吗？",{"type":25,"children":312},[313],{"text":314,"type":29},"不是。时长只需覆盖完整业务路径并留出合理余量；过长会减少可用代理 IP 数量，也会降低任务调度灵活性。",{"type":31,"level":21,"children":316},[317],{"text":318,"type":29},"价格不同一定是代理 IP 导致的吗？",{"type":25,"children":320},[321],{"text":322,"type":29},"不一定。Cookie、账号、货币、配送地址、促销、库存和 A\u002FB 测试都可能改变价格。",{"type":31,"level":21,"children":324},[325],{"text":326,"type":29},"同一商品要重复采样吗？",{"type":25,"children":328},[329],{"text":330,"type":29},"建议重复采样，并记录相同字段。单次结果不足以区分短时促销、缓存和稳定地区差异。",{"type":31,"level":21,"children":332},[333],{"text":334,"type":29},"动态住宅代理 IP 能保证保持到设定时间吗？",{"type":25,"children":336},[337],{"text":338,"type":29},"不能保证完全一致。应按实测会话中断率设置任务长度和重试方案，必要时缩短单条路径。",{"type":25,"children":340},[341],{"text":342,"type":29},"---",{"type":25,"children":344},[345],{"text":346,"type":29},"完成小样本验证后，可查看[动态住宅代理 IP](https:\u002F\u002Fwww.quanqiudaili.com\u002Fproduct\u002Fdynamic-residential)并按当前业务路径评估轮换方案。",{"source":6,"id":348,"documentId":349,"translationKey":349,"locale":9,"slug":350,"path":351,"_path":351,"title":352,"description":353,"excerpt":353,"image":354,"publishedAt":355,"updatedAt":356,"author":17,"tags":357,"featured":19,"sortOrder":20,"readingTime":21,"bodyType":22,"bodyBlocks":358,"bodyHtml":153},2484,"e8solm25h3e0r6jbsq1peuf9","crawler-429-too-many-requests-proxy","\u002Fblog\u002Fcrawler-429-too-many-requests-proxy","爬虫 429 Too Many Requests：判断频率还是代理 IP 池","爬虫遇到 429 Too Many Requests 时，先固定账号和代理出口做降频实验，再以低频基线更换代理 IP，区分频率、配额和代理 IP 池问题。","https:\u002F\u002Fcms.quanqiudaili.com\u002Fuploads\u002Flarge_429_too_many_requests_cover_978d5b6a35.webp","2026-09-03T03:46:20.903Z","2026-09-03T03:46:20.890Z",[],[359,363,367,371,401,405,409,413,417,421,425,429,433,437,441,445,449,453,457,461,465,469,473,477,481,485,489,493,497,501,505,509,513,516,520,524,528,532,536,540,544,548,551],{"type":25,"children":360},[361],{"text":362,"type":29},"遇到 429，先固定账号、URL、Header、Cookie 和代理 IP，只降低并发、延长请求间隔并遵守 `Retry-After`；如果 429 稳定下降，优先处理频率或账号\u002FAPI 配额。只有在低频基线下，错误仍稳定集中于少数出口，才继续检查代理 IP 池质量、地区识别或出口复用。",{"type":31,"level":32,"children":364},[365],{"text":366,"type":29},"先保存 429 的完整证据",{"type":25,"children":368},[369],{"text":370,"type":29},"429 表示请求过多，但限制维度可能是账号、API 密钥、接口、IP、设备或多项组合。保存：",{"type":248,"format":249,"children":372},[373,377,381,385,389,393,397],{"type":252,"children":374},[375],{"text":376,"type":29},"状态码与响应体摘要；",{"type":252,"children":378},[379],{"text":380,"type":29},"`Retry-After` 和其他限流响应头；",{"type":252,"children":382},[383],{"text":384,"type":29},"账号\u002FAPI 密钥的脱敏标识；",{"type":252,"children":386},[387],{"text":388,"type":29},"URL、接口、请求方法和参数摘要；",{"type":252,"children":390},[391],{"text":392,"type":29},"并发数、请求间隔、时间窗口；",{"type":252,"children":394},[395],{"text":396,"type":29},"代理出口 IP、国家\u002F城市；",{"type":252,"children":398},[399],{"text":400,"type":29},"请求总数、成功数、429 数量和耗时。",{"type":25,"children":402},[403],{"text":404,"type":29},"如果响应中包含 `X-RateLimit-Limit`、`X-RateLimit-Remaining` 或 `X-RateLimit-Reset` 等字段，也应一并保存；这些字段有助于判断限制是按账号、接口还是时间窗口计算。",{"type":25,"children":406},[407],{"text":408,"type":29},"HTTP 429 与 `Retry-After` 的语义可参考 [RFC 6585 第 4 节](https:\u002F\u002Fwww.rfc-editor.org\u002Frfc\u002Frfc6585#section-4)。服务端没有义务说明采用哪一种计数维度，因此需要对照实验。",{"type":25,"children":410},[411],{"text":412,"type":29},"若同时出现 403 或验证码，可结合[爬虫 403\u002F429\u002F验证码排查](https:\u002F\u002Fwww.quanqiudaili.com\u002Fblog\u002Fcrawler-403-429-captcha-proxy-ip-troubleshooting)分开记录，避免把不同错误合并成 429。",{"type":31,"level":32,"children":414},[415],{"text":416,"type":29},"阶段 A：固定代理 IP，逐步降低请求频率",{"type":25,"children":418},[419],{"text":420,"type":29},"保持账号、代理 IP、URL、Header、Cookie 和请求参数不变，按阶梯降低并发和请求频率。每一档使用相同样本量与观察时间，不要在中途换出口。",{"type":25,"children":422},[423],{"text":424,"type":29},"示例记录表：",{"type":41,"children":426},[427],{"text":428,"type":29},"| 组别 | 并发 | 请求间隔 | 请求数 | 429 数 | 成功数 | 结论 |\n| --- | ---: | ---: | ---: | ---: | ---: | --- |\n| A1 | 当前值 | 当前值 | 同样本量 | 记录 | 记录 | 基线 |\n| A2 | 降低一档 | 延长一档 | 同样本量 | 记录 | 记录 | 与 A1 比较 |\n| A3 | 再降低一档 | 再延长一档 | 同样本量 | 记录 | 记录 | 验证趋势是否重复 |",{"type":25,"children":430},[431],{"text":432,"type":29},"若降频后 429 在连续时间窗口内都明显下降，而出口未变，主要矛盾更可能是频率、账号或接口配额。若响应带 `Retry-After`，等待时间应优先遵守服务端提示。",{"type":31,"level":32,"children":434},[435],{"text":436,"type":29},"阶段 B：保持低频基线，仅更换代理 IP",{"type":25,"children":438},[439],{"text":440,"type":29},"保持阶段 A 得到的低频基线，固定账号、URL、Header、Cookie、参数、并发和间隔，只替换同地区代理 IP。",{"type":41,"children":442},[443],{"text":444,"type":29},"| 结果模式 | 更可能的原因 | 下一步 |\n| --- | --- | --- |\n| 所有出口在同一时段都 429 | 账号、接口或全站限流 | 暂停并检查配额\u002F服务端提示 |\n| 同一账号跨多个出口都 429 | 账号或 API 密钥配额 | 换 IP 通常无效 |\n| 少数出口跨账号重复 429 | 出口复用或质量问题 | 隔离出口并扩大同地区样本 |\n| 更换出口短暂恢复，随后再次 429 | 频率策略仍不合适 | 回到阶段 A，降低总体负载 |\n| 429 响应实际是验证页面 | 状态分类错误 | 单独记录验证码\u002F挑战，不当作普通 429 |",{"type":31,"level":32,"children":446},[447],{"text":448,"type":29},"429 的重试与停止规则",{"type":31,"level":21,"children":450},[451],{"text":452,"type":29},"429 响应的基础退避代码",{"type":25,"children":454},[455],{"text":456,"type":29},"```python\nimport random\nimport time",{"type":25,"children":458},[459],{"text":460,"type":29},"def wait_before_retry(response, attempt):\nretry_after = response.headers.get(\"Retry-After\")\nif retry_after and retry_after.isdigit():\nwait = float(retry_after)\nelse:\nwait = min(2 ** attempt, 60)\ntime.sleep(wait + random.uniform(0, 0.5))\n```",{"type":31,"level":21,"children":462},[463],{"text":464,"type":29},"生产环境如何设置重试上限与共享限流",{"type":25,"children":466},[467],{"text":468,"type":29},"生产实现还应支持 HTTP 日期格式的 `Retry-After`，并设置：",{"type":25,"children":470},[471],{"text":472,"type":29},"1. 最大重试次数；\n2. 单任务总时限；\n3. 同账号或接口的共享限流器；\n4. 连续 429 后的暂停和报警；\n5. 重试队列的容量上限。",{"type":25,"children":474},[475],{"text":476,"type":29},"不要让多个工作进程各自判断后立即重试，否则总体流量仍可能超过限制。账号或 API 密钥维度的限流，应在所有工作进程之间共享状态。",{"type":31,"level":32,"children":478},[479],{"text":480,"type":29},"把 403、429、验证码和空结果分开",{"type":41,"children":482},[483],{"text":484,"type":29},"| 结果 | 主要含义 | 是否按 429 重试 |\n| --- | --- | --- |\n| 429 | 请求过多或配额限制 | 按服务端提示有限重试 |\n| 403 | 权限、账号、地区或访问策略 | 不直接套用 429 退避 |\n| 验证码\u002F挑战页 | 环境、会话或请求行为变化 | 暂停并单独诊断 |\n| HTTP 200 但空结果 | 业务无数据或解析问题 | 不应计为请求成功或 429 |\n| 5xx | 临时服务端错误 | 可有限重试，单独统计 |",{"type":25,"children":486},[487],{"text":488,"type":29},"把所有失败合并为一个错误率，会导致错误扩容代理 IP 池或错误提高重试次数。",{"type":25,"children":490},[491],{"text":492,"type":29},"建议把每个实验窗口固定为相同请求数，例如每组 100 次，并记录窗口开始和结束时间。只有在多个窗口都出现同样趋势时，才把“降频有效”或“少数出口异常”当作可复用结论。",{"type":31,"level":32,"children":494},[495],{"text":496,"type":29},"评估代理 IP 池容量的方法",{"type":25,"children":498},[499],{"text":500,"type":29},"评估代理 IP 池容量时，应使用单个代理 IP 的安全承载量、基线成功率、冷却占比和重试放大，而不是用“总请求量÷IP 数”直接估算。",{"type":31,"level":32,"children":502},[503],{"text":504,"type":29},"用脱敏复测数据区分频率限制与代理池因素",{"type":25,"children":506},[507],{"text":508,"type":29},"以下为脱敏记录示例（接口、账号和 IP 已隐去，数字为示例、非 SLA）：同一账号与单一出口、并发 5 时 50 次请求出现 429 12 次；降至并发 1、间隔 8 秒后为 2 次；保持低频改用 5 个出口仍为 2 次。该对照更支持频率因素，不能据此断言代理池无问题。",{"type":25,"children":510},[511],{"text":512,"type":29},"429 状态码与 `Retry-After` 可核对 [RFC 6585](https:\u002F\u002Fwww.rfc-editor.org\u002Frfc\u002Frfc6585.html)，HTTP 语义可核对 [RFC 9110](https:\u002F\u002Fwww.rfc-editor.org\u002Frfc\u002Frfc9110.html)；并发、间隔和样本量属于经验设计。",{"type":31,"level":32,"children":514},[515],{"text":120,"type":29},{"type":31,"level":21,"children":517},[518],{"text":519,"type":29},"429 和 403 是同一种问题吗？",{"type":25,"children":521},[522],{"text":523,"type":29},"不是。429 主要表示请求过多或配额限制；403 还可能涉及权限、账号、地区或访问策略。",{"type":31,"level":21,"children":525},[526],{"text":527,"type":29},"没有 Retry-After 怎么办？",{"type":25,"children":529},[530],{"text":531,"type":29},"采用指数退避和随机抖动，并设置最大次数与任务总时限。仍持续 429 时暂停任务。",{"type":31,"level":21,"children":533},[534],{"text":535,"type":29},"账号配额会导致 429 吗？",{"type":25,"children":537},[538],{"text":539,"type":29},"会。若限制绑定账号或 API 密钥，更换代理 IP 通常不会解除配额限制。",{"type":31,"level":21,"children":541},[542],{"text":543,"type":29},"单次 429 能判断代理 IP 池质量吗？",{"type":25,"children":545},[546],{"text":547,"type":29},"不能。至少要完成降频实验，并在低频基线下比较多个同地区出口和时间窗口。",{"type":25,"children":549},[550],{"text":342,"type":29},{"type":25,"children":552},[553],{"text":554,"type":29},"确认问题确实集中在出口后，再按任务规模评估[动态住宅代理 IP](https:\u002F\u002Fwww.quanqiudaili.com\u002Fproduct\u002Fdynamic-residential)。",{"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":153},2483,"p940djzj71fql92iswg7jyyg","market-intelligence-country-city-proxy-ip","\u002Fblog\u002Fmarket-intelligence-country-city-proxy-ip","市场情报采集如何选择国家和城市代理 IP？","市场情报采集应先按研究问题筛选国家，再确认目标站是否存在城市差异，并用分层抽样、地区交叉验证和失败分类校验代理 IP 结果。","https:\u002F\u002Fcms.quanqiudaili.com\u002Fuploads\u002Flarge_cover_c830f3842a.webp","2026-09-03T03:45:59.382Z","2026-09-03T03:45:59.369Z",[],[567,571,575,579,583,587,591,595,599,603,607,611,615,619,623,627,631,635,639,643,647,651,655,673,677,681,685,689,693,697,701,705,709,713,717,721,725,729,733,737,741,745,748,752,756,760,764,768,772,776,780,783],{"type":25,"children":568},[569],{"text":570,"type":29},"市场情报采集应先按业务问题选择国家，再决定是否下钻到城市。只有目标站会因城市改变价格、库存、配送、广告、搜索结果或门店信息时，城市层级代理 IP 才有额外价值；如果研究问题只比较国家语言、货币或国家层级页面，先做国家层级采样更节省资源，也更容易解释结果。",{"type":31,"level":32,"children":572},[573],{"text":574,"type":29},"先明确要用采集验证哪些市场差异",{"type":25,"children":576},[577],{"text":578,"type":29},"不要先看代理 IP 库存再决定市场。先明确要验证什么：",{"type":41,"children":580},[581],{"text":582,"type":29},"| 研究问题 | 国家层级是否通常足够 | 何时需要城市层级 |\n| --- | --- | --- |\n| 页面语言、货币、国家站重定向 | 通常足够 | 同一国家不同城市返回不同站点时 |\n| 商品价格和促销 | 先做国家层级 | 城市、邮编或门店影响价格时 |\n| 库存、配送时效、门店信息 | 往往不够 | 结果依赖城市、邮编或配送地址时 |\n| 本地搜索与广告展示 | 先做国家基线 | 目标结果明显按城市变化时 |\n| 竞争品牌覆盖 | 取决于研究范围 | 需要比较核心城市与非核心城市时 |",{"type":25,"children":584},[585],{"text":586,"type":29},"每个问题都要定义“页面出现什么差异才算有效”，例如货币变化、配送日期变化、门店列表变化或搜索结果排序变化。",{"type":31,"level":32,"children":588},[589],{"text":590,"type":29},"按市场价值和研究目标筛选国家",{"type":31,"level":21,"children":592},[593],{"text":594,"type":29},"先按业务价值划分市场层级",{"type":25,"children":596},[597],{"text":598,"type":29},"可以把候选国家分为四类：",{"type":25,"children":600},[601],{"text":602,"type":29},"1. **核心市场**：当前收入、客户或业务量较高，需要持续监测。\n2. **增长市场**：业务增长较快或准备投入，需要增加采样。\n3. **竞争对照市场**：用于比较竞争强度、价格或产品策略。\n4. **基准市场**：作为长期不变的参照组，帮助识别全站变化。",{"type":31,"level":21,"children":604},[605],{"text":606,"type":29},"用评分模型确定采样优先级",{"type":25,"children":608},[609],{"text":610,"type":29},"如果国家很多，可用评分表排序。下面权重仅为方法示例，企业应按自己的研究目标调整：",{"type":41,"children":612},[613],{"text":614,"type":29},"| 维度 | 示例权重 | 评分依据 |\n| --- | ---: | --- |\n| 当前业务价值 | 35% | 收入、客户量或业务优先级 |\n| 增长潜力 | 25% | 增长趋势与市场计划 |\n| 竞争变化速度 | 20% | 价格、广告和新品变化频率 |\n| 语言\u002F货币\u002F规则差异 | 10% | 是否能代表独立市场环境 |\n| 数据可获得性 | 10% | 页面稳定性、公开程度和合规条件 |",{"type":25,"children":616},[617],{"text":618,"type":29},"先统一各维度的评分范围，例如都使用 1–5 分，再按下式计算：",{"type":25,"children":620},[621],{"text":622,"type":29},"```text\n国家优先级得分 = 各维度评分 × 对应权重后求和\n```",{"type":25,"children":624},[625],{"text":626,"type":29},"缺失数据应标记为待核实，不要直接按 0 分处理，以免把“没有数据”误判为“没有价值”。",{"type":25,"children":628},[629],{"text":630,"type":29},"总分只用于排优先级，不代表代理 IP 质量，也不代表该市场一定值得长期投入。最终还要确认目标国家资源、地区识别和目标站可访问性。",{"type":31,"level":32,"children":632},[633],{"text":634,"type":29},"用小样本判断是否需要细分到城市",{"type":31,"level":21,"children":636},[637],{"text":638,"type":29},"设计国家与城市层级的对照样本",{"type":25,"children":640},[641],{"text":642,"type":29},"该对照的目的不是证明某个城市一定更优，而是判断城市差异在控制其他变量后是否仍然存在。",{"type":25,"children":644},[645],{"text":646,"type":29},"每个入选国家先选择一个核心城市和一个非核心城市，在同一时间窗口测试同一页面。保持浏览器版本、语言、Cookie、账号状态和请求参数一致，只改变城市出口。",{"type":31,"level":21,"children":648},[649],{"text":650,"type":29},"确定扩大城市采样的判定阈值",{"type":25,"children":652},[653],{"text":654,"type":29},"若连续多个样本在以下字段出现稳定差异，才扩大城市覆盖：",{"type":248,"format":249,"children":656},[657,661,665,669],{"type":252,"children":658},[659],{"text":660,"type":29},"商品价格、库存或配送时效；",{"type":252,"children":662},[663],{"text":664,"type":29},"搜索结果、广告或本地商家列表；",{"type":252,"children":666},[667],{"text":668,"type":29},"页面语言、货币、重定向或地区提示；",{"type":252,"children":670},[671],{"text":672,"type":29},"门店、服务范围或本地内容模块。",{"type":25,"children":674},[675],{"text":676,"type":29},"判定时至少覆盖多个时间窗口，并重复比较两个以上核心城市与非核心城市样本。差异要在同一业务字段上重复出现；单次价格跳变、缓存变化、验证码或网络失败不能作为城市差异证据。",{"type":25,"children":678},[679],{"text":680,"type":29},"达到上述条件后再扩大城市数量；如果城市之间没有稳定差异，继续国家层级采样即可。城市粒度不是越细越好，过度下钻会增加样本量、成本和归因难度。",{"type":31,"level":32,"children":682},[683],{"text":684,"type":29},"按国家与城市层级设计分层样本",{"type":25,"children":686},[687],{"text":688,"type":29},"一个可执行的起始结构是：",{"type":41,"children":690},[691],{"text":692,"type":29},"| 层级 | 样本角色 | 用途 |\n| --- | --- | --- |\n| 每国核心城市 | 业务与人口集中地区 | 观察主流市场表现 |\n| 每国非核心城市 | 与核心城市形成对照 | 判断差异是否只集中在大城市 |\n| 基准国家\u002F城市 | 长期保持不变 | 识别站点全局变化 |\n| 重复时间窗口 | 工作日\u002F周末或业务高低峰 | 排除短时促销与缓存影响 |",{"type":25,"children":694},[695],{"text":696,"type":29},"先用小样本确认差异，再增加城市和时间窗口。不要一开始同时扩大国家、城市、页面和账号数量，否则数据异常时很难定位原因。",{"type":31,"level":32,"children":698},[699],{"text":700,"type":29},"用页面信号交叉验证代理 IP 的国家和城市",{"type":25,"children":702},[703],{"text":704,"type":29},"先用 IP 检测页记录出口国家和城市，再与目标站页面语言、货币、重定向、配送地区或本地内容交叉验证。不同地理位置数据库可能给同一 IP 标注不同城市，因此“检测库城市不一致”不能直接判为代理 IP 错误。",{"type":25,"children":706},[707],{"text":708,"type":29},"建议记录三个层次：",{"type":25,"children":710},[711],{"text":712,"type":29},"1. 代理 IP 检测库显示的国家和城市；\n2. 目标站实际识别或展示的地区；\n3. 业务结果是否符合研究预期。",{"type":25,"children":714},[715],{"text":716,"type":29},"如果国家一致、城市有差异，应再换一个同城市代理 IP 复测；如果目标站明确跳转到错误国家，再联系服务方核实或更换代理 IP。",{"type":31,"level":32,"children":718},[719],{"text":720,"type":29},"按失败类型分别统计并处理",{"type":41,"children":722},[723],{"text":724,"type":29},"| 结果 | 是否计入业务差异 | 处理方式 |\n| --- | --- | --- |\n| 页面正常但无相关内容 | 可以单独统计 | 标记“无结果”，不等同网络失败 |\n| 验证码或挑战页 | 不计为有效业务结果 | 记录频率、Cookie、账号和出口 |\n| 连接超时或 5xx | 不计为有效业务结果 | 有限重试，保留耗时和错误 |\n| 账号\u002F权限提示 | 不计为地区差异 | 先处理账号或平台规则 |\n| 地区与购买国家不符 | 暂不计入 | 同地区换代理 IP 复测并记录 |",{"type":25,"children":726},[727],{"text":728,"type":29},"公开数据采样也应遵守目标站条款、访问频率、robots 或官方接口要求。代理 IP 只改变网络出口，不能改变访问权限。",{"type":25,"children":730},[731],{"text":732,"type":29},"在样本表中建议同时保留“购买国家、检测库国家、目标站展示国家、城市字段和业务结果”五列。地区不一致时先定位是哪一列发生偏差，再决定重测代理 IP 还是修正研究条件。",{"type":25,"children":734},[735],{"text":736,"type":29},"涉及采集责任边界时，可进一步查看[相关采集合规说明](https:\u002F\u002Fwww.quanqiudaili.com\u002Fblog\u002Fproxy-ip-risk-boundary-account-ads-kyc-scraping)。",{"type":31,"level":32,"children":738},[739],{"text":740,"type":29},"国家或城市代理 IP 资源不足时的替代方案",{"type":25,"children":742},[743],{"text":744,"type":29},"城市层级资源不足时，先退回国家层级采样或减少城市数量，不要用不确定的城市标注填补样本。国家、城市覆盖与识别结果应以当前后台、购买页或客户经理确认为准，不能把历史库存当成长期承诺。",{"type":31,"level":32,"children":746},[747],{"text":120,"type":29},{"type":31,"level":21,"children":749},[750],{"text":751,"type":29},"城市层级定位一定比国家层级更好吗？",{"type":25,"children":753},[754],{"text":755,"type":29},"不一定。只有研究对象在城市层面产生可验证差异时，城市级采样才值得增加成本。",{"type":31,"level":21,"children":757},[758],{"text":759,"type":29},"城市显示不一致就是代理 IP 错了吗？",{"type":25,"children":761},[762],{"text":763,"type":29},"不一定。先比较多个检测库和目标站实际展示，再使用同城市其他代理 IP 复测。",{"type":31,"level":21,"children":765},[766],{"text":767,"type":29},"样本应优先覆盖哪些市场？",{"type":25,"children":769},[770],{"text":771,"type":29},"先覆盖核心市场和增长市场，再加入竞争对照市场与一个长期基准市场。",{"type":31,"level":21,"children":773},[774],{"text":775,"type":29},"无结果可以直接记为采集失败吗？",{"type":25,"children":777},[778],{"text":779,"type":29},"不建议。无内容、验证码、网络失败和账号限制应分开统计，否则成功率和市场结论都会失真。",{"type":25,"children":781},[782],{"text":342,"type":29},{"type":25,"children":784},[785],{"text":786,"type":29},"完成小样本地区验证后，可按当前国家和城市需求评估[动态住宅代理 IP](https:\u002F\u002Fwww.quanqiudaili.com\u002Fproduct\u002Fdynamic-residential)。",[788],{"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":789,"featured":19,"sortOrder":20,"readingTime":21,"bodyType":22,"bodyBlocks":790,"bodyHtml":153},[],[791,794,797,800,803,806,809,812,815,818,821,824,827,830,833,836,839,842,845,848,851,854,857,860,863,866,869,872,875,878,881],{"type":25,"children":792},[793],{"text":28,"type":29},{"type":31,"level":32,"children":795},[796],{"text":35,"type":29},{"type":25,"children":798},[799],{"text":39,"type":29},{"type":41,"children":801},[802],{"text":44,"type":29},{"type":31,"level":32,"children":804},[805],{"text":48,"type":29},{"type":25,"children":807},[808],{"text":52,"type":29},{"type":31,"level":32,"children":810},[811],{"text":56,"type":29},{"type":25,"children":813},[814],{"text":60,"type":29},{"type":31,"level":32,"children":816},[817],{"text":64,"type":29},{"type":25,"children":819},[820],{"text":68,"type":29},{"type":31,"level":32,"children":822},[823],{"text":72,"type":29},{"type":25,"children":825},[826],{"text":76,"type":29},{"type":25,"children":828},[829],{"text":80,"type":29},{"type":25,"children":831},[832],{"text":84,"type":29},{"type":25,"children":834},[835],{"text":88,"type":29},{"type":31,"level":32,"children":837},[838],{"text":92,"type":29},{"type":41,"children":840},[841],{"text":96,"type":29},{"type":25,"children":843},[844],{"text":100,"type":29},{"type":31,"level":32,"children":846},[847],{"text":104,"type":29},{"type":25,"children":849},[850],{"text":108,"type":29},{"type":31,"level":32,"children":852},[853],{"text":112,"type":29},{"type":25,"children":855},[856],{"text":116,"type":29},{"type":31,"level":32,"children":858},[859],{"text":120,"type":29},{"type":31,"level":21,"children":861},[862],{"text":124,"type":29},{"type":25,"children":864},[865],{"text":128,"type":29},{"type":31,"level":21,"children":867},[868],{"text":132,"type":29},{"type":25,"children":870},[871],{"text":136,"type":29},{"type":31,"level":21,"children":873},[874],{"text":140,"type":29},{"type":25,"children":876},[877],{"text":144,"type":29},{"type":31,"level":21,"children":879},[880],{"text":148,"type":29},{"type":25,"children":882},[883],{"text":152,"type":29},null,[],[]]