OpenUptime

Down in one region only

A customer in another country says your site is down, and it loads fine for you. Both can be true. This guide narrows the cause down in four steps, then sets up a check that sees what that customer sees.

Common causes

CauseWhat it looks like
A DNS change that has not reached every resolverSome users get the old address and time out or see the old site. See DNS changes that have not arrived yet.
A CDN or load balancer edge in troubleSame name, same answer, but the edge near the user fails: 5xx, resets, or slow first bytes.
A route problem between a network and your hostConnections from one provider or country hang at connect; everyone else is fine.
A firewall, WAF or geo ruleOne region gets 403, a challenge page or a reset, often after a rule change.
IPv6 broken while IPv4 worksUsers on networks that prefer IPv6 fail; the AAAA record points at something that does not answer.

1. See where the request stops

Ask the affected user, or a machine in that region, to run this and send you the line it prints:

curl -s -o /dev/null https://www.example.com/ \
  -w "dns %{time_namelookup}s  connect %{time_connect}s  tls %{time_appconnect}s  first byte %{time_starttransfer}s  total %{time_total}s  status %{http_code}  ip %{remote_ip}\n"
  • A large dns time or a different ip from yours points at DNS.
  • connect that never finishes points at routing or a firewall that drops packets.
  • A tls time with an error points at the certificate the edge in that region serves.
  • A fast connect and a slow or failing first byte points at the edge or the origin behind it.

2. Compare DNS answers

Different public resolvers can hold different answers while a change is spreading. If they disagree, the problem is DNS, not your server:

dig +short www.example.com @1.1.1.1
dig +short www.example.com @8.8.8.8
dig +short www.example.com @9.9.9.9

The Compare resolvers view of the DNS lookup tool asks Cloudflare, Google, AliDNS and DNSPod at once from a browser and says whether they agree.

3. Test one address directly

Send the request to the address the user got, without changing any DNS. If that address fails and yours works, the edge or server behind it is the problem; tell your CDN or host which address and region:

# ask a specific edge or origin address, bypassing your own DNS answer
curl -s -o /dev/null -w "%{http_code} %{remote_ip}\n" \
  --resolve www.example.com:443:203.0.113.10 https://www.example.com/

Run it once with -4 and once with -6 to rule out a broken IPv6 path.

4. Check recent rule changes

If the region gets a 403 or a challenge page, look at firewall, WAF, bot and country rules changed around the time the reports started. A rate limit keyed on a shared office or mobile carrier address can block a whole city.

Watch it from several regions

One checker in one place cannot tell "my site is down" from "the route to my checker is down". A multi-region monitor checks the same URL from several regions and opens an incident only when most of them fail, or when as many fail as you set under Go down when. The incident names the regions that failed and why, so the next report from a customer already has its answer.

To reproduce a report straight away, use Test once from on the monitor page, or the API. Nothing is recorded and the monitor's status does not change:

curl -X POST https://openuptime.app/api/monitors/<id>/check \
  -H "x-api-key: ou_..." -H "Content-Type: application/json" \
  -d '{"region":"eu-central"}'

{ "ok": false, "region": "eu-central",
  "result": { "ok": false, "latencyMs": 10000, "error": "Request timed out" } }

只在某个地区故障

另一个国家的客户说你的网站打不开,而你这边一切正常。这两件事可能同时成立。本文用四步缩小原因范围,再设置一个能看到客户所见情况的检测。

常见原因

原因表现
DNS 变更还没传到所有解析服务器部分用户拿到旧地址,超时或看到旧站点。见 DNS 变更还没生效。
CDN 或负载均衡的某个节点出问题域名和解析结果都一样,但用户附近的节点失败:5xx、连接被重置或首字节很慢。
某个网络到你主机之间的路由问题来自某个运营商或国家的连接卡在建立连接阶段,其他人都正常。
防火墙、WAF 或地区规则某个地区收到 403、验证页或连接重置,常发生在规则修改之后。
IPv6 不通而 IPv4 正常优先使用 IPv6 的网络上的用户失败;AAAA 记录指向一个不响应的地址。

1. 看请求停在哪一步

请受影响的用户,或该地区的一台机器运行下面的命令,并把输出的那一行发给你:

curl -s -o /dev/null https://www.example.com/ \
  -w "dns %{time_namelookup}s  connect %{time_connect}s  tls %{time_appconnect}s  first byte %{time_starttransfer}s  total %{time_total}s  status %{http_code}  ip %{remote_ip}\n"
  • dns 耗时很长,或 ip 和你的不同,说明是 DNS 问题。
  • connect 一直不结束,说明是路由问题,或防火墙丢弃了数据包。
  • tls 阶段报错,说明该地区节点提供的证书有问题。
  • 连接很快但 first byte 很慢或失败,说明是节点或其背后的源站有问题。

2. 对比 DNS 解析结果

变更传播期间,不同的公共解析服务器可能给出不同的结果。如果它们不一致,问题在 DNS,而不是你的服务器:

dig +short www.example.com @1.1.1.1
dig +short www.example.com @8.8.8.8
dig +short www.example.com @9.9.9.9

在浏览器里,DNS 查询工具的“对比解析器”可以同时向 Cloudflare、Google、阿里 DNS 和 DNSPod 查询,并告诉你结果是否一致。

3. 直接测试某个地址

不改任何 DNS,直接把请求发到用户拿到的那个地址。如果这个地址失败而你的正常,问题在它背后的节点或服务器;把地址和地区告诉你的 CDN 或主机商:

# ask a specific edge or origin address, bypassing your own DNS answer
curl -s -o /dev/null -w "%{http_code} %{remote_ip}\n" \
  --resolve www.example.com:443:203.0.113.10 https://www.example.com/

分别加 -4 和 -6 各运行一次,排除 IPv6 线路的问题。

4. 检查最近的规则改动

如果该地区收到 403 或验证页,请查看在报告出现前后修改过的防火墙、WAF、机器人和国家规则。按共享的办公网络或移动运营商地址限流,可能会挡住整个城市的用户。

从多个区域持续检测

只在一个地方检测,无法区分“我的网站挂了”和“到检测点的线路挂了”。多区域监控会从多个区域检测同一个 URL,只在多数区域失败,或失败区域数达到你在判定为故障的条件中设置的数量时才创建事件。事件会写明哪些区域失败以及原因,下一次客户报告时,你已经有了答案。

想立刻复现客户的报告,可以在监控页使用单次测试位置,或调用 API。这不会记录结果,也不会改变监控状态:

curl -X POST https://openuptime.app/api/monitors/<id>/check \
  -H "x-api-key: ou_..." -H "Content-Type: application/json" \
  -d '{"region":"eu-central"}'

{ "ok": false, "region": "eu-central",
  "result": { "ok": false, "latencyMs": 10000, "error": "Request timed out" } }

Last updated 2026-11-16最后更新:2026-11-16