OpenUptime

Uptime monitoring for e-commerce

For a store, a page that loads but cannot sell is as bad as a page that is down. These are the checks that catch that, and the habits that keep alerts useful during a sale.

What to monitor

The page loads and still sells

Use an HTTP monitor on a product page and require text that only appears when the page is healthy, such as the add-to-cart label. A second monitor can require that an error message does not appear. A shop that returns 200 with an empty catalogue is then reported as down.

  • HTTP
  • keyword must contain
  • keyword must not contain

Shoppers in different markets

Check the store from the regions where your customers are, so a CDN or routing problem in one market shows up. With several regions the monitor goes down when most of them fail, or you can set how many must fail. The region of each failure is written into the incident.

  • multi-region
  • minimum failing regions

Slow is the first sign of trouble

Every monitor has a slow-response threshold, 1500 ms by default. Slow checks mark the monitor as degraded, and you can choose to open a minor incident after a number of consecutive slow checks. During a sale this tells you about strain before it becomes an outage.

  • slow-response alerts

Domain, feeds and order sync

A DNS monitor confirms that your domain still resolves as expected. Heartbeat monitors cover jobs that run without a page, such as the product feed export or the order sync to your warehouse: the job pings when it finishes, and a missed ping alerts.

  • DNS
  • heartbeat

Deploys during a sale

Put a maintenance window around planned releases. Checks keep running and are recorded, but no incident opens and no alert fires for the covered monitors.

  • maintenance windows

Status page

A public status page lets shoppers and support staff check whether the store is having trouble before they write in. Show the storefront, checkout, and payments callback as separate components. Email and RSS subscribers hear when an incident opens or resolves, and the embeddable badge can sit in your footer or help centre.

Alert routing

Send checkout failures to the channel that wakes someone, such as PagerDuty or a phone-notified Telegram chat, and slow-response warnings to a quieter Slack channel. Per-monitor routing makes that split possible.

Limits to know about

  • Checks are single requests; OpenUptime cannot walk through a cart and payment flow.
  • No TLS certificate expiry warning; keep a renewal reminder for your store's certificate.
  • Checks run at most once a minute.

Questions

Can OpenUptime test the checkout flow?

Not as a whole flow. It checks one request at a time. You can monitor the cart page, the checkout page and your payment callback endpoint as separate monitors and require the right keyword in each.

How do I get warned about a slow store before it goes down?

Set the slow-response threshold on the monitor and choose to open a minor incident after several consecutive slow checks. The monitor shows as degraded in the meantime.

Does OpenUptime monitor my SSL certificate expiry?

No. Set a renewal reminder with your certificate provider or use another tool for it. An HTTPS monitor fails once the certificate is rejected.

Set it up in OpenUptime

Start on the hosted service, or run it yourself. The source is on GitHub under the MIT license.

电商站点的可用性监控

对店铺来说,页面能打开却无法下单,与页面打不开一样糟糕。下面是能发现这种问题的检测方式,以及在促销期间让告警保持有用的做法。

监控什么

页面能打开,也能下单

对商品页使用 HTTP 监控,并要求页面包含只有正常时才会出现的文字,例如“加入购物车”。再加一个监控,要求错误提示不得出现。这样返回 200 但商品目录为空的店铺也会被判定为故障。

  • HTTP
  • keyword must contain
  • keyword must not contain

不同市场的顾客

从顾客所在的区域检测店铺,这样某个市场的 CDN 或线路问题会被发现。选择多个区域后,多数区域失败时监控才判定为故障,你也可以自行设置需要几个区域失败。事件中会写明失败发生在哪些区域。

  • multi-region
  • minimum failing regions

变慢往往是出问题的第一个信号

每个监控项都有响应变慢的阈值,默认 1500 毫秒。检测变慢时监控会标记为降级,你也可以设置连续变慢若干次后创建一个轻微事件。促销期间,这能让你在宕机之前就察觉压力。

  • slow-response alerts

域名、商品数据源与订单同步

DNS 监控用于确认域名仍然按预期解析。心跳监控覆盖没有页面的任务,例如商品数据源导出或同步订单到仓库系统:任务完成时发送请求,漏掉就会告警。

  • DNS
  • heartbeat

促销期间发布

为计划内的发布设置维护窗口。检测会继续运行并被记录,但窗口覆盖的监控项不会创建事件,也不会触发告警。

  • maintenance windows

状态页

公开状态页能让顾客和客服在联系你之前先查看店铺是否出了问题。把店铺前台、结账、支付回调作为不同的组件展示。邮件和 RSS 订阅者会在事件出现或解决时收到通知,可嵌入的徽章也可以放在页脚或帮助中心。

告警路由

把结账故障发到会叫醒人的渠道,例如 PagerDuty 或会推送到手机的 Telegram 群,把响应变慢的提醒发到较安静的 Slack 频道。按监控项路由使这种区分成为可能。

需要了解的限制

  • 检测针对单个请求;OpenUptime 无法走完购物车和支付流程。
  • 没有 TLS 证书到期预警;请为店铺证书另设续期提醒。
  • 检测最快每分钟一次。

常见问题

OpenUptime 能测试结账流程吗?

不能测试完整流程。它每次检测一个请求。你可以把购物车页、结账页和支付回调接口分别建成监控项,并要求各自包含正确的关键字。

如何在店铺宕机之前就收到变慢的警告?

在监控项上设置响应变慢的阈值,并选择连续变慢若干次后创建轻微事件。在此期间监控会显示为降级。

OpenUptime 会监控 SSL 证书到期吗?

不会。请在证书提供方处设置续期提醒,或使用其他工具。证书被拒绝后,HTTPS 监控才会失败。

在 OpenUptime 中设置

可以直接使用托管服务,也可以自己运行。源码在 GitHub,采用 MIT 许可证。