OpenUptime

Glossary

Maintenance window

A maintenance window is a period set aside in advance for planned work, such as upgrades or migrations, during which the service may be unavailable.

Pick a time when the fewest users are affected, which for a global product may not be night in your own time zone. Announce it in advance on the status page, say what will be affected and for how long, and post when it is over.

Alerts should be held during the window so on-call is not paged for planned work, but they should resume as soon as it ends: a service still down after the window is a real outage.

Whether the window counts against uptime depends on the SLA. Many exclude announced maintenance up to a limit.

In OpenUptime

OpenUptime maintenance windows can be one-off or recurring, are shown on status pages and suppress alerts for the monitors they cover while they run.

Questions

How far in advance should maintenance be announced?

A common practice is several days for routine work and longer for anything with a large impact. Follow what your SLA or customer contracts require.

Can maintenance be done without downtime?

Often. Rolling deploys, database replicas and blue-green switches let most changes happen without users noticing. Windows are still useful for the risky ones.

Monitor it with OpenUptime

Checks from one or several regions, alerts to the tools you use, and a status page for your customers. Free to start, open source under the MIT license.

术语表

维护窗口(Maintenance window)

维护窗口是提前预留的一段时间,用于升级、迁移等计划内工作,期间服务可能不可用。

选择影响用户最少的时间,对于全球化的产品,这未必是你所在时区的夜里。提前在状态页上公告,说明会影响什么、持续多久,结束后再发布通知。

维护期间应暂停告警,免得值班人员因为计划内工作被叫醒;但窗口一结束就要立即恢复:窗口结束后服务仍然宕机,就是真正的故障。

维护窗口是否计入停机时间,取决于 SLA。很多 SLA 会在一定额度内排除提前公告的维护。

在 OpenUptime 中

OpenUptime 的维护窗口可以是一次性的或周期性的,会显示在状态页上,并在进行期间暂停所涉及监控的告警。

常见问题

维护应该提前多久公告?

常见做法是例行维护提前几天,影响较大的更早公告。如果 SLA 或客户合同有要求,按要求执行。

维护能做到不停机吗?

很多时候可以。滚动发布、数据库副本和蓝绿切换能让大多数变更在用户无感的情况下完成。对风险较高的变更,维护窗口仍然有用。

用 OpenUptime 监控

从一个或多个区域检查,告警发到你常用的工具,还有面向客户的状态页。免费开始使用,以 MIT 许可证开源。