Downtime
Downtime is the time during which a service is unavailable or does not work as expected.
Planned downtime is announced in advance, for maintenance or migrations. Unplanned downtime comes from failures: a crashed process, an expired certificate, a full disk, a bad deploy, a provider outage.
Not all downtime is total. A service can be up in one region and down in another, or answer the home page while checkout fails. Whether partial outages count, and how, is a choice to write down before you report a number.
Downtime is the sum of how long it takes to notice a problem, to start working on it and to fix it. Monitoring shortens the first part, alert routing the second, rollbacks and runbooks the third.
At a glance
allowed downtime = (100% − uptime target) × length of the period
In OpenUptime
In OpenUptime an incident opens when a monitor goes down and resolves when it recovers, so each outage has a start, an end and a timeline. Maintenance windows suppress alerts while planned work is under way.
Work out the downtime a target allows →
Questions
Does planned maintenance count as downtime?
It depends on the agreement. Many SLAs exclude announced maintenance windows; others count every minute. Check the definition before comparing numbers.
What causes most downtime?
Changes are a common cause: deploys, configuration edits and expired certificates or domains. That is why rollbacks and expiry monitoring prevent a large share of outages.
Related terms
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.
停机时间（Downtime）
停机时间是指服务不可用、或无法按预期工作的那段时间。
计划内停机会提前公告，用于维护或迁移。计划外停机来自故障：进程崩溃、证书过期、磁盘写满、发布出错、云服务商故障等。
停机并不总是完全中断。服务可能在一个区域正常、另一个区域不可用，也可能首页能打开而结账失败。部分故障算不算、怎么算，需要在报告数字之前先约定好。
停机时间由三段组成：发现问题、开始处理、修复完成。监控能缩短第一段，告警分发缩短第二段，回滚和处理手册缩短第三段。
速览
允许的停机时间 =（100% − 可用率目标）× 周期长度
在 OpenUptime 中
在 OpenUptime 中，监控宕机时会自动创建事件，恢复时自动解决，所以每次故障都有开始时间、结束时间和时间线。维护窗口期间会暂停告警。
常见问题
计划内维护算停机吗？
取决于协议。很多 SLA 不计入提前公告的维护窗口，也有的每一分钟都算。比较数字前先看清定义。
大多数停机是什么原因造成的？
变更是常见原因：发布、配置修改，以及证书或域名过期。所以快速回滚和到期监控能避免相当一部分故障。