Downtime is any stretch of time when your website or service is not available to the people trying to use it. That includes the site not loading at all, and also partial failures, like a checkout that errors while the rest of the site looks fine.
Planned versus unplanned downtime
Planned downtime is scheduled: a server upgrade, a database migration, a host's maintenance window. You can warn customers and pick a quiet hour. Unplanned downtime is the outage nobody chose, and it is the kind that hurts. Most uptime guarantees count only the unplanned kind.
The usual causes of a site going down
- An expired domain or SSL certificate. The domain stops resolving, or browsers show a full-page security warning that most visitors will not click past.
- DNS errors. A mistyped record during a change sends visitors nowhere.
- A bad update. A plugin, theme or code change that breaks the site the moment it goes live.
- Traffic spikes or resource limits, especially on shared hosting.
- Attacks such as a DDoS flooding the server.
- Outages at the provider, including a cloud region or a third-party service your site depends on.
A well-known example
On 28 February 2017, an incorrectly entered command during routine work on Amazon S3 in AWS's US-East-1 region took down a large part of the service for several hours. Thousands of unrelated websites and apps that stored files there broke at the same time, and AWS's own status dashboard initially could not show the problem because it relied on S3 too. It is a good illustration that downtime often comes from a dependency rather than your own server.
How much downtime should worry you
That depends on what the site does. A few minutes offline overnight on an informational site may go unnoticed. The same outage on a site that takes bookings or payments during business hours costs sales and trust. Know which pages make you money and make sure those are the ones you monitor.
Mistakes that make downtime longer
- Auto-renew turned off, or a card on file that has expired, for the domain and hosting.
- No rollback plan. If an update breaks things, you should be able to restore the previous version in minutes.
- Making changes at peak hours, such as editing DNS on a Friday afternoon.
- Nobody knows where the logins are. During an outage, hunting for the registrar or host password wastes the most time.
Related terms
- Uptime: how availability is measured and promised.
- DDoS: an attack designed to cause downtime.
- DNS: a frequent source of self-inflicted outages.
- SSL certificate: its expiry blocks visitors.