When a website goes down, the first question everyone asks is: how long will this last? The answer depends almost entirely on what caused the outage. Knowing the cause tells you whether to wait five minutes or five hours.
Deployment errors: 15 to 45 minutes
The most common cause of sudden downtime is a bad deployment — a code or configuration change that broke something in production. Engineering teams catch these quickly through automated alerts and monitoring. The fix is usually a rollback to the previous version, which takes 15 to 45 minutes for most modern deployments.
Signs of a deployment-caused outage: it starts suddenly at an unusual hour on a weekday, affects specific features rather than the whole site, and resolves cleanly when the rollback completes.
Infrastructure failures: 30 minutes to 4 hours
Cloud provider incidents — AWS, Google Cloud, Azure, Cloudflare — tend to last longer than deployment errors because the fix is in someone else's hands. The affected company cannot resolve it themselves; they are waiting on the infrastructure provider.
These outages typically last 30 minutes to four hours. Major incidents (like the 2021 AWS us-east-1 event or the 2021 Meta BGP failure) can extend to six or eight hours, but these are rare.
Traffic spike crashes: minutes to hours
When a site gets hit with more traffic than its infrastructure can handle — from a viral moment, a product launch, or a DDoS attack — the outage duration depends on how quickly engineers can scale capacity or implement mitigation.
For well-architected cloud services, auto-scaling kicks in within minutes and the site recovers. For fixed-capacity infrastructure, recovery might require manual intervention and could take hours.
Database failures: 30 minutes to several hours
Database failures are among the most disruptive types of outages because so many application functions depend on data persistence. A database crash, replication failure, or storage issue can take considerable time to diagnose and repair.
Signs of a database-related outage: the site loads (HTML, CSS, images are served) but functionality is broken — you cannot log in, content does not load, or actions return errors.
DNS failures: minutes to 48 hours
DNS failures are unusual because the fix itself can take time to propagate. If a domain's DNS records are misconfigured or the DNS provider has an outage, correcting the record is only part of the fix. DNS changes must propagate across resolvers globally, which can take anywhere from minutes (with a low TTL) to 48 hours (with a high TTL or aggressive resolver caching).