High availability, often shortened to HA, is a design that keeps a service running when parts of it fail. For a database, this usually means keeping a standby copy in another location, often a different zone, that can take over automatically if the primary fails. The aim is to reduce downtime and the need for manual intervention.
HA improves resilience, but it does not eliminate interruption. A failover can still cause a short outage while the standby takes over and applications reconnect, and some in-flight work may fail. Applications need to handle connection errors and retry sensibly to benefit fully.
HA is also not a replacement for backups. A standby usually copies changes from the primary, so a mistake such as an accidental deletion or corruption can be replicated to it. Separate backups and point-in-time recovery protect against a different kind of failure. HA typically costs more because it runs additional resources.
When choosing it, check what level of failure the setup covers, such as a single instance, a zone or a region, how long failover takes and what it costs. Test failover in a non-production environment and confirm what availability the provider commits to in its service level agreement.