This page describes operations that impact availability in Cloud SQL.
Cloud SQL helps you maximize the uptime of your database without any changes to your application. With a 99.99% uptime SLA and near-zero downtime for several planned operations, Cloud SQL Enterprise Plus edition offers greater resilience and business continuity for your applications.
Maximizing uptime means reducing the downtime for your database. Downtime is typically caused by both unplanned events and planned operations.
Unplanned events and zonal outage
For applications that require high availability, we recommend configuring the instance for regional availability, also referred to as the high availability (HA) configuration in Cloud SQL. For more information about configuring high availability in Cloud SQL, see About high availability (HA).
For instances configured with regional availability, if the instance encounters a failure, then Cloud SQL automatically brings up the instance in the secondary zone with the same IP address and no data loss so that applications can resume operations on the database. This protects against unplanned events such as the failure of a single VM host, or the failure of one or more infrastructure components that causes the outage of an entire zone.
Cloud SQL high availability is available for both Cloud SQL Enterprise Plus edition and Cloud SQL Enterprise edition.
If you don't configure regional availability for an instance, then the instance is configured for zonal availability, also referred to as a standalone instance.
Cloud SQL recovers standalone instances (instances configured with zonal availability) from VM host failures automatically. However, Cloud SQL doesn't recover standalone instances from a zonal outage automatically. To re-establish a standalone instance in a healthy zone, you must restore any standalone instances manually. You can recover a standalone instance from a zonal outage manually by performing either point-in-time-recovery or by promoting a read replica to become the standalone instance. For more information, see Recovery options for standalone instances.
Planned operations
For planned operations such as maintenance or some instance reconfigurations, Cloud SQL Enterprise Plus edition offers near-zero downtime for primary instances with no application changes or proxy requirements. With near-zero downtime, a planned operation typically causes less than a second of downtime.
The following planned operations benefit from the near-zero downtime on a Cloud SQL Enterprise Plus edition primary instance. These downtimes are applicable to both standalone (zonal) and high availability (regional) instances.
Maintenance, which includes the following operations with sub-second downtime:
- Automatic maintenance that Cloud SQL performs every quarter
- Self-service maintenance that you can initiate at any time
- Maintenance simulation that you can initiate to test the sub-second downtime behavior
Near-zero downtime
With near-zero downtime, a planned operation typically causes less than a second of downtime for a Cloud SQL Enterprise Plus edition primary instance.
The operation typically takes a few minutes to complete, and your instance stays available during that time. The downtime, meaning the period when the database can't accept connections, is near-zero (less than one second), and the process doesn't require you to change the endpoints that your applications connect to.
Prerequisites and constraints
The near-zero downtime feature is only available with Cloud SQL Enterprise Plus edition. For more information about Cloud SQL editions, see Introduction to Cloud SQL editions.
Near-zero downtime is not available on read replica instances.
Instances that have read replicas are supported only if they meet the following criteria:
- Private services access connectivity enabled. Public-IP replicas and Private Service Connect replicas are supported if they also have private services access enabled.
- At most 7 read replicas.
- No cascadable replicas.
Supports between 1 and 100 databases per instance.
Available for Cloud SQL Enterprise Plus edition edition only.
Available for the following SQL Server editions:
- SQL Server Enterprise 2019
- SQL Server Enterprise 2022
- SQL Server Enterprise 2025
If you are using the Cloud SQL Auth Proxy or Cloud SQL Language Connectors, then ensure that they are updated to their latest version.
Your SQL Server internal server name
SERVERPROPERTY('servername')might change during maintenance. If your application references the hostname or machine name, you might need to update these values after maintenance. Your Cloud SQL connection strings and instance IP address won't change.Supports user databases only. System databases can't be added to availability groups.
The following database types aren't supported:
- Read-only databases
- Single-user databases
- Databases designated as
AUTO_CLOSE
The following features aren't supported:
- Transactional replication setup.
- Self-loop linked servers that use a servername.
- Private Service Connect outbound connectivity.
What's next
View all the Google Cloud services available in locations worldwide.