BlackTree Security · Infrastructure · Automation · AI

Cloud SQL’s Default Has Changed

Existing instances move only when their version is eligible, automatic upgrades are enabled and regular maintenance reaches them.

On 10 October 2026, Google made MySQL 8.0.46 Cloud SQL’s default for MySQL 8.0. This changes the managed service’s target, not the upstream release date.

Which instances follow the new default?

Instance state Upgrade route
New MySQL 8.0 instance with no minor Uses the default with automatic upgrades enabled.
Existing 8.0.35 or later instance with automatic upgrades on Moves to the default in regular scheduled maintenance.
Existing 8.0.35 or later instance with the setting off The default change does not opt it in.
Existing 8.0.34 or earlier instance Upgrade the minor manually first; enable the automatic option after reaching 8.0.35 or later.

Automatic moves use regular scheduled maintenance, not self-service maintenance.

Maintenance still requires a recovery plan

Cloud SQL maintenance restarts an instance and causes downtime. Impact varies by edition, workload, data size, replicas and client behaviour. Enterprise Plus has a near-zero downtime route only for eligible configurations with prerequisites.

Operators can set a window and rollout timing; notifications require opt-in. An overlapping operation cancels maintenance, and disk use above 97 per cent causes it to be skipped. Connection, pooling, timeout and query settings affect the interruption; replicas need separate checks.

Build the upgrade record before the event

BlackTree recommends:

  1. Capture eligibility. Record the instance, region, edition, current minor version and automatic upgrade setting. Put instances below 8.0.35 on a separate manual path.
  2. Control maintenance. Record the window, rollout timing, notification state, replica topology, available capacity and any operation that could overlap the event.
  3. Test application recovery. Exercise drivers, authentication, connection pools, representative queries, scheduled jobs and background workers in a representative environment. Rehearse the restart behaviour appropriate to the edition.
  4. Verify the result. After maintenance, confirm the running minor, application health, replication, jobs and monitoring. Keep the result with the service record.

The default is a planning signal. The instance version, enabled setting and observed maintenance result remain the evidence for each service.

Sources

Related: PostgreSQL 14 Has Different Support Deadlines in the Cloud.

Leave a Reply

Your email address will not be published. Required fields are marked *