There are three main types of RDS maintenance as follows.
RDS underlying hardware maintenance
This is the replacement of the underlying host.
How to check the maintenance
Email notification : A notification is sent including the maintenance schedule, affected instance ID, and the instance’s availability zone.
Check on PHD (Personal Health Dashboard) (not available on the RDS Console)
Scheduled maintenance cannot be postponed.
Downtime depends on the availability zone configuration.
In the case of Single AZ, downtime occurs throughout the maintenance period.
In the case of Multi AZ, downtime depends on the configuration.
Mitigation methods
Stop and start the instance in advance during a time convenient for the user.
Manually perform failover in advance.
OS and Security patch
How to check the maintenance
RDS Console
PHD
AWS CLI
Maintenance windows can be adjusted to postpone the schedule.
Most maintenance is performed online, but downtime can occur depending on the availability zone configuration.
DB engine maintenance
This is an upgrade at the DB engine level.
Downtime occurs.
Downtime lasts throughout the maintenance period because the DB engine needs to be restarted.
Downtime also occurs in Multi-AZ because the primary and secondary instances are upgraded simultaneously.
The maintenance schedule can be postponed.
Minor version upgrades can be postponed by adjusting the maintenance window.
RDS deprecation engine version upgrades cannot be postponed.
Notices are given 3 to 6 months before version support ends.