AWS::RDS::DBCluster - [BUG] - Resolving drift with Parameter Group always restarts the instances

[blimmer]

Name of the resource

AWS::RDS::DBCluster

Resource Name

No response

Issue Description

I've been using the Aurora Blue/Green deployment to conduct major version upgrades. Since CloudFormation does not natively support this (see #1483), the process is basically:

1. Create a new parameter group for the new major version via CDK.
2. Create the Blue/Green cluster manually via the console (using the parameter group from step 1).
3. Switchover to the green cluster via the console.
4. Resolve CDK drift by syncing the EngineVersion and DBClusterParameterGroupName values.

In effect, Step 4 is a no-op, since this change was applied manually via the console.

Expected Behavior

When I update the EngineVersion property to sync with reality (e.g., a major version via the process above, or to sync a CloudFormation template with an auto-minor upgrade that occurred), CloudFormation "knows" that it's a no-op and simply applies the template without trying to force an upgrade to a different version.

Therefore, when I sync the DBClusterParameterGroupName from the old value in the template to the new value in reality, I do not expect a reboot of all instances to occur (which causes an interruption to availability).

Observed Behavior

When I update DBClusterParameterGroupName to sync it with reality, CloudFormation always triggers a reboot of all cluster instances, even though the parameter group was already applied.

Test Cases

You can follow my steps above, inspired by the blogs linked here: #1483 (comment)

When you resolve the drift by updating the CloudFormation template to have the same DBClusterParameterGroupName as reality, you'll observe that all DB cluster instances will restart.

Other Details

No response