rds: Upgrade(2.142.0 -> 2.143.0) added an explicit dependency between some of our DB constructs.

[mamta925]

Describe the bug

In CDK upgrade	Package	Type	Update	Change
aws-cdk (source)	devDependencies	minor	2.142.0 -> 2.143.0
aws-cdk-lib (source)	dependencies	minor	2.142.0 -> 2.143.0

---

This upgrade added an explicit dependency between some of our DB constructs.

"DbDbClusterreaderB277327D": { "DeletionPolicy": "Delete", "DependsOn": [ "DbDbClusterwriterCE53952E", ], "Properties": { "DBClusterIdentifier": { "Ref": "DbDbCluster8FBC7FAD",

Now, cloudformation wants to delete and recreate the database in order to apply that change, which is obviously not convenient.

So… any ideas how we can deploy this change without deleting the DB and having to re-load the data?

Expected Behavior

It should not delete DB

Current Behavior

Stack update is failing because of DB recreation

Reproduction Steps

const dbCluster = new rds.DatabaseCluster(this, "DbCluster", {
  serverlessV2MinCapacity: 1,
  serverlessV2MaxCapacity: 10,
  writer: rds.ClusterInstance.serverlessV2("writer"),
  readers: [rds.ClusterInstance.serverlessV2("reader")],
  storageEncryptionKey: dbStorageEncryptionKmsKey,
  removalPolicy: RemovalPolicy.DESTROY,
  copyTagsToSnapshot: true,
  clusterIdentifier: dbName,
  securityGroups: [dbSecurityGroup],
  credentials: dbClusterCredentials,
  vpc,
  vpcSubnets: privateSubnets,
  defaultDatabaseName: dbDatabaseName,
  engine: rds.DatabaseClusterEngine.auroraMysql({
    version: rds.AuroraMysqlEngineVersion.VER_3_05_2,
  }),
});

Our old DB constuct , now updating

cdk versions

Possible Solution

not sure if this is causing issue : https://github.com/aws/aws-cdk/issues/30260

Additional Information/Context

No response

CDK CLI Version

aws-cdk (source) devDependencies minor 2.142.0 -> 2.143.0

Framework Version

No response

Node.js Version

20

OS

Linux

Language

TypeScript

Language Version

No response

Other information

No response

[colifran]

I think this PR is where the explicit dependency was added. @pahud any thoughts on this?

[pahud]

Yes that PR ensures the writer would always be created before the reader.

Hi @mamta925

Did you mean you created this in 2.142.0 and now you just tried to upgrade to 2.143.0 and CDK is trying to replace your nodes?

const dbCluster = new rds.DatabaseCluster(this, "DbCluster", {
  serverlessV2MinCapacity: 1,
  serverlessV2MaxCapacity: 10,
  writer: rds.ClusterInstance.serverlessV2("writer"),
  readers: [rds.ClusterInstance.serverlessV2("reader")],
  storageEncryptionKey: dbStorageEncryptionKmsKey,
  removalPolicy: RemovalPolicy.DESTROY,
  copyTagsToSnapshot: true,
  clusterIdentifier: dbName,
  securityGroups: [dbSecurityGroup],
  credentials: dbClusterCredentials,
  vpc,
  vpcSubnets: privateSubnets,
  defaultDatabaseName: dbDatabaseName,
  engine: rds.DatabaseClusterEngine.auroraMysql({
    version: rds.AuroraMysqlEngineVersion.VER_3_05_2,
  }),
});

[pahud]

trying to reproduce this in my account

[pahud]

Hi @mamta925

I created the cluster with 2.142.0 as below:

export class DummyStack extends Stack {
  constructor(scope: Construct, id: string, props?: StackProps) {
    super(scope, id, props);

    const vpc = ec2.Vpc.fromLookup(this, 'Vpc', { isDefault: true });

    const dbCluster = new rds.DatabaseCluster(this, "DbCluster", {
      serverlessV2MinCapacity: 1,
      serverlessV2MaxCapacity: 10,
      writer: rds.ClusterInstance.serverlessV2("writer"),
      readers: [rds.ClusterInstance.serverlessV2("reader")],
      // storageEncryptionKey: dbStorageEncryptionKmsKey,
      removalPolicy: RemovalPolicy.DESTROY,
      copyTagsToSnapshot: true,
      // clusterIdentifier: dbName,
      // securityGroups: [dbSecurityGroup],
      // credentials: dbClusterCredentials,
      vpc,
      // vpcSubnets: privateSubnets,
      // defaultDatabaseName: dbDatabaseName,
      engine: rds.DatabaseClusterEngine.auroraMysql({
        version: rds.AuroraMysqlEngineVersion.VER_3_05_2,
      }),
    });

  }
}

Then I upgraded to 2.143.0. The cdk diff indicates the reader node is going to have a DependOn on the writer with in-place update and it took 10 seconds on the deployment. No node is being replaced.

Now, cloudformation wants to delete and recreate the database in order to apply that change, which is obviously not convenient.

What made you think CFN was trying to delete and recreate the instance?

[mamta925]

Hey, yes this part only, As soon as I added below code(removed dependency ) then every thing worked

   const  cfnDbReader = dbCluster.node.findChild("reader").node.defaultChild as CfnResource;
    cfnDbReader.addOverride("DependsOn", undefined);