Aurora version upgrade failing due to parameter group

[aniruddhaEFS]

SST version: 4.14.1
Related to #6261 #6585 #6586

When upgrading Aurora 16.8 to 17.7, I get parameter group related error during deploy

sdk-v2/provider2.go:572: sdk.helper_schema: updating RDS Cluster Instance (xxxx-xxxx-xxxx-xxx-aurorainstance-batbauuv): operation error RDS: ModifyDBInstance, https response error StatusCode: 400, RequestID: eba3e97a-de23-4c35-9de1-06478a6dcf29, api error InvalidParameterCombination: The parameter group xxxx-xxxx-xxxx-xxx-auroraparametergroup-duxxhkxh with DBParameterGroupFamily aurora-postgresql17 can't be used for this instance. Use a parameter group with DBParameterGroupFamily aurora-postgresql16.: provider=aws@7.20.0

Compared to #6261, the only difference I can see is that I have transform>clusterParameterGroup

new sst.aws.Aurora('aurora', {
      engine: 'postgres',
      dataApi: true,
      database: 'xxxx',
      vpc,
      scaling,
      version: '16.8',
      replicas,
      transform: {
        cluster: {
          backupRetentionPeriod,
          enabledCloudwatchLogsExports: ['postgresql'],
          performanceInsightsEnabled: true,
          performanceInsightsRetentionPeriod,
        },
        clusterParameterGroup: (options: aws.rds.ClusterParameterGroupArgs) => {
          options.parameters = [
            {
              name: 'rds.logical_replication',
              value: '1',
              applyMethod: 'pending-reboot',
            },
            {
              name: 'wal_sender_timeout',
              value: '60000',
            },
            {
              name: 'max_worker_processes',
              value: '16', //'GREATEST(${DBInstanceVCPU*2},8)',
              applyMethod: 'pending-reboot',
            },
            {
              name: 'shared_preload_libraries',
              value: 'pg_stat_statements,pglogical,pgaudit',
              applyMethod: 'pending-reboot',
            },
            {
              name: 'pgaudit.log',
              value: 'all',
            },
            {
              name: 'pgaudit.log_parameter',
              value: '1',
            },
            {
              name: 'log_statement',
              value: 'all',
            },
            {
              name: 'log_duration',
              value: '1',
            },
          ];
        },
      },
    });

The deployment fails but the state does not get corrupted, so it's not a big blocker.

[vimtor]

did you check #6586?

maybe that gives you some ideas, but it feels like transform parameter groups is to blame here

[aniruddhaEFS]

Yes, it looks like transform>clusterParameterGroup is not handled correctly.

I need the parameter changes for DMS replication to work, so one alternative could be to do the migration step by step - pause DMS, remove transform>clusterParameterGroup during upgrade, re-add transform>clusterParameterGroup, resume DMS. Not ideal but it could work.