PCDBE — frequently asked questions

While Cloud Functions can be triggered by HTTP or other events, Google's recommended pattern for scheduled Cloud SQL exports uses Pub/Sub as an intermediary to decouple the scheduler from the function, improving reliability and scalability.

Cloud Composer is designed for complex DAGs with dependencies. For simple, independent daily tasks like exports, it adds unnecessary cost and maintenance complexity compared to Serverless options.

No. User accounts are for humans. Workloads must use service accounts to authenticate via OAuth2 tokens generated by the metadata server.

No. Best practice is to create dedicated service accounts for each distinct workload or application to enforce least privilege.

While Datastream and Dataflow provide robust streaming, they require building and maintaining a complex pipeline. Database Migration Service is more 'managed' and 'easy to set up' for the initial replication step.

Federated tables allow BigQuery to query data stored in other systems, such as Cloud SQL, without physically moving the data. This reduces latency and setup effort.

PSC keeps traffic within Google's private network backbone, avoiding the public internet and reducing the attack surface for sensitive data.

Yes, Cloud Interconnect is also valid for hybrid connectivity, but PSC is often simpler to configure for specific service connections like Cloud SQL.

Upgrading the edition of a Cloud SQL instance is generally a non-disruptive operation. However, it is best practice to verify the specific maintenance window settings.

No, Standard and Enterprise editions store WALs on the instance's local persistent disk. Only Enterprise Plus supports native Cloud Storage retention for WALs.

AlloyDB is optimized for transactional PostgreSQL workloads. Bigtable is specifically designed for high-throughput, low-latency access to large amounts of versioned time-series data.

No. Autoscaling manages resource allocation within a cluster or region. Multi-region deployment provides geographic redundancy, lower latency for global users, and protection against regional failures.

No. PSA connects your VPC to Google's managed services. It does not create routes between two customer-owned VPCs unless they are already peered.

Yes, the proxy uses Google Cloud APIs to authenticate and create a secure tunnel, but it still requires network connectivity to the database instance.

Modern conversion workspaces use advanced compilers to automatically translate Oracle PL/SQL code, reducing manual effort. Legacy workspaces may require more manual intervention or offer less sophisticated translation capabilities.

Change Data Capture (CDC) allows continuous replication of data changes from the source to the target while the migration is ongoing. This ensures that only minimal changes need to be applied during the final cutover, reducing downtime.

Advanced DR is available only in the Cloud SQL for MySQL Enterprise Plus edition. It is not supported in Standard or Basic editions.

No, the read replica must run the same minor version as the primary instance to maintain compatibility and support the designated failover role.

No, read replicas in Memorystore Standard Tier are confined to the same region. They protect against zone failures but not regional ones.

Local snapshots are not automatically available in other regions. You must explicitly schedule exports to a Cloud Storage bucket in the target region or a dual-region bucket.