Your team is executing a zero-downtime database migration from an on-premises sharded relational database to Cloud Spanner. The forward migration strategy uses continuous Change Data Capture (CDC) replication to keep Cloud Spanner in sync with the source database. To ensure business continuity during cutover, you must implement a robust fallback mechanism that allows the application to fail back to the source database with minimal downtime and without data loss if issues occur after cutover. Which strategy should you implement for the reverse replication and validation framework?

Capture change streams from Cloud Spanner, buffer and temporally order the distributed change events, reverse any data transformations or sharding mappings, and continuously write changes back to the source database.: ### What Reverse Replication with Change Streams Is **Reverse replication** is an essential continuity mechanism in large-scale database migrations that streams changes made in the new target database (such as **Cloud Spanner**) back to the legacy source system. This enables immediate fallback with minimal downtime if performance degradation or functional defects arise post-cutover. ### How It Meets the Requirements - **Ensures Zero Data Loss on Fallback**: By continuously capturing change events from Spanner, all updates written after cutover are mirrored back to the source database. - **Handles Distributed Ordering**: Because **Cloud Spanner** is a globally distributed system, change stream records must be buffered and ordered temporally (using Spanner commit timestamps) before being applied to the source. - **Accommodates Schema Differences**: It explicitly reverses schema transformations, type conversions, and source sharding routing rules established during the forward migration. ### Key Advantages and Capabilities - **Minimal Downtime Fallback**: If an issue arises after cutover, traffic can be redirected back to the source database immediately. - **Preserves Sharding Schemes**: Appropriately re-routes writes back across partitioned on-premises source shards. - **Guarantees Transactional Integrity**: Temporal ordering ensures that interdependent transactions are executed in the correct sequence on the target relational database. - **Decoupled Architecture**: Utilizes change streams without placing query locking overhead on operational Spanner workloads. ### Why This Is Optimal Cloud Spanner's distributed architecture requires explicit ordering and transformation reversal to write back cleanly to a traditional single-node or sharded relational database. Capturing change streams and temporally ordering events provides a deterministic, zero-data-loss rollback path.

Cdc Migration Reverse Replication Fallback Strategy | professional-cloud-data-engineer

AdvancedProfessional Cloud Data Engineer

An enterprise is planning a multi-workload migration strategy to Google Cloud and must select the optimal data transfer mechanisms based on data source type, transfer volume, and network availability: Workload 1: A 350 TB historical file archive stored on an on-premises Network Attached Storage (NAS) system. The archive must be moved to Cloud Storage within two weeks, but the facility's available outbound internet bandwidth is constrained to 100 Mbps and cannot be upgraded. Workload 2: A mission-critical on-premises transactional PostgreSQL database that powers an active production system. It must be migrated to Cloud SQL for PostgreSQL with minimal downtime using continuous replication until the final cutover. Workload 3: Ongoing daily reporting feeds from SaaS advertising platforms (Google Ads and Salesforce) that must be automatically ingested on a recurring schedule directly into analytical tables without maintaining custom ETL infrastructure. Which combination of services should the organization select for these workloads?

Planning migration and validation to Google Cloud (e.g., BigQuery Data Transfer Service, Database Migration Service, Transfer Appliance, Google Cloud networking, Datastream)

AdvancedProfessional Cloud Data Engineer

Your enterprise is planning a large-scale migration of several petabytes of transactional and analytical data from an on-premises data center to Google Cloud. The migration architecture must satisfy the following technical and operational requirements: Throughput and Latency: Provide dedicated 10 Gbps+ bandwidth with low, predictable latency to meet tight migration cutover windows. Security and Privacy: Ensure migration traffic remains completely private and never traverses the public internet. Google Services Access: Enable on-premises migration agents to transfer data directly into Google Cloud storage and database services without assigning public IP addresses to internal hosts. Cost Optimization: Minimize network egress and data transfer costs during validation and ongoing replication phases. Which network architecture should you configure to support this data migration?

AdvancedProfessional Cloud Data Engineer

An enterprise organization is designing a centralized data repository in BigQuery to store sensitive financial transactions across multiple regions. The data governance and security framework requires the following: Regional business analysts must be able to query specific analytical aggregations without having read access to the underlying source dataset or base tables. Data engineers requiring direct base table access must have row-level access dynamically restricted to their assigned region using native BigQuery row-level security (RLS). Data classification metadata and billing chargeback keys must be maintained at both the dataset and individual table levels. Which combination of administrative and security configurations meets these governance requirements?

AdvancedProfessional Cloud Data Engineer

An organization is designing an automated Dataflow ingestion pipeline to de-identify sensitive structured records containing Personally Identifiable Information (PII) before storing them in BigQuery for analytics. The privacy compliance team establishes the following technical constraints: Identifier Tokenization: The 10-digit numerical patient_account_id column must retain referential integrity for analytical joins, remain reversible for authorized re-identification pipelines, and preserve its original 10-digit numeric character set to maintain backward compatibility with legacy schema validators. Quasi-Identifier Generalization: The patient_age column must be abstracted into discrete 10-year cohorts to mitigate re-identification risk while preserving population distribution metrics. Key Security: Plaintext cryptographic keys must not be exposed in configuration templates; all token encryption keys must be encrypted using Cloud Key Management Service (Cloud KMS). Which transformation configuration in Cloud Sensitive Data Protection (Cloud DLP) meets all of these requirements?

AdvancedProfessional Cloud Data Engineer

A financial services organization ingests millions of analytical audit files daily into a Google Cloud Storage Standard bucket. Strict compliance regulations mandate that audit files must be retained immutably for exactly 7 years (2,555 days), preventing any user—including bucket and project administrators—from deleting or modifying them before the retention period expires. To optimize storage costs, the team identified the following operational requirements: Audit files larger than 10 MiB should automatically move from Standard to Coldline storage after 30 days of inactivity, avoiding transition costs on smaller files where operation fees exceed storage savings. Files should be permanently purged automatically as soon as the 7-year compliance period lapses. Which combination of Cloud Storage configurations satisfies all compliance and cost-optimization requirements?

IntermediateProfessional Cloud Architect

An e-commerce company is evolving its data strategy from simple raw log archiving in Cloud Storage to an active operational intelligence and machine learning ecosystem. Multiple data science and analytics teams need to spin up ephemeral, short-lived compute clusters running Apache Spark and Hive to process massive datasets and drive personalized customer recommendations. The architecture must preserve table schemas, column statistics, and partition definitions across transient cluster lifecycles while enabling seamless metadata interoperability with Google Cloud services like BigQuery and Dataplex Universal Catalog. Which solution should you implement to manage the centralized data lake metadata?

Professional Cloud Data Engineer

Cdc Migration Reverse Replication Fallback Strategy

Related questions