GCP Database Engineer study materials
GCP Database Engineer study resources

GCP Database Engineer: Migrate data solutions

GCP Database Engineer · Work through your exam, one topic at a time.

All guide sections

Migrate data solutions

This domain tests planning and executing migrations with minimal downtime and verified correctness. The exam distinguishes homogeneous migrations, heterogeneous conversions, online versus offline transfer, and the tools Google Cloud provides for each.

Pick the migration method by source and downtime tolerance

  • Database Migration Service (DMS): homogeneous migrations such as MySQL, PostgreSQL, and SQL Server to Cloud SQL, using an initial snapshot plus continuous replication and a promotion cutover.
  • Spanner Migration Tool: assesses and converts MySQL or PostgreSQL schemas and data for Spanner, including type and key mapping.
  • BigQuery Data Transfer Service: scheduled ingestion from SaaS and Google sources; not a database CDC tool.
  • Storage Transfer Service: online object transfer from other clouds or on-premises; Transfer Appliance for large offline datasets where network transfer is impractical.

Validate before cutover

Run a test migration and compare row counts, checksums, charset and collation behavior, and application queries. For DMS, confirm binary log retention on the source, network connectivity to the destination, and a rollback plan. For heterogeneous moves, validate stored procedures, triggers, and application SQL that may not convert cleanly.

Worked example

A 2 TB on-premises MySQL database must move to Cloud SQL with under one hour of downtime. Do not use a single mysqldump; a full dump and reload would exceed the window. Instead, configure DMS with a continuous replication job, let the initial snapshot and CDC catch up, run the application in read-only mode briefly, verify replication lag is near zero, then promote Cloud SQL and repoint the application. Keep the source available for rollback until validation passes.

Traps that cost marks

Using offline dumps for large databases with tight downtime; forgetting that DMS requires source log retention; ignoring definer/invoker and lower_case_table_names differences; skipping post-migration row-count and query validation; and assuming Transfer Appliance is for online incremental sync rather than bulk offline transfer.

Key rules

Prefer continuous replication for minimal downtime, always test and validate, and size the transfer method to data volume and network capacity.