Data and Database Migration
Domain scope
This domain covers data and database migration: classification, movement methods, schema conversion, validation, cutover, and rollback. The exam tests consistency and recoverability, not just copy speed.
Core rules
- Classify data by sensitivity, residency, retention, and owner. Apply encryption, access control, and audit before bulk movement.
- Choose offline for small or tolerant datasets; online with change data capture for large or near-zero-downtime cuts. CDC must handle schema changes and deletes, not only inserts.
- Convert schema objects, then test sequences, triggers, stored procedures, constraints, collations, and time zones. These are frequent breakpoints.
- Validate at three levels: row counts, checksums or key-business aggregates, and application-level reconciliation. Record acceptance thresholds before cutover.
- Plan rollback with a reverse replication path or a frozen source. Define the point of no return and who can declare it.
Worked example
A 2 TB Oracle order database moves to a managed PostgreSQL service with a four-hour window. The team runs schema conversion, a full load, and CDC from the source. Before cutover it compares row counts, per-table checksums, and daily order totals. During the window it stops writes, drains CDC lag to zero, runs application reconciliation, switches connection strings, and keeps reverse replication for 48 hours. The rollback trigger is failed reconciliation on more than 0.1% of orders or CDC lag above five minutes at the point of no return.
Traps
Treating data migration as the last task; copying data without owners and classification; assuming CDC captures deletes and DDL; validating only row counts; and having no reverse path after the source is frozen.