Question

Difficulty: Very hardDatabase Migration and Schema Conversion using DMS and SCT

A Solutions Architect is planning the heterogeneous database migration of an on-premises Oracle 19c database to an Amazon Aurora PostgreSQL-Compatible Edition DB cluster. The source database supports a critical OLTP application experiencing a high volume of transactions, exceeding 12,00012,000 write transactions per second. The schema contains complex PL/SQL objects, standard relational tables, and several tables containing Large Binary Objects (LOBs) up to 5 MB5\text{ MB} in size. The migration must minimize downtime for the cutover, maintain high replication performance without degrading the source database's CPU utilization, and guarantee that LOB data is not truncated.

Which two actions must the Solutions Architect take to configure the migration components successfully? (Select two.)

  1. Configure the AWS DMS source endpoint with extra connection attributes to use the AWS DMS Binary Reader instead of Oracle LogMiner to process redo and archive logs directly from the file system.Answer
  2. B
    Configure the AWS DMS replication task to use Limited LOB mode with a Max LOB size of 1,024 KB1,024\text{ KB} to optimize initial transfer speed, allowing DMS to automatically query the source database to fetch the remaining LOB content exceeding this limit during replication.
  3. Enable database-level minimal supplemental logging and table-level primary key supplemental logging on the source Oracle database before starting the replication task.Answer
  4. D
    Install the AWS SCT agent on the source database server and configure it to stream real-time transaction changes directly to the target Aurora PostgreSQL DB cluster using PostgreSQL logical replication slots.
  5. E
    Enable Amazon Aurora Auto Scaling on the target DB cluster and configure the AWS DMS replication task to write directly to the Aurora Reader endpoint to distribute the write load across multiple read replicas.

Answer

Configure the AWS DMS source endpoint to use the Binary Reader instead of LogMiner, and enable database-level minimal supplemental logging along with table-level primary key supplemental logging on the source Oracle database.
For high-volume transaction databases (exceeding 10,000 write TPS), using the default Oracle LogMiner API introduces severe CPU and memory overhead on the source database and can lead to lag in capturing changes. The Solutions Architect should use the AWS DMS Binary Reader, which reads the redo and archive logs directly from the file system. Furthermore, AWS DMS CDC requires minimal supplemental logging at the database level and table-level primary key supplemental logging to ensure the logs contain enough context to reproduce SQL statements on the target.

Step-by-Step Solution

1
Configure Oracle supplemental logging on the source database.
Database-level minimal supplemental logging and table-level primary key supplemental logging are enabled, which ensures update and delete events write full row identifiers to the redo logs.
Required for AWS DMS CDC to successfully parse and apply updates/deletes to target tables.
2
Set up extra connection attributes on the AWS DMS Oracle source endpoint to enable the AWS DMS Binary Reader.
DMS switches from using the Oracle LogMiner API to reading redo and archive log files directly from the directory structures.
Avoids the CPU overhead and performance limitations of the LogMiner API under transaction rates exceeding 10,000 TPS.
3
Configure target endpoint routing to the Aurora writer endpoint and select appropriate LOB replication tasks.
The replication task writes to the active writer and successfully processes LOBs up to 5 MB without truncation (either through Full LOB mode or configured Inline LOB settings).
Aurora Reader endpoints do not support writes, and Limited LOB mode with lower thresholds truncates data.

Key Concept

Heterogeneous migration from Oracle to Aurora PG with high throughput and LOB constraints requires bypassing LogMiner with DMS Binary Reader and enabling mandatory supplemental logging.
Rate this question