AWS DMS copies data and can apply ongoing changes to support a controlled cutover.
Full load and change data capture
AWS Database Migration Service (DMS) moves data between supported data stores and can keep a target updated while the source changes. Think of a full loadA full load copies existing source data to a target. as the initial copy and change data captureChange data capture reads ongoing changes from a source data store so they can be replicated. (CDC) as the stream of later changes.
How it works
- Use source and target endpointsA DMS endpoint holds connection, type and location information for a data store.. At least one endpoint must be an AWS service.
- Select a replication instanceA DMS replication instance reads data from the source and loads it into the target., which reads and loads data, or use DMS Serverless to provision and scale capacity automatically. Multi-AZ gives a provisioned instance a synchronous standby in another Availability Zone.
- Choose the task type. Full load plus CDC copies existing data while caching changes for each table once its load starts, applies those changes, then continues replication from source logs. Configure the source for CDC as required by its engine.
- After replication reaches a steady state, stop the application, allow remaining changes to reach the target, and restart the application against the target.
| Task | Use |
|---|---|
| Full load only | Copy existing data, including views when needed. |
| Full load plus CDC | Copy existing data and continue applying changes. |
| CDC only | Replicate changes after the target already has its initial data. |
Exam cues
- If the question says reduce migration cutover downtime, think full load plus CDC, because ongoing changes are applied after the initial copy.
- If the question says automatic failover of a provisioned DMS instance, think Multi-AZ, because it maintains a synchronous standby.
- If the question says the target already has its initial data, think CDC only, because the task replicates ongoing changes.
- If the question says replication capacity should scale without instance management, think DMS Serverless.
- If the question says a MongoDB-compatible database, think Amazon DocumentDB. If it says a data warehouse for large analytical queries, think Amazon Redshift. Redshift omits some features suited to small transactions, such as efficient single-row data manipulation. DMS moves data rather than serving as the destination database.
Remember
- CDC is not real time and has no latency SLA. Latency is the delay in applying changes.
- Use a full-load-only task to migrate views.
Trap: CDC is not real time and has no latency SLA. Use a full-load-only task when the migration must include views.
DMS reads source data and captures changes.
DMS loads the target and applies captured changes.
Da documentação da AWS (18)
- AWS DMS migrates data between supported data stores and can replicate ongoing changes.
- At least one endpoint must be on an AWS service; DMS cannot migrate between two on-premises databases.
- DMS supports full-load-only tasks, full load plus ongoing replication and ongoing-replication-only tasks.
- A full load plus CDC task loads existing data, applies cached changes and then applies ongoing changes as transactions.
- During full load DMS starts capturing changes for a table when that table's full load starts and applies cached changes after its load completes.
- CDC reads ongoing changes from source database logs using engine-specific methods; source configuration requirements vary by engine.
- DMS CDC is not real-time replication and has no latency SLA.
- After replication reaches steady state, applications can be stopped while remaining changes are applied and restarted against the target.
- DMS Schema Conversion assesses and converts schemas and code for a different target engine; objects it cannot convert automatically require manual conversion.
- Amazon DocumentDB is a managed database service for MongoDB-compatible databases.
- Amazon Redshift is a managed data warehouse service for analyzing data.
- Amazon Redshift is designed for analytical queries over large datasets and omits some features suited to transactional processing, including efficient single-row data manipulation.
- A Multi-AZ DMS replication instance has a synchronous standby in a different Availability Zone for availability and failover.
- DMS Serverless automatically provisions and scales replication capacity and provides built-in high availability.
- CDCLatencySource measures the delay between the source and the replication instance.
- High CDCLatencyTarget indicates a delay applying change events to the target.
- FreeableMemory includes memory in buffers and cache that can be freed and is not the same as actual free memory.
- Views can be migrated with full-load-only tasks but not with tasks that include CDC.
Monitoring and troubleshooting
AWS Database Migration Service (DMS) provides endpointA DMS endpoint holds connection, type and location information for a data store. tests, task and table states, logs, and metrics to help diagnose migration problems. Use the task status to find errors, then check individual tables and metrics to narrow down the issue.
How it works
- Define source and target endpoints with the data stores’ connection, type, and location information. Test each endpoint connection using a VPC and replication instanceA DMS replication instance reads data from the source and loads it into the target..
- Check the task status and table states. A task can continue loading other tables when one or more fail. Enable task logging in the settings to send task information to CloudWatch.
- Compare replication metrics. Similar
CDCLatencySourceandCDCLatencyTargetvalues suggest checking the source first; a higher target value suggests checking the target. Also checkFreeableMemoryandSwapUsagefor signs of an overloaded replication instance.
Exam cues
- If the question says a task is running with errors, think table states and enabled task logs, because other tables can keep loading while some fail.
- If the question says
CDCLatencyTargetis higher thanCDCLatencySource, think target investigation, because the comparison points to the target side. - If the question says endpoint connectivity needs checking before migration, think endpoint tests using the replication instance and VPC.
- If the question says
FreeableMemoryis approaching zero andSwapUsageis rising, think replication instance overload.
Remember
- Table states distinguish a full loadA full load copies existing source data to a target. in progress, a completed load, and a table error.
- Task logging to CloudWatch requires logging to be enabled in task settings.
- Similar source and target CDC latencyReplication latency is the delay between changes in a source and their replication. values suggest checking the source first.
Trap: A task running with errors can still load other tables. Its status does not mean every table migrated successfully; check the table states and task logs.
Da documentação da AWS (7)
- A DMS task running with errors can continue loading other tables while one or more tables fail to migrate.
- DMS table states distinguish an in-progress full load, a completed load and a table error.
- DMS can log task information to CloudWatch when logging is enabled in the task settings.
- Comparing CDCLatencySource and CDCLatencyTarget helps locate replication latency; similar values suggest checking the source first, while a higher target value suggests checking the target.
- DMS can test endpoint connections using a VPC and replication instance.
- Source and target endpoints provide connection, data store type and location information.
- FreeableMemory approaching zero or increasing SwapUsage can indicate that the replication instance is overloaded.
Flashcards
Nada para revisar aqui agora. Escolha outros tópicos ou outro filtro.
·
Todos os 11 flashcards em lista
How do Amazon DocumentDB, Amazon Redshift and AWS DMS differ?
DocumentDB runs MongoDB-compatible databases. Redshift is a data warehouse for analytics. DMS migrates data and can replicate ongoing changes.
A team must migrate a database while replicating ongoing changes to reduce cutover downtime. Which AWS DMS task fits?
Use a full load plus CDC task to copy existing data and then apply ongoing changes.
A target already contains the initial database copy. Which AWS DMS task migrates subsequent changes?
A CDC-only task replicates ongoing changes.
A team needs automatic failover for a provisioned AWS DMS replication instance. What should it configure?
Choose Multi-AZ to provision a synchronous standby in a different Availability Zone.
A team does not want to manage replication instance capacity for AWS DMS. What feature should it consider?
DMS Serverless automatically provisions and scales replication capacity.
Does AWS DMS CDC guarantee real-time replication with a latency SLA?
No. CDC latency varies and has no SLA.
A database is moving to a different engine. How can AWS DMS help with the schema and data?
Use DMS Schema Conversion to assess and convert the schema, and DMS to migrate data. Objects that cannot be converted automatically need manual conversion.
An AWS DMS task is running with errors. Does that mean every table migrated successfully?
No. One or more tables may have failed while the task continues loading others. Check table states and task logs.
AWS DMS CDCLatencyTarget is higher than CDCLatencySource. Which side should the team investigate first?
Investigate the target side for latency.
What should a team check before starting an AWS DMS migration to confirm connectivity?
Test the source and target endpoints from the replication instance.
AWS DMS FreeableMemory approaches zero and SwapUsage rises. What might this indicate?
The replication instance may be overloaded.