AWS Certified Database Specialty DBS-C01 (DBS-C01) — Questions 751–825

Option C is correct because ReplicationLatency measures the time between the last update on the source table and the last update on the replica table. Option A is wrong because ThrottledRequests indicates throttling. Option B is wrong because ConsumedWriteCapacityUnits measures usage.

Option D is wrong because SuccessfulRequestLatency measures request latency.

Options A, D, and E are correct. Enabling DynamoDB Accelerator (DAX) reduces read load. Implementing exponential backoff helps retry gracefully.

Using DynamoDB global tables can distribute write load. Option B is wrong because disabling auto scaling would not help. Option C is wrong because switching to on-demand capacity might help but is not always cost-effective and does not address hot keys.

Option A (DMLThroughput) measures the rate of write operations. Option D (CommitLatency) measures the time to commit transactions, which directly affects write latency. Option B is incorrect because it is a read metric.

Option C is incorrect because it measures storage. Option E is incorrect because it is a read latency metric.

Options B and D are correct. A high disk queue depth indicates I/O bottleneck. Increasing IOPS (Option B) or enabling Auto Scaling (Option D) can alleviate the bottleneck.

Option A is wrong because more connections may increase I/O contention. Option C is wrong because query caching is not a primary solution for I/O. Option E is wrong because Multi-AZ does not improve I/O performance.

DynamoDB is a key-value and document database optimized for high-velocity writes and low-latency point lookups. With a primary key of device_id and a sort key of timestamp, the query for the most recent reading can use ScanIndexForward: false and limit 1, which is efficient. RDS and Aurora are relational and not ideal for high-throughput writes.

ElastiCache is a cache, not a durable database. Redshift is for analytics.

Option B is correct because increasing the instance size provides more resources for replication. Option A is wrong because pausing replication would increase lag. Option C is wrong because switching to full load would cause downtime.

Option D is wrong because disabling logging would break CDC.

Options B and D are correct. Option B: Pre-creating a global table can distribute write traffic across regions, reducing load on a single table. Option D: Using DynamoDB Accelerator (DAX) can offload read traffic, reducing read capacity consumption.

Option A is wrong because switching to provisioned capacity requires capacity planning but does not automatically prevent throttling. Option C is wrong because increasing read capacity units is not applicable to on-demand mode. Option E is wrong because decreasing write capacity is counterproductive.

Options A and C are correct because CloudWatch Logs can monitor error logs in real time and set metric filters for specific error patterns. Option B (S3 export) is for archiving, not real-time monitoring. Option D (CloudTrail) is for API activity, not database logs.

Option E (X-Ray) is for tracing, not log monitoring.

Option A is correct for schema conversion. Option B is correct for ongoing replication. Option C is correct for data migration.

Option D is incorrect because it is for file transfer. Option E is incorrect because it is for large data transfer.

When the create-db-instance command returns, the DB instance state is 'creating'. It takes several minutes for the instance to become 'available'. Option A (available) is incorrect because creation is not instant.

Option C (stopped) is incorrect. Option D (deleting) is incorrect.

DMS can migrate BFILEs if the source endpoint is configured with the directory path. Using SCT to convert any incompatible schema is recommended. Uploading BFILEs to S3 and using Oracle Directory objects with DMS can handle the migration.

Option B (direct update) is not secure. Option E (CloudEndure) is for servers.

Option C is correct. The latest restorable time is 2023-03-15T04:30:00Z, which is within the backup window. Option A is wrong because it is after the latest restorable time.

Option B is wrong because it is earlier than the latest restorable time. Option D is wrong because the backup window ends at 06:00, but the latest restorable time is 04:30.

Option A is correct because RDS encryption at rest can only be enabled when creating a new DB instance, not on an existing one. Option B is wrong because enabling encryption requires creating an encrypted snapshot and restoring as a new instance, not just modifying. Option C is wrong because encryption cannot be enabled on an existing instance.

Option D is wrong because encryption is configured at instance creation, not through parameter groups.

Options A, B, and D are correct. A: SCT converts the schema. B: DMS with CDC provides minimal downtime.

D: Use SCT to assess and convert Oracle-specific features. Option C (modify application to use Aurora-specific features) is not necessary. Option E (Snowball) is not for minimal downtime.

Option B is correct. Amazon Timestream is optimized for time-series data, supports high write throughput, and provides built-in aggregation functions for time-based queries. Option A (DynamoDB) is not ideal for time-series aggregations without additional processing.

Option C (RDS for PostgreSQL) can handle the workload but may require more management and scaling effort. Option D (ElastiCache) is a cache, not a durable database.

Option B is correct because KMS automatic key rotation rotates the key material annually. Option A is wrong because manual rotation requires creating a new key and updating the RDS instance. Option C is wrong because CloudHSM is not used with RDS.

Option D is wrong because the option group does not control key rotation.

Option B is correct because AWS DMS can replicate ongoing changes after an initial full load, minimizing downtime. Option A is wrong because mysqldump would be slow and cause more downtime. Option C is wrong because S3 is not directly used for live migration.

Option D is wrong because RDS does not support direct restore of physical backups from on-premises.

Enabling RDS Enhanced Monitoring does not capture query logs. Enabling automatic backups does not log queries. Enabling RDS Performance Insights does not log queries.

To capture SELECT queries, you need to enable PostgreSQL query logging by setting the appropriate parameter group parameters (e.g., log_statement = 'all' or 'mod') and then export logs to CloudWatch Logs.

Option B (AWS DMS with ongoing replication) is correct because it supports minimal downtime by continuously replicating changes. Option A (native pg_dump/pg_restore) requires downtime. Option C (AWS Database Migration Service with full load only) does not replicate ongoing changes.

Option D (SCT with Snowball) is for large-scale schema conversion, not minimal-downtime migration.

Option A is correct because using native Cassandra replication allows seamless migration with zero downtime by adding Keyspaces as a new datacenter. Option B is wrong because DMS does not support Cassandra as a source for Keyspaces. Option C is wrong because CQL COPY is not suitable for high throughput writes and requires downtime.

Option D is wrong because S3 export/import is a batch process, not real-time.

Option D is correct because setting up replication from EC2 to RDS using MySQL native replication allows near-zero downtime cutover. Option A is wrong because dump/restore requires downtime. Option B is wrong because snapshot restore requires downtime.

Option C is wrong because DMS might add complexity and is not necessary if native replication works.

Multi-AZ deployment with a standby replica provides automatic failover in RDS MySQL. Option C is correct. Option A (Read Replica) does not provide automatic failover.

Option B (Single-AZ) lacks high availability. Option D (Multi-AZ with synchronous replication) is correct but the description is more accurate for Aurora; for RDS MySQL, Multi-AZ uses synchronous replication to a standby.

RDS does not allow changing the encryption key of an existing encrypted DB instance. The only way is to take a snapshot, copy it with a new KMS key, and restore from that snapshot. Option A is wrong because you cannot modify the KMS key directly.

Option B is wrong because restoring to a point in time uses the same key. Option D is wrong because enabling encryption on an unencrypted instance requires a snapshot restore, but this instance is already encrypted.

Option B is correct. To enable encryption on an existing unencrypted Redshift cluster, you must unload the data to Amazon S3, create a new encrypted cluster, and reload the data. Option A is incorrect because Redshift does not support enabling encryption on an existing cluster directly.

Option C is incorrect because snapshots preserve encryption status; restoring an unencrypted snapshot to an encrypted cluster is not possible. Option D is incorrect because modifying the cluster does not support enabling encryption.

Option B is correct because AWS DMS supports SSL encryption and CDC for minimal downtime. Option A is wrong because S3 does not provide database migration. Option C is wrong because EC2 instance is not needed.

Option D is wrong because VPN alone does not migrate data.

Option C is correct because connection pooling can accumulate connections if not properly configured, leading to exhaustion. Option A is wrong because 1000 is the default for db.r5.large. Option B is wrong because network ACLs affect connectivity, not the number of connections once established.

Option D is wrong because the error indicates connections are being accepted but exhausted.

Option B is correct because a materialized view can pre-aggregate the daily sales, and refreshing it daily avoids scanning the full table. Option A is wrong because indexes on date columns help but still require scanning large portions. Option C is wrong because partitioning by month still requires scanning the partition.

Option D is wrong because upgrading to a larger instance is costly and may not achieve sub-30 seconds.

Option B is correct because VACUUM reclaims space from deleted rows without adding cost. Option A is wrong because adding nodes increases cost. Option C is wrong because increasing slice count is not possible without changing node type.

Option D is wrong because compression reduces storage but may not immediately free space.

Option B is correct because enabling Enhanced Monitoring at the instance level provides OS-level metrics that can help identify resource bottlenecks causing high write latency. Option A is wrong because profiling queries in the slow query log may show slow queries but not necessarily latency at the OS level. Option C is wrong because creating additional replicas does not reduce write latency on the primary.

Option D is wrong because switching to Amazon DynamoDB is a major architectural change, not a troubleshooting step.

Option C is correct because manual remediation of unconverted code is often necessary, and SCT provides an assessment report to guide this. Option A is wrong because DMS does not convert stored procedures. Option B is wrong because PostgreSQL has its own procedural languages.

Option D is wrong because Lambda is not designed for schema conversion.

Option C is correct because Amazon DynamoDB offers flexible schema, high throughput, and optional strong consistency. Option A (RDS MySQL) is relational with fixed schema. Option B (Neptune) is graph-dedicated.

Option D (DocumentDB) is MongoDB-compatible but not as optimized for consistent low-latency reads across a wide range of access patterns.

Option D is correct because using a write sharding pattern with a suffix distributes writes across partitions, avoiding hot spots. Option A is wrong because increasing read capacity does not help write throttling. Option B is wrong because DynamoDB Accelerator (DAX) only caches reads.

Option C is wrong because global tables replicate data, not reduce throttling.

Option A (increasing the number of nodes) and Option C (distributing data evenly) are correct. Option B is incorrect because Sort keys affect order, not disk spills. Option D is incorrect because WLM queues affect concurrency, not disk operations.

Option E is incorrect because column compression reduces storage but may not reduce disk spills.

Option C is correct because a high volume of write operations can cause the primary instance to become overloaded, leading to failover. Option A is incorrect because DatabaseConnections alone does not directly cause failover. Option B is incorrect because ReadLatency is a symptom, not a direct cause of failover.

Option D is incorrect because FreeableMemory is not a typical cause of failover.

Option B is correct because RDS for SQL Server supports Transparent Data Encryption (TDE) which can use CloudHSM as the key store. Multi-AZ deployments support cross-region read replicas, allowing DR. Option A is wrong because RDS encryption at rest cannot use CloudHSM directly.

Option C is wrong because moving to DynamoDB is unnecessary and does not use CloudHSM. Option D is wrong because using an RDS Custom instance with TDE and manual replication is more complex and not recommended.

Option C is correct because DynamoDB supports updating the encryption key for an existing table using the UpdateTable API with the SSESpecification parameter. You can switch from AWS managed key to a customer managed key. Option A is incorrect because you cannot change the encryption configuration during table creation after the table exists.

Option B is incorrect because exporting to S3 and re-importing is unnecessary and causes more downtime. Option D is incorrect because DynamoDB streams are used for change data capture, not for encryption changes.

Option B is correct because AWS DMS supports ongoing replication from MongoDB to DocumentDB using change streams. Option A is wrong because mongodump/mongorestore is a one-time import with downtime. Option C is wrong because AWS DataSync is for file-based transfers, not databases.

Option D is wrong because AWS Glue is for ETL, not live replication.

Amazon DynamoDB is a NoSQL key-value and document database that offers flexible schema and high throughput with low latency. Option A (Aurora) is relational. Option B (RDS MySQL) is relational.

Option D (ElastiCache) is a cache, not a primary database.

Option C is correct because AWS Database Migration Service (DMS) with ongoing replication allows migration with minimal downtime by continuously replicating changes from the source to the target until cutover. Option A is wrong because creating a read replica is not possible from on-premises. Option B is wrong because exporting and importing will cause significant downtime.

Option D is wrong because changing the master user is unrelated to migration.

Option C is correct because the RDS security group should allow inbound traffic from the EC2 security group. Option A is wrong because NACLs are stateless and less granular. Option B is wrong because RDS does not support EC2 classic security groups.

Option D is wrong because IAM roles are for authentication, not network access.

Using IAM policies to restrict access is a security best practice. Encrypting the table at rest with a KMS key protects data. Monitoring with CloudTrail provides audit trail.

VPC endpoints help but are not a security practice for the table itself. Encryption in transit is done by DynamoDB automatically via HTTPS. Fine-grained access control can be achieved with IAM conditions, not attribute-based access control on the table itself.

Option A is correct because scaling up the instance provides more CPU capacity. Option B is wrong because read replicas help with read scaling but not CPU on the writer. Option C is wrong because gp3 storage does not affect CPU.

Option D is wrong because changing the query cache is unlikely to significantly reduce CPU.

AWS DMS with ongoing replication allows for minimal downtime. Option A is incorrect because backup/restore does not support ongoing replication. Option B is incorrect because S3 and Glue are not designed for database migration.

Option C is incorrect because AWS SCT is for schema conversion, not data migration.

Option C is correct because CloudTrail can record data-plane events for DynamoDB (e.g., GetItem, PutItem) when data events are enabled. Option A is wrong because CloudWatch Logs can store logs but does not capture API calls directly. Option B is wrong because VPC Flow Logs capture network traffic, not API calls.

Option D is wrong because Config records configuration changes, not data-plane API calls.

Option A is correct. The policy allows the user to connect to the specified database as user 'db_user1'. Option B is wrong because it only allows connecting to one database user.

Option C is wrong because it allows only one database user. Option D is wrong because the resource is specific to a database instance and user.

The ORA-1555 error indicates that undo data is overwritten before DMS can read it. Increasing undo retention ensures the snapshot data is available longer. Reducing batch size would not address the undo issue.

Increasing logging frequency does not affect undo. Using a different connection method does not solve the underlying undo retention problem.

Amazon RDS events are stored in the RDS console under 'Events' and also in CloudTrail. But the best place to see failover reasons is the RDS console events. Option B is correct because the RDS console shows events like 'failover' with reasons.

Option A is wrong because CloudTrail records API calls, not internal failover reasons. Option C is wrong because CloudWatch Logs does not automatically log failover reasons. Option D is wrong because StatusInsis not a service.

Amazon Timestream is a time-series database optimized for IoT data, with fast ingestion and built-in analytics functions. DynamoDB is not optimized for time-series queries. RDS is not designed for high-velocity time-series data.

Redshift is for batch analytics.

Option A is correct because changing the isolation level to READ COMMITTED eliminates gap locks for non-unique indexes, reducing lock contention. Option B is wrong because increasing innodb_lock_wait_timeout only increases timeout, not reduces locks. Option C is wrong because adding an index on transaction_date may increase lock granularity? Actually, if the index is not unique, gap locks still occur.

Option D is wrong because partitioning does not affect gap locks.

Option C is correct. The error indicates the flash recovery area is full. The most efficient automated way is to modify the RDS DB instance to increase allocated storage, which also increases the recovery area.

Option A is wrong because manually deleting logs is not automated. Option B is wrong because archiving to S3 is not a standard RDS feature. Option D is wrong because restarting does not free space.

Options A and D are correct because you can create a snapshot, copy it with encryption, and restore; or use DMS to migrate to an encrypted target. Option B is wrong because you cannot directly enable encryption on an existing instance. Option C is wrong because enabling encryption on a replica requires the source to be encrypted.

Option E is wrong because the console does not allow direct modification.

Option B is correct because enabling slow query log helps identify long-running queries causing high CPU and connections. Option A is wrong because Multi-AZ failover does not resolve performance issues. Option C is wrong because enhancing CloudWatch metrics does not directly fix timeouts.

Option D is wrong because read replicas are for read scaling, not for reducing connection timeouts on the primary.

DynamoDB encryption at rest using AWS managed key does not support customer-controlled rotation. Option C is correct. Option A is wrong because AWS managed keys rotate automatically every year, not on customer schedule.

Option B is wrong because KMS automatic rotation is annual and cannot be set to 90 days. Option D is wrong because disabling encryption is not a valid option.