DBS-C01 Monitoring and Troubleshooting — Timed 10 Questions

Question 1

An organization is using Amazon DynamoDB with on-demand capacity. They notice that the 'ThrottledRequests' metric is non-zero during a specific hour each day. The table stores session data and has a partition key of 'user_id'. The workload is uniform. What is the most likely cause of throttling?

10 minute time limit — choose an answer to begin.

Accepted Answer

The account-level read/write capacity limit is exceeded.. With on-demand capacity, throttling can occur when the account-level read/write capacity limit is exceeded, even with a uniform workload. The account-level limit (default 40,000 RCU or 40,000 WCU) applies across all on-demand tables in an AWS account. If the workload spikes during a specific hour, it may exceed this limit, causing throttling. Although per-partition limits exist (3,000 RCU or 1,000 WCU), a uniform workload means partition key distribution is even, so hot partitions are unlikely. Thus, the most likely cause is the account-level limit being reached.

Answer

The table has on-demand capacity and cannot throttle.

Answer

A global secondary index is consuming write capacity.

Answer

The partition key is not evenly distributed.

Question 2

You are managing an Amazon RDS for PostgreSQL Multi-AZ DB instance that handles a high-traffic e-commerce application. Recently, the database has been experiencing intermittent slowdowns during peak hours. You have enabled Enhanced Monitoring and Performance Insights. After reviewing the Performance Insights dashboard, you notice that the 'db.sql.queries.avg_latency' metric spikes during the slowdowns, and the top SQL queries are all simple SELECT statements on a frequently accessed 'orders' table. The table has over 10 million rows and is indexed on 'order_id', 'customer_id', and 'order_date'. The average query latency for these SELECT statements jumps from 5 ms to over 500 ms during the spikes. You also observe that the 'ReadIOPS' metric on the DB instance is consistently below the provisioned IOPS limit of the gp2 storage. The DB instance type is db.r5.large with 16 GB memory. The 'DatabaseConnections' metric shows that the number of connections is well within the max_connections limit (set to 200). However, the 'CPUCreditBalance' for the underlying EC2 instance, which is a T3 medium, drops to near zero during the spikes. The 'CPUUtilization' metric is below 50%. Which of the following is the MOST likely cause and the appropriate action to resolve the issue?

Accepted Answer

The issue is caused by a missing index on the 'orders' table. Add a composite index on the columns used in the WHERE clauses.. The issue is caused by a missing composite index on the 'orders' table. Although single-column indexes exist on 'order_id', 'customer_id', and 'order_date', the simple SELECT statements likely filter on multiple columns simultaneously, forcing the database to use a single index and then perform additional filtering in memory, resulting in high latency. Adding a composite index on the columns used in the WHERE clauses (e.g., (customer_id, order_date)) allows index-only scans and eliminates unnecessary processing. The other options are inconsistent with the symptoms: the instance is already a dedicated CPU type (db.r5.large), so CPU credit exhaustion does not apply; memory and connection metrics show no bottleneck.

Answer

The issue is due to CPU credit exhaustion on the T3 instance. Change the DB instance class from db.t3.medium to a dedicated CPU instance type such as db.r5.large.

Answer

The issue is due to a memory bottleneck causing queries to spill to disk. Increase the instance memory by moving to a db.r5.xlarge instance.

Answer

The issue is caused by a connection pool exhaustion. Increase the max_connections parameter and use connection pooling.

Question 3

A company is running a production Amazon RDS for Oracle DB instance. The database specialist is setting up monitoring to detect and troubleshoot performance issues. Which TWO metrics should be used to identify whether the database is experiencing I/O bottlenecks?

Accepted Answer

ReadLatency. The correct answers are ReadLatency (C) and WriteLatency (D). These metrics directly measure the time taken for I/O operations; high values indicate I/O bottlenecks. Option A (DatabaseConnections) measures the number of connections, not I/O performance. Option B (CPUUtilization) measures CPU usage, which can be high for many reasons, not necessarily I/O. Option E (FreeStorageSpace) indicates storage capacity, not performance. Therefore, only C and D are correct.

Answer

DatabaseConnections

Answer

CPUUtilization

Answer

FreeStorageSpace

Question 4

A company is troubleshooting a performance issue with an Amazon RDS for MariaDB instance. The CloudWatch metric 'ReadIOPS' is consistently high, but 'WriteIOPS' is low. Which TWO actions could help improve read performance?

Accepted Answer

Add a read replica to offload read queries.. (Add a read replica) offloads read queries from the primary instance, reducing read IOPS. Option D (Increase the DB instance class to one with more memory) can improve caching (e.g., InnoDB buffer pool), reducing disk reads. Option B (Increase allocated storage) may increase baseline IOPS but not directly address read performance. Option C (Use Amazon ElastiCache) is a valid caching solution but not specific to RDS and is not one of the two best actions. Option E (Enable Multi-AZ) enhances availability but does not improve read performance, as the standby replica is not used for reads.

Answer

Increase the allocated storage to improve IOPS.

Answer

Use Amazon ElastiCache to cache query results.

Answer

Enable Multi-AZ to improve read performance.

Question 5

A team is troubleshooting an Amazon Aurora MySQL cluster where the writer instance fails over unexpectedly every few days. The failover is not triggered by manual intervention or maintenance events. CloudWatch metrics show that the failover occurs when the replica lag on the reader instances spikes to over 10 seconds. What is the MOST likely cause of the failover?

Accepted Answer

The reader instances are under-provisioned and cannot keep up with the write rate from the writer. If reader instances are under-provisioned, they cannot apply changes from the writer fast enough, leading to replica lag. This replica lag can cause the system to initiate a failover if it exceeds a threshold, as the writer may be deemed unhealthy due to the inability of readers to keep up.

Option A is incorrect because over-provisioned reader instances would have ample resources and would not cause replica lag; idle connections do not trigger failovers.

Option B is incorrect because a storage subsystem bottleneck in a single Availability Zone is unlikely in Aurora, which uses shared storage across three AZs; even if it occurred, it would affect both writer and readers equally, not specifically cause writer failover.

Option C is incorrect because in Aurora, replication is at the storage layer and is not blocked by long-running transactions on the writer. Such transactions do not directly cause replica lag on readers.

Answer

The reader instances are over-provisioned, causing them to become idle and drop connections

Answer

The Aurora cluster is experiencing a storage subsystem bottleneck in a single Availability Zone

Answer

The writer instance is running a long-running transaction that blocks replication

Question 6

A financial services company runs a critical application on Amazon DynamoDB. Recently, they observed increased latency and throttled requests. Upon reviewing CloudWatch metrics, they see that WriteCapacityUnits consumed is consistently below the provisioned capacity, but ReadCapacityUnits consumed is frequently at 100%. The application performs equal mix of strongly consistent reads and writes. What is the most likely cause of the throttling?

Accepted Answer

Hot partitions causing uneven read traffic distribution.. Throttling on reads with provisioned capacity still available indicates hot partitions. Option B is wrong because adaptive capacity helps but does not eliminate hot partitions. Option C is wrong because writes are not throttled. Option D is wrong because strongly consistent reads consume twice the read capacity units, but the issue is hot partitions.

Answer

Adaptive capacity is automatically adjusting partition throughput.

Answer

Write capacity is insufficient for the workload.

Answer

Strongly consistent reads are consuming twice the RCU.

Question 7

Refer to the exhibit. An IAM policy attached to an IAM role used by an AWS Lambda function. The Lambda function is invoked by Amazon API Gateway and reads from a DynamoDB table. The function is experiencing permission errors when calling DynamoDB GetItem. What is the most likely cause?

Accepted Answer

The policy does not grant access to DynamoDB indexes if the function uses a query on a Global Secondary Index.. The IAM policy in the exhibit grants `dynamodb:GetItem` and `dynamodb:Query` actions on the table ARN only. When a Lambda function performs a Query operation on a Global Secondary Index (GSI), it must have permission on the index ARN (e.g., `arn:aws:dynamodb:region:account:table/table-name/index/index-name`) in addition to the table ARN. Without the index ARN in the policy's `Resource` element, DynamoDB will deny the request, resulting in permission errors. Therefore, the most likely cause is that the policy does not grant access to DynamoDB indexes, matching option B.

Answer

The IAM role trust policy does not allow Lambda to assume the role.

Answer

The dynamodb:GetItem action is not listed in the policy.

Answer

The Lambda function ARN in the resource field is incorrect.

Question 8

Which THREE metrics should be monitored to troubleshoot a slow Amazon Redshift query? (Choose three.)

Accepted Answer

DiskSpaceUsage. Options B, C, and D are correct. DiskSpaceUsage can indicate disk spills that slow queries, WLMQueueWaitTime measures time queries wait in a queue, and QueryDuration directly measures execution time. Option A (DatabaseConnections) tracks connection count, not query performance. Option E (NetworkThroughput) measures network throughput, which is not directly related to query slowness.

Answer

DatabaseConnections

Answer

NetworkThroughput

Question 9

A company is using Amazon RDS for MySQL with Multi-AZ deployment. The application team reports that write latency has increased significantly over the past hour. CloudWatch shows elevated 'ReadLatency' and 'WriteLatency' metrics. The DB instance is a db.r5.large with 500 GB General Purpose SSD (gp2) storage. Which action is most likely to resolve the issue?

Accepted Answer

Increase the allocated storage size to 1,000 GB. Elevated latency with gp2 storage often indicates exhausted I/O credits, and increasing storage size raises the baseline IOPS. Option A is incorrect because Multi-AZ is already deployed. Option B is incorrect because read replicas do not reduce write latency. Option D is incorrect because instance class scaling does not directly affect storage IOPS.

Answer

Enable Multi-AZ deployment

Answer

Create a read replica to offload read traffic

Answer

Scale up the DB instance class to db.r5.xlarge

Question 10

A database administrator is troubleshooting a slow-running query on an Amazon RDS for MySQL DB instance. Which CloudWatch metric should be monitored to determine if the query is experiencing storage I/O bottlenecks?

Accepted Answer

ReadLatency and WriteLatency. ReadLatency and WriteLatency indicate I/O performance. Option B is wrong because CPUUtilization measures processor usage. Option C is wrong because NetworkThroughput measures network traffic. Option D is wrong because DatabaseConnections measures connections.

Answer

CPUUtilization

Answer

NetworkThroughput

Answer

DatabaseConnections