DBS-C01 Monitoring and Troubleshooting Practice Test 14 — 15 Questions

Question 1

A database specialist is troubleshooting an Amazon DynamoDB table that is experiencing elevated latency for read operations. The table uses on-demand capacity mode and has a global secondary index (GSI) with a partition key that is different from the base table. The specialist notices that some read requests are throttled and that the GSI's consumed read capacity is much higher than expected. Which two actions should the specialist take to diagnose and mitigate the issue? (Choose two.)

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Examine the ConsumedReadCapacityUnits and ThrottledRequests metrics for the GSI in Amazon CloudWatch.. The specialist should check CloudWatch metrics specific to the GSI, such as ConsumedReadCapacityUnits and ThrottledRequests, to confirm and quantify the issue. Additionally, reviewing the GSI's partition key design for hot partitions is crucial because uneven data distribution can cause throttling and high capacity consumption. These two actions together diagnose the problem and suggest a mitigation path.

Answer

Enable DynamoDB Streams on the table to capture the read requests that are being throttled.

Answer

Increase the provisioned read capacity of the base table to reduce throttling on the GSI.

Answer

Enable strong consistency on the GSI to ensure that reads are not throttled due to eventual consistency.

Question 2

A database specialist is managing an Amazon DocumentDB cluster. The application team reports that queries are returning stale data. The specialist suspects that the application is reading from a replica that has replication lag. The specialist needs to confirm whether replication lag is occurring. Which metric should the specialist monitor in Amazon CloudWatch?

Accepted Answer

DBClusterReplicaLagMaximum. Amazon DocumentDB provides specific CloudWatch metrics for replication lag. DBClusterReplicaLagMaximum reports the maximum lag across all replicas in the cluster, which is ideal for identifying if any replica is behind and causing stale reads. Other metrics like DBInstanceReplicaLag also exist but are per-instance. DBClusterReplicaLagMaximum gives a cluster-wide view, making it the best choice to confirm replication lag.

Answer

DBClusterReplicaLag

Answer

ReplicaLag

Answer

DBInstanceReplicaLag

Question 3

A company has an Amazon Aurora MySQL cluster with a writer and a reader instance. The application connects to the reader endpoint for read-only queries. Users report that read queries sometimes fail with 'ERROR 1290 (HY000): The MySQL server is running with the --read-only option so it cannot execute this statement'. What is the most likely cause?

Accepted Answer

The application is issuing write statements to the reader endpoint.. The error message indicates that the MySQL server is running with the --read-only option, which is the default for Aurora reader instances. The application is likely sending write statements to the reader endpoint, which is intended for read-only queries. To resolve, the application should route write operations to the writer endpoint. This is a common mistake when applications do not separate read and write connections.

Answer

The reader instance is undergoing a reboot or is in the process of being promoted to a writer.

Answer

The application is connecting to the writer endpoint instead of the reader endpoint.

Answer

The Aurora cluster's reader endpoint is incorrectly configured to point to the writer instance.

Question 4

A company runs an Amazon Aurora MySQL DB cluster with one writer and three readers. The application experiences intermittent connection timeouts to the reader instances. A database specialist needs to determine whether the issue is due to the reader instances being overloaded or due to network connectivity. Which two actions should the specialist take to gather the necessary information? (Choose two.)

Accepted Answer

Enable and review Aurora MySQL error logs for each reader instance to look for connection errors or aborted connections.. To diagnose intermittent connection timeouts to Aurora readers, the specialist should check both database-level connection metrics and error logs. The DatabaseConnections metric indicates if readers are nearing connection limits, while error logs reveal server-side connection errors. Together, they help differentiate between overload and network problems. The other actions are either less direct or remediation steps rather than diagnostic data collection.

Answer

Use Amazon CloudWatch Logs Insights to analyze VPC Flow Logs for rejected traffic to the reader instances.

Answer

Run a traceroute from the application servers to the reader endpoints to measure network latency and packet loss.

Answer

Modify the reader instances to use a larger instance class to increase max_connections.

Question 5

A database specialist is troubleshooting an Amazon DynamoDB table that is experiencing throttling on read operations. The table uses provisioned capacity mode with auto scaling enabled. The specialist needs to identify the cause of the throttling. (Choose two.)

Accepted Answer

Check the 'ConsumedReadCapacityUnits' metric to see if it consistently exceeds the provisioned capacity.. ReadThrottleEvents directly counts read requests that exceed provisioned capacity, confirming read throttling. ConsumedReadCapacityUnits shows if the table is consistently at its capacity limit, indicating the need for more capacity. Together, these metrics pinpoint the cause. Other metrics like SuccessfulRequestLatency or ThrottledRequests are less specific, and auto scaling target utilization is a configuration detail, not a direct cause.

Answer

Verify that the table's auto scaling policy has a target utilization set to 100%.

Answer

Review the 'SuccessfulRequestLatency' metric for read operations to identify increased latency.

Answer

Analyze the 'ThrottledRequests' metric to see if it includes read operations.

Question 6

A database administrator is monitoring an Amazon RDS for SQL Server DB instance using Enhanced Monitoring. The administrator notices that the 'WriteIOPS' metric is consistently high, but the 'DiskQueueDepth' metric is low. The DB instance is using General Purpose SSD (gp2) storage. What is the most likely cause of the high WriteIOPS?

Accepted Answer

The DB instance is experiencing a high volume of write operations that are being served from the storage cache.. High WriteIOPS with low DiskQueueDepth on a gp2 volume suggests that the storage is handling the write load efficiently, likely due to gp2 burst capacity. gp2 volumes burst above baseline IOPS for short periods, and during bursts, queue depth remains low because the volume can serve the IOPS. This is a normal performance characteristic, not an indication of throttling or memory issues.

Answer

The DB instance is experiencing storage throttling due to exceeding the baseline IOPS of the gp2 volume.

Answer

The DB instance is using a read replica to offload read traffic, causing increased writes on the primary.

Answer

The DB instance has insufficient memory, causing frequent page writes to disk.

Question 7

A database specialist manages an Amazon Aurora MySQL cluster with a writer and two readers. Users report that analytical queries are occasionally returning stale data, and the specialist suspects that some read traffic is being served by the writer instance instead of the readers. The cluster endpoint is being used by the application for all connections. Which action should the specialist take to ensure that read-only queries are always directed to the reader instances?

Accepted Answer

Create a custom endpoint that includes only the reader instances and update the application to use it for read queries.. The issue is that the application uses the cluster endpoint, which always routes to the writer. To ensure reads go to readers, the application must use an endpoint that includes only reader instances. A custom endpoint allows selecting specific readers, providing control and avoiding the writer. The reader endpoint also works, but a custom endpoint is more precise if the goal is to exclude certain readers or include only specific ones.

Answer

Enable Aurora Auto Scaling on the reader instances and set the application to use the reader endpoint.

Answer

Configure the writer instance to accept read-only queries by setting the read_only parameter to 1.

Answer

Modify the application to connect to the cluster endpoint for reads and the reader endpoint for writes.

Question 8

A healthcare company uses Amazon ElastiCache for Redis (cluster mode disabled) with one primary and two replicas. The application reports intermittent latency spikes, and CloudWatch shows `ReplicationLag` on one replica growing to several seconds while the primary's `CurrConnections` and `EngineCPUUtilization` remain normal. The replica's `SaveInProgress` metric is frequently 1. Which action should the database specialist take to reduce replication lag?

Accepted Answer

Scale the affected replica to a larger node type with more memory and faster disk, and schedule snapshots during low-traffic periods.. The replica's frequent `SaveInProgress` combined with rising `ReplicationLag` points to background snapshotting consuming CPU and disk on the replica, slowing replication. Scaling the replica to a larger node and moving snapshots to low-traffic windows reduces that contention. This addresses the root cause without removing backups or changing the cluster topology.

Answer

Increase the number of replicas from two to three to distribute the replication load.

Answer

Disable automatic backups on the cluster to eliminate the background save that is delaying replication.

Answer

Enable cluster mode on the existing cluster to shard the data and reduce replication overhead per node.

Question 9

A company runs an Amazon RDS for PostgreSQL DB instance with a primary and a read replica. The read replica is used for reporting queries. A database specialist notices that the read replica is experiencing high replication lag, and the ReplicaLag metric in CloudWatch is increasing. The primary instance shows normal CPU and write throughput. Which action should the specialist take first to resolve the replication lag?

Accepted Answer

Identify and terminate long-running queries on the read replica that are blocking the replication apply process.. Replication lag on a PostgreSQL read replica can be caused by conflicts between read queries and the replication apply process. Long-running reporting queries can hold locks that block WAL application, causing the replica to fall behind. Terminating such queries allows replication to proceed. This is a common first step because it is non-disruptive to the primary and can quickly resolve lag without scaling resources.

Answer

Increase the size of the read replica instance to a larger instance class.

Answer

Enable automated backups on the read replica to reduce replication lag.

Answer

Modify the primary instance to increase the max_wal_senders parameter.

Question 10

A database specialist manages an Amazon Aurora MySQL cluster with a writer and two readers. Application logs show intermittent errors: 'ERROR 1040 (HY000): Too many connections' when connecting to the reader endpoint. Amazon CloudWatch metrics show DatabaseConnections on each reader fluctuating between 80 and 120. The application uses a connection pool with a maximum of 50 connections per instance. What is the most likely cause of the connection errors?

Accepted Answer

The Aurora MySQL cluster's max_connections parameter is set too low for the total number of connections from all application instances.. The error 'Too many connections' is directly related to the max_connections parameter on the Aurora MySQL instances. Even with a connection pool limit per application instance, the aggregate number of connections from multiple application instances can exceed the database's max_connections. Increasing max_connections or reducing the pool size per instance resolves the issue.

Answer

The reader endpoint is not properly configured to route connections to the readers, causing connections to be sent to the writer instance.

Answer

The application's connection pool is not closing idle connections, causing a buildup of connections on the readers.

Answer

The Aurora reader endpoint is distributing connections unevenly because the application's connection pool is not using the reader endpoint correctly.

Question 11

A database specialist is analyzing an Amazon Aurora MySQL cluster that has a writer and one reader. The application connects to the reader endpoint for read-only queries. The specialist observes that the 'AuroraReplicaLag' metric for the reader is consistently around 100 milliseconds. The application requires that read queries return data that is at most 50 milliseconds stale. Which action should the specialist take to meet the staleness requirement?

Accepted Answer

Configure the application to use the writer endpoint for all read queries.. Aurora MySQL reader instances replicate asynchronously, and the observed lag of 100 milliseconds exceeds the application's 50-millisecond staleness requirement. To guarantee that reads are within 50 milliseconds, the application must read from the writer instance, which has no replication lag. Other options do not reduce replication lag or are not applicable to Aurora MySQL.

Answer

Enable the 'aurora_read_replica_read_committed' parameter on the reader instance.

Answer

Increase the size of the reader instance to reduce replication lag.

Answer

Enable the 'hot_standby_feedback' parameter on the writer instance.

Question 12

A company uses Amazon DynamoDB with a table that has a provisioned read capacity of 100 RCUs and write capacity of 50 WCUs. The application performs frequent small reads and writes, but during peak hours, the table experiences throttling. CloudWatch metrics show that consumed read capacity is consistently near 100 RCUs, but consumed write capacity is only around 20 WCUs. The database specialist needs to reduce throttling without over-provisioning. Which action is MOST cost-effective?

Accepted Answer

Enable DynamoDB Auto Scaling for both read and write capacity with a target utilization of 70%.. The table is throttled on reads while writes are underutilized. DynamoDB Auto Scaling automatically adjusts read and write capacity based on utilization, targeting a specified percentage. With a 70% target, it will scale reads up to handle peaks and scale writes down, optimizing cost. This is more cost-effective than manual over-provisioning or switching to on-demand, which could be pricier for this predictable workload.

Answer

Increase the write capacity to 100 WCUs to balance the capacity.

Answer

Switch the table to on-demand capacity mode.

Answer

Increase the provisioned read capacity to 200 RCUs and keep write capacity at 50 WCUs.

Question 13

A database specialist is responsible for an Amazon ElastiCache for Redis cluster that is used as a session store for a web application. The specialist receives reports that users are being logged out unexpectedly, and the application logs show errors indicating that session data is missing. The Redis cluster is configured with a single shard and one replica. The specialist checks the ElastiCache console and notices that the cluster's evictions metric is high. What is the most likely cause of the missing session data?

Accepted Answer

The Redis cluster's maxmemory-policy is set to allkeys-lru, causing session keys to be evicted when memory is full.. The high evictions metric indicates that Redis is evicting keys due to memory pressure. The default maxmemory-policy for ElastiCache for Redis is volatile-lru, but if it is set to allkeys-lru, any key can be evicted, including session keys without expiration. For session stores, it is recommended to use volatile-lru with an expiration time on session keys, or to use a dedicated Redis cluster with sufficient memory and no eviction for session data. Monitoring evictions and memory usage is crucial.

Answer

The Redis cluster's maxmemory-policy is set to volatile-lru, causing only keys with an expiration time to be evicted.

Answer

The Redis cluster's replica is failing over, causing temporary data loss during the failover process.

Answer

The Redis cluster's maxmemory-policy is set to noeviction, causing write operations to fail when memory is full.

Question 14

A database specialist is monitoring an Amazon Aurora PostgreSQL DB cluster. The cluster experiences a sudden increase in write latency. The specialist needs to identify the cause. Which two actions should the specialist take to troubleshoot the issue? (Choose two.)

Accepted Answer

Check the Amazon CloudWatch metric for DiskQueueDepth on the writer instance.. To troubleshoot write latency in Aurora PostgreSQL, checking DiskQueueDepth in CloudWatch reveals if the storage is saturated, and reviewing Performance Insights wait events identifies specific I/O bottlenecks. These two actions provide concrete data on the cause. Enhanced Monitoring and ReplicaLag are less directly relevant, and scaling is a remediation, not diagnosis.

Answer

Enable Enhanced Monitoring to get OS-level metrics.

Answer

Check the ReplicaLag metric for Aurora Replicas.

Answer

Increase the instance size of the writer instance.

Question 15

A database specialist is investigating intermittent connection timeouts to an Amazon Aurora MySQL DB cluster. The application logs show 'Too many connections' errors. The specialist checks Amazon CloudWatch metrics and sees that the DatabaseConnections metric frequently hits the max_connections limit. The application uses a connection pool with a maximum of 100 connections per instance, and there are 5 application instances. The Aurora cluster has one writer and two readers. What is the most likely cause of the connection timeouts?

Accepted Answer

The application is not using the reader endpoint for read-only queries, causing all connections to be directed to the writer instance.. The 'Too many connections' error occurs when the number of active connections exceeds the max_connections limit. In an Aurora MySQL cluster, the writer and reader instances each have their own max_connections setting. If the application uses the cluster endpoint for all connections, including read-only queries, all connections go to the writer, potentially exhausting its connection limit. By using the reader endpoint for read queries, connections are distributed across reader instances, reducing the load on the writer and preventing the error.

Answer

The Aurora cluster's reader instances are not properly configured to accept connections from the application's security group.

Answer

The Aurora cluster's max_connections parameter is set too low for the total number of connections from all application instances.

Answer

The application's connection pool is not properly closing connections, leading to connection leaks.