Troubleshooting Scenario Questions

Enable auto scaling on the table to automatically adjust capacity.

Why wrong: Auto scaling is for provisioned capacity mode, not on-demand.

Enable DynamoDB Accelerator (DAX) to reduce read load on the table.

Why wrong: DAX is a caching layer but does not resolve throttling due to hot partitions.

Switch the table to provisioned capacity mode and increase the read capacity units.

Why wrong: Switching to provisioned capacity does not address the underlying hot partition issue.

Run the Redis INFO command to see memory usage

Why wrong: Memory usage may be relevant, but cache hit ratio is the primary metric.

Review the ElastiCache event logs for node failures

Why wrong: Node failures would cause complete cache misses, not just reduced hit ratio.

Check the Redis AOF persistence logs

Why wrong: Persistence logs are for durability, not cache performance.

Monitor the 'FreeStorageSpace' metric in percent and set a threshold of 10.

Why wrong: There is no percent metric for FreeStorageSpace; it is in bytes.

Monitor the 'DiskSpaceUtilization' metric and set a threshold of 90.

Why wrong: There is no 'DiskSpaceUtilization' metric for RDS; use 'FreeStorageSpace'.

Monitor the 'FreeStorageSpace' metric in gigabytes and set a threshold of 50.

Why wrong: CloudWatch metrics for RDS FreeStorageSpace are in bytes, not gigabytes.

Amazon RDS Enhanced Monitoring

Why wrong: Incorrect. Enhanced Monitoring provides OS-level metrics, not query performance insights.

AWS CloudTrail

Why wrong: Incorrect. CloudTrail tracks API calls, not database performance.

Amazon CloudWatch Logs

Why wrong: Incorrect. CloudWatch Logs are for log data, not database query analysis.

Modify the DB instance to reduce the backup retention period to 0 days.

Why wrong: Reducing backup retention to 0 days will delete all automated backups from S3, but this does not free up DB instance storage. Additionally, it disables automated backups, losing backup capability.

Delete manual snapshots from the RDS console.

Why wrong: Manual snapshots are stored in Amazon S3, not in the instance storage, so deleting them has no effect.

Disable automated backups to stop storage consumption.

Why wrong: Disabling automated backups stops future backups and may delete existing ones (from S3), but does not free instance storage.

Increase the provisioned read capacity units to handle the load.

Why wrong: Throttling is on writes, not reads; increasing read capacity does not help.

Switch the table to provisioned capacity mode and increase write capacity.

Why wrong: On-demand already auto-scales; the issue is hot partitions, not overall capacity.

Enable DynamoDB Accelerator (DAX) to cache write operations.

Why wrong: DAX caches reads, not writes; it does not reduce write throttling.

First, check network ACLs. Second, verify security group inbound rules. Third, test connectivity using telnet. Finally, verify EC2 network configuration.

Why wrong: This order is incorrect because network ACLs are evaluated after security groups, and testing before verifying network configuration may miss routing issues.

First, test connectivity using telnet. Second, verify security group inbound rules. Third, check network ACLs. Finally, verify EC2 network configuration.

Why wrong: This order is incorrect because testing without prior verification of security groups and ACLs may give misleading results and wastes time on troubleshooting without context.

First, verify EC2 network configuration. Second, verify security group inbound rules. Third, check network ACLs. Finally, test connectivity using telnet.

Why wrong: This order is incorrect because starting with EC2 network configuration is premature; security groups and NACLs are more common causes and should be checked first.

The KMS key ARN in the resource field is incorrect.

Why wrong: The KMS key ARN in the resource field is likely correct; the policy uses a wildcard or placeholder and the key is in the same account.

The policy is missing an encryption context that matches the secret's encryption context.

Why wrong: Encryption contexts are not relevant to the issue; the policy does not require an encryption context for decryption.

The policy only allows Decrypt, but the application needs Encrypt permission.

Why wrong: The application only needs Decrypt permission to decrypt the secret; it does not need Encrypt.

The DB instance is not Multi-AZ, causing failover delay.

Why wrong: Multi-AZ affects availability, not performance.

The DB instance status is 'available', meaning it is not accepting connections.

Why wrong: 'available' means the instance is running and ready for connections.

The engine version 8.0.27 has a known performance bug.

Why wrong: No known bug, and version is recent.

Disable the MySQL query cache to free up CPU.

Why wrong: Disabling query cache is not a diagnostic step.

Increase the instance class to provide more CPU resources.

Why wrong: This is a remediation, not an investigation.

Enable Multi-AZ to distribute the load to a standby instance.

Why wrong: Multi-AZ does not help with CPU investigation.

The DB instance is in a public subnet and the application server is in a private subnet, so they cannot communicate.

Why wrong: Incorrect. Instances in public and private subnets within the same VPC can communicate using private IPs; subnet type does not prevent connectivity.

The security group for the DB instance does not allow inbound traffic from the application server's security group.

Why wrong: Incorrect. The stem explicitly states that the security group allows inbound traffic from the application server's security group, so this is not the cause.

The DB instance is not part of a DB subnet group that includes the private subnet.

Why wrong: Incorrect. A DB subnet group defines which subnets the DB instance can be placed in, but it does not affect connectivity after the instance is launched.

Queries_per_second

Why wrong: Throughput doesn't indicate concurrency issues.

Slow_queries

Why wrong: Slow queries are a subset; not all high threads are slow.

Threads_connected

Why wrong: Connected threads may be idle; running threads are active.

Increase the number of database connections.

Why wrong: More connections can increase I/O contention.

Enable query caching in PostgreSQL.

Why wrong: Query caching is not a standard PostgreSQL feature and does not directly resolve disk queue depth.

Enable Multi-AZ deployment.

Why wrong: Multi-AZ provides high availability, not I/O performance improvement.

Create a read replica to offload read traffic

Why wrong: Read replicas don't reduce writer CPU.

Increase the instance size to handle the load

Why wrong: Scaling up is a reaction, not a diagnostic step.

Modify the DB instance class to a burstable type

Why wrong: Changing class is a fix, not diagnosis.

DatabaseConnections

Why wrong: This metric shows the number of connections, but not which queries are causing high CPU.

ReadIOPS

Why wrong: ReadIOPS indicates disk I/O, not directly CPU usage by queries.

CPUUtilization

Why wrong: CPUUtilization confirms high CPU but does not identify the specific query causing it.