CCNA Monitoring and Troubleshooting Questions

Option B is correct because long-running transactions on the writer hold locks, increasing deadlocks. Option A is wrong because read replicas do not cause locks on writer. Option C is wrong because different instance sizes do not cause deadlocks.

Option D is wrong because Performance Insights is diagnostic, not a cause.

Option C is correct because enabling auto scaling allows DynamoDB to adjust capacity based on traffic patterns. Option A is wrong because increasing only the secondary region may not be sufficient without auto scaling. Option B is wrong because turning off global tables would prevent replication.

Option D is wrong because switching to on-demand is expensive and not the best practice for predictable workloads.

Option C is correct because the table is in us-west-2, but the developer may be querying from a different region or the table name is case-sensitive. Option A is wrong because the table is ACTIVE. Option B is wrong because the ARN is present.

Option D is wrong because the command succeeded.

Option C is correct because ReplicationLatency metric shows delay between regions. Option A is wrong because Streams are used for replication; check the metric first. Option B is wrong because auto scaling affects capacity, not replication.

Option D is wrong because table size doesn't directly cause replication delay.

Adding a read replica offloads read traffic from the primary. Increasing the instance size can provide more memory for caching, reducing read IO. Option C (increasing storage) may help if it increases IOPS, but not necessarily.

Option D (enabling Multi-AZ) does not improve read performance. Option E (using Amazon ElastiCache) is a valid approach but not specific to RDS.

Option D is correct because CacheHits and CacheMisses together give the hit rate. Option A is wrong because CPUUtilization is for resource usage. Option B is wrong because Evictions indicates memory pressure, not hit rate.

Option C is wrong because CurrConnections shows active connections.

Option A is correct because adding certain options may require a reboot. Option B is incorrect because the option group modification is likely valid; the issue is that it requires a reboot. Option C is incorrect because modifying the DB instance again would not help.

Option D is incorrect because the DB instance is not in a failed state.

Option C is correct because an explicit Deny overrides any Allow. The Deny statement specifically denies DeleteDBInstance on that resource. Option A is wrong because the policy allows CreateDBInstance.

Option B is wrong because DescribeDBInstances is allowed. Option D is wrong because the Deny is explicitly on the production-db ARN.

Option B is correct because Lambda can be triggered directly from DynamoDB Streams. Option A is wrong because Kinesis Data Firehose is for loading streaming data into destinations, not for triggering Lambda. Option C is wrong because SQS is for message queuing, not directly from DynamoDB Streams.

Option D is wrong because Step Functions coordinates workflows, but does not directly trigger from DynamoDB Streams without Lambda.

Option C is correct because reducing the maximum connections in the parameter group and rebooting immediately limits the active connections, preventing the database from being overwhelmed. Option A is wrong because modifying the security group does not affect the number of connections. Option B is wrong because increasing the instance size may help but takes time to provision.

Option D is wrong because terminating all connections will disrupt the application.

Option A is correct because physical reads cause disk I/O. Option C is correct because write operations generate I/O. Option D is correct because dirty buffer flushes cause writes.

Option B is incorrect because logical reads are from memory, not disk. Option E is incorrect because network throughput is not disk I/O.

Option B is correct because Enhanced Monitoring provides OS-level metrics, while Performance Insights provides query-level details. Option A is wrong because CloudWatch Logs is for log files. Option C is wrong because AWS CloudTrail is for API calls.

Option D is wrong because VPC Flow Logs capture network traffic.

Options B and D are correct. FullTableScans (B) directly counts full scan operations. TableScanRows (D) indicates the number of rows read via full scans.

Option A is wrong because BufferCacheHitRatio indicates cache efficiency, not full scans. Option C is wrong because UserCommits shows transaction commits. Option E is wrong because RedoLogSpaceUsage relates to write-ahead logging.

Options A and D are correct. Amazon CloudWatch provides metrics and alarms for DynamoDB. AWS CloudTrail logs API calls but does not monitor performance metrics.

Option B is incorrect because AWS Config tracks resource changes, not performance. Option C is incorrect because VPC Flow Logs capture network traffic. Option E is incorrect because Amazon Inspector is a security assessment service.

Option D is correct because increasing RCU or using DAX can reduce read throttling. Option A is wrong because auto scaling target should be 70%, not 90%. Option B is wrong because decreasing WCU doesn't help reads.

Option C is wrong because GP2 is for RDS, not DynamoDB.

Option B is correct because Performance Insights shows wait events and SQL queries. Option A is wrong because CloudWatch Logs doesn't show lock waits directly. Option C is wrong because Enhanced Monitoring shows OS metrics.

Option D is wrong because AWS Trusted Advisor is for best practices.

Option B is correct because the port 5432 is the default PostgreSQL port, but the security group may not allow inbound traffic on that port. Option A is wrong because the endpoint is correct. Option C is wrong because Multi-AZ does not affect connectivity.

Option D is wrong because there is no encryption requirement implied.

Option C is correct. The output shows a failover event followed by a restart, which is typical after a failover. Option A is incorrect because the events show a failover, not a manual restart.

Option B is incorrect because the failover was completed, not failed. Option D is incorrect because there is no indication of a snapshot restore.

Option C is correct because RDS Performance Insights shows active session information including connections. Option A is wrong because CloudTrail tracks API calls, not connections. Option B is wrong because VPC Flow Logs track network traffic, not database connections.

Option D is wrong because Trusted Advisor gives best practice checks, not real-time connections.

Option A is correct because auto scaling may have a lag in provisioning capacity; pre-warming can help. Option C is correct because a hot partition can cause throttling even if overall capacity is not exhausted. Option B is incorrect because write sharding is not a built-in DynamoDB feature.

Option D is incorrect because global tables add complexity and are not a direct fix for throttling. Option E is incorrect because decreasing read capacity does not help writes.

Option C is correct because CloudWatch metrics such as WriteThrottleEvents, ConsumedWriteCapacityUnits, and ProvisionedWriteCapacityUnits (for on-demand, the provisioned is the maximum) help identify throttling causes. Option A is wrong because DynamoDB Streams track changes, not throttling. Option B is wrong because CloudTrail logs API calls but not throttling events at the item level.

Option D is wrong because VPC Flow Logs capture network traffic, not database throttling.

Options B, C, and D are correct. WLMQueueWaitTime measures time queries wait in queue, QueryDuration measures execution time, and DiskSpaceUsage indicates if disk spills are causing slowness. Option A (NetworkThroughput) is about network, not query performance.

Option E (DatabaseConnections) is about connection count, not query speed.

Option B is correct because the Deny statement for all DynamoDB actions overrides the Allow for PutItem. Deny statements always take precedence. Option A is wrong because the table name is correct.

Option C is wrong because the role is assumed. Option D is wrong because the user exists.

Option D is correct because concurrent writes to the same item in different regions cause conflicts. Option A is wrong because provisioned throughput affects throttling, not conflicts. Option B is wrong because network latency does not cause conflicts.

Option C is wrong because eventual consistency does not cause conflicts.

Option B is correct because enabling slow query log captures queries that exceed a threshold, allowing analysis. Option A is wrong because Enhanced Monitoring shows OS metrics, not query details. Option C is wrong because Performance Insights shows wait events and SQL, but slow query log specifically targets slow queries.

Option D is wrong because CloudWatch metrics show overall performance, not per-query.

Option A is correct because CPUUtilization is a primary indicator of resource saturation. Option B is wrong because DatabaseConnections shows concurrent connections, not resource usage. Option C is wrong because FreeableMemory is important but CPU is more likely the first bottleneck.

Option D is wrong because SwapUsage is relevant but not the first metric to check.

Option B is correct because modifying storage to 1,000 GB does not require downtime and automatically adds more space. Option A is wrong because deleting logs may not free enough space and requires manual intervention. Option C is wrong because taking a snapshot and restoring to a larger instance causes downtime.

Option D is wrong because enabling storage auto-scaling does not immediately solve the current full state.

Option B is correct because DynamoDB Auto Scaling adjusts the provisioned throughput automatically in response to traffic patterns, reducing throttling. Option A is wrong because switching to eventually consistent reads may reduce throttling but changes the consistency model, which may not be acceptable. Option C is wrong because increasing read capacity is a manual action and not efficient.

Option D is wrong because increasing write capacity does not address read throttling.

Option B is correct because increasing the instance size of the reader can improve its ability to apply changes faster. Option A is wrong because increasing the writer size may not help if the reader is the bottleneck. Option C is wrong because reducing transaction size helps but may not be feasible.

Option D is wrong because Multi-AZ is always enabled for Aurora.

Option B is correct because Performance Insights provides database performance analysis and can identify missing indexes. Option A is wrong because RDS Enhanced Monitoring provides OS-level metrics. Option C is wrong because CloudWatch Logs are for application logs.

Option D is wrong because AWS CloudTrail logs API calls.

Option B is correct. The average CPU utilization increases from 75.5% to 98.2% over the hour, indicating a steady increase. Option A is incorrect because it is not decreasing.

Option C is incorrect because it is not constant. Option D is incorrect because it does not fluctuate; it increases monotonically.

Option D is correct because monitoring FreeStorageSpace and transaction log usage reveals what consumes space. Option A is wrong because backup storage is separate. Option B is wrong because Enhanced Monitoring shows OS metrics, not disk usage by database.

Option C is wrong because CloudTrail is for API activity.

Option A is correct because adding nodes distributes workload and increases CPU and I/O resources. Option B is wrong because VACUUM reorganizes data but does not free space if the disk is full; it may require space. Option C is correct because optimizing sort keys reduces the amount of data scanned, speeding up queries.

Option D is wrong because disabling concurrency scaling reduces performance. Option E is wrong because Multi-AZ is not available for Redshift. So correct: A, C.

Option C: Using Amazon EBS io2 Block Express volumes with higher IOPS and lower latency can reduce the time for the secondary to catch up. Option A: EBS Multi-Attach is not supported for io2? Actually io2 supports multi-attach but for shared volumes, not typically used for MongoDB. Option B: Using instance store volumes may have higher performance but are ephemeral and data loss is possible.

Option D: Snapshots are for backups, not for improving failover performance.

Option C is correct because automated backups with point-in-time recovery enable recovery to any point within the retention period, meeting 5-minute RPO. Option A is wrong because Multi-AZ is for high availability, not backup. Option B is wrong because read replicas don't provide backup.

Option D is wrong because manual snapshots are taken on-demand, not continuously.

Option C is correct because creating read replicas offloads SELECT queries from the primary, reducing CPU. Option A is wrong because increasing instance size again may not be cost-effective and the issue is read-heavy. Option B is wrong because enabling Performance Insights only helps monitoring, not performance.

Option D is wrong because Multi-AZ is for failover, not read scaling.

Option D is correct because on-demand capacity mode automatically scales but can throttle if traffic is unevenly distributed across partitions, leading to hot partitions. Option A is incorrect because on-demand mode does not have a per-table limit in the same way provisioned capacity does. Option B is incorrect because DynamoDB does not have a global write limit per account that would cause throttling on a single table.

Option C is incorrect because the table uses on-demand capacity mode, not provisioned.

Option C is correct because implementing exponential backoff reduces request rate temporarily. Option A is wrong because switching to on-demand takes 30 minutes. Option B is wrong because deleting GSI does not help writes to main table.

Option D is wrong because DAX caches reads, not writes.

Option B is correct because using strongly consistent reads ensures the latest data is read. Option A is wrong because increasing WCU does not affect consistency. Option C is wrong because DAX caches data and may serve stale data.

Option D is wrong because DynamoDB Streams are for change data capture, not consistency.

Option C is correct because using a single partition key creates a hot partition, leading to throttling even with on-demand capacity (which has per-partition limits). The best solution is to redesign the partition key to distribute writes. Option A is incorrect because auto scaling is not needed for on-demand.

Option B is incorrect because DAX caches reads, not writes. Option D is incorrect because creating a GSI does not affect write throttling on the base table.

Option A is correct because checking the STV_SLICES table shows disk usage per slice, revealing data distribution skew. Option B is wrong because the STL_QUERY table logs query text, not distribution. Option C is wrong because the STV_WLM_SERVICE_STATE table shows queue state, not distribution.

Option D is wrong because CloudWatch metrics like CPUUtilization do not indicate distribution skew.

Option A is correct because evictions occur when memory is full and the cache evicts keys based on the eviction policy. This would cause session loss. Option B is wrong because replication lag would cause stale data, not loss.

Option C is wrong because snapshotting may cause latency but not evictions. Option D is wrong because encryption in transit does not affect memory.

Option B is correct because the high CPU and low memory indicate the instance is overworked. Scaling up to a larger instance provides more CPU and memory resources. Option A is incorrect because adding a reader instance helps with read scaling but does not reduce CPU/memory pressure on the primary writer.

Option C is incorrect because creating an index may improve query performance but the team already ruled out query optimization. Option D is incorrect because increasing storage does not directly affect CPU or memory.

Options A, B, and D are correct. Auto-scaling cannot handle sudden bursts (A), hot partitions (B), or exceeded max capacity (D). Option C is wrong because eventual consistency doesn't cause throttling.

Option E is wrong because GSI consumes write capacity from the table.

Option A is correct because increasing the maximum read capacity prevents throttling. Option B is correct because enabling DynamoDB Accelerator (DAX) reduces read load. Option E is correct because implementing exponential backoff helps handle throttling gracefully.

Option C is wrong because decreasing target utilization would cause auto scaling to scale out earlier, but it would not prevent throttling during spikes because auto scaling reacts after the spike. Option D is wrong because write capacity is not the issue.

Option A and Option C are correct. ReadIOPS and WriteIOPS measure the input/output operations per second, which can indicate storage performance issues. Option B (DatabaseConnections) is about connections, not storage performance.

Option D (CPUUtilization) is about CPU usage. Option E (NetworkReceiveThroughput) is about network throughput.

Option B is correct because DynamoDB Accelerator (DAX) caches reads, reducing read capacity consumption. Option A is wrong because increasing read capacity costs more. Option C is wrong because auto-scaling still incurs cost.

Option D is wrong because changing to On-Demand may be more expensive for predictable workloads.

Option D is correct because high WriteLatency and growing oplog suggest that the secondary instance is too slow to apply operations, causing replication lag. Checking the secondary's metrics helps diagnose. Option A is wrong because enabling Multi-AZ does not directly address write latency.

Option B is wrong because increasing the instance class may help but should be done after diagnosis. Option C is wrong because increasing storage does not improve write performance.

Option A is correct because low free space can cause performance degradation on gp2; increasing storage increases baseline IOPS. Option B is correct because high 'DiskQueueDepth' indicates I/O bottleneck; switching to Provisioned IOPS provides consistent I/O. Option C is correct because enabling Multi-AZ does not improve I/O performance, it adds overhead.

Option D is wrong because Performance Insights is a monitoring tool, not a fix. Option E is correct because deleting old data frees up space and can improve performance. So correct: A, B, E.

Option A is correct because with on-demand capacity, DynamoDB can handle sudden traffic spikes without throttling. Increasing read capacity is not applicable because on-demand mode automatically adjusts. Optimistic locking is already used implicitly with condition expressions.

Adding a global secondary index does not directly reduce UpdateItem latency.

Option A is correct because Aurora's fast failover requires the JDBC driver to use the cluster endpoint with the Aurora hostlist provider. Option B is wrong because increasing instance size does not reduce failover time. Option C is wrong because Multi-AZ is already inherent in Aurora.

Option D is wrong because increasing backup retention does not affect failover.

Option A is correct because high CPUUtilization indicates the cache nodes are overloaded. Option B is wrong because CacheHits are a measure of effectiveness, not provisioning. Option C is wrong because Evictions occur when memory is full but CPU is more direct for throughput.

Option D is wrong because ReplicationLag indicates replication issues, not capacity.

Option A is correct because high 'Threads_running' often indicates queries waiting on locks or I/O. Option B is wrong because 'Connections' shows total connections, not concurrent active queries. Option C is wrong because 'Queries' shows total queries over time, not concurrency.

Option D is wrong because 'Slow_queries' shows only long-running queries.

Option B is correct because enabling Performance Insights provides a detailed analysis of database performance, helping to identify the root cause of high CPU utilization. Option A is wrong because increasing instance size without understanding the cause may lead to unnecessary costs. Option C is wrong because creating a read replica does not directly address CPU utilization on the primary instance.

Option D is wrong because switching to a different database engine is a major change and not a troubleshooting step.

Option B is correct because a large transaction can cause increased latency without high CPU or connection count, as it may be waiting on disk I/O or replication. Option A is wrong because Multi-AZ failover is automatic and would show a spike then recovery. Option C is wrong because insufficient storage I/O would show in 'BurstBalance' or 'WriteIOPS' metrics, not just latency.

Option D is wrong because normal CPU and connections rule out a runaway query in terms of CPU, though a query could be I/O-bound. However, a large transaction is a common cause of such symptoms.

Option B is correct because checking for unused indexes or bloat is a typical starting point for storage issues. Option A is wrong because modifying storage is a solution, not a troubleshooting step. Option C is wrong because deleting logs may not recover much space.

Option D is wrong because enabling auto scaling is a preventive measure, not a troubleshooting step.

Option A is correct because ConcurrencyScalingActiveQueries indicates queries running on concurrency scaling clusters. Option B is wrong because WLMQueueLength shows queue wait, not concurrency scaling. Option C is wrong because QueryDuration measures query execution time.

Option D is wrong because DiskSpaceUsage is for storage, not concurrency.

Option D is correct because the network ACL must allow inbound traffic on ephemeral ports for the response. Option A is wrong because the security group already allows inbound on 3306. Option B is wrong because the error is 'Connection timed out', not authentication.

Option C is wrong because if the VPC had no internet gateway, the error would be different for private connections.

Option C is correct. The output is the DNS endpoint of the RDS instance. Option A is wrong because it's not an IP address.

Option B is wrong because Resource ID looks different. Option D is wrong because the reader endpoint includes '-ro'.

Option C is correct because RDS events log failover reasons. Option A is wrong because Enhanced Monitoring does not capture failover events. Option B is wrong because CloudTrail logs API calls, not failover reasons.

Option D is wrong because error logs may not include failover cause.

Option D is correct because Performance Insights with the db.sql_tokenized dimension allows grouping by user and application. Option A is wrong because RDS Enhanced Monitoring does not show SQL-level details. Option B is wrong because VPC Flow Logs capture network traffic, not database connections.

Option C is wrong because CloudTrail logs API calls, not database connections.

Option C is correct because ReadThrottleEvents counts throttled read requests. Option A is wrong because ConsumedReadCapacityUnits shows successful reads, not throttled ones. Option B is wrong because ThrottledWriteEvents counts write throttles.

Option D is wrong because SuccessfulRequestLatency measures latency, not throttles.

Option D is correct because RDS supports modifying storage settings without downtime, and the change occurs during the next maintenance window. Option A is wrong because creating a snapshot does not increase storage. Option B is wrong because gp3 is not supported for SQL Server.

Option C is wrong because read replicas do not increase storage on the primary.

Option A is correct because RA3 nodes use managed storage; resizing the cluster to a different node type or adding more nodes can increase storage capacity. Option B is wrong because VACUUM reorganizes data but does not free space if the disk is full; it may even require temporary space. Option C is wrong because Analytics is not a feature that affects storage.

Option D is wrong because disabling concurrency scaling does not free disk space.

Options A and C are correct because connection timeout and max allowed packet can cause aborted connections. Option B is wrong because max_connections limits total connections but not aborts. Option D is wrong because query cache is not related to connections.

Option E is wrong because binlog retention is for replication.

Option A is correct because 'ThrottledWriteRequests' directly indicates throttling. Option B is correct because 'WriteThrottleEvents' is another metric for throttling events. Option C is correct because 'ConsumedWriteCapacityUnits' helps understand if provisioned capacity is being fully used.

Option D is wrong because 'ReadThrottleEvents' is for reads, not writes. Option E is wrong because 'SuccessfulRequestLatency' measures latency, not throttling.

Option C is correct because increasing the instance class of the writer can improve performance. Option A is wrong because read replicas do not help with write latency. Option B is wrong because Auto Scaling does not apply to Aurora instance classes automatically.

Option D is wrong because Multi-AZ is already inherent in Aurora.