CCNA Monitoring and Troubleshooting Questions

Option D is correct because the table has no items (ItemCount: 0). Option A is wrong because the table is active. Option B is wrong because provisioned throughput is not exceeded (ItemCount is 0).

Option C is wrong because the attribute definition exists.

Option D is correct because if the write concern is set to 'majority' and a replica is down or lagging, the write cannot be acknowledged. Option A is wrong because the primary is not overloaded; the issue is write concern. Option B is wrong because read preference does not affect writes.

Option C is wrong because the number of replicas is sufficient for majority if all are healthy.

Options A, C, and D are correct. Using AWS Direct Connect provides a dedicated network connection, reducing latency. Compressing data reduces the amount of data transferred, mitigating latency impact.

Using multiple threads parallelizes the data transfer, improving throughput. B is wrong because increasing the instance size does not reduce network latency. E is wrong because using a VPN may introduce additional overhead.

Option C is correct because DynamoDB global tables replicate writes asynchronously; if the replica table in eu-west-1 is not properly configured as a global table replica, replication will fail. Option A is incorrect because global tables do not require VPC endpoints. Option B is incorrect because auto scaling does not affect replication.

Option D is incorrect because DynamoDB Streams are used for global tables replication, but disabling them would stop replication, not cause lag.

Option B is correct because increasing the storage IOPS on both instances reduces the time to write transaction logs to disk, which is the bottleneck. The primary generates logs, and the standby applies them; both need fast I/O. Option A is insufficient because disabling synchronous commit might cause data loss.

Option C is incorrect because the standby is already the same size; increasing its size alone may not help if the bottleneck is I/O. Option D is incorrect because reducing the batch size reduces write volume, affecting application performance.

Option A is correct because high replica lag indicates the replica is falling behind, often due to insufficient I/O capacity on the writer or replica. Option B is wrong because read replicas handle SELECT queries. Option C is wrong because binary log retention does not affect replica lag directly.

Option D is wrong because the DB cluster parameter group is shared.

Option B is correct because auto scaling uses a target utilization percentage; if it's set too high, it may not trigger scaling soon enough. Option A is incorrect because auto scaling controls the provisioned capacity, not the table class. Option C is incorrect because auto scaling does not require manual approval.

Option D is incorrect because auto scaling can increase capacity up to the maximum limit, regardless of the GSI.

Option D is correct because `Evictions` metric shows the number of evicted keys due to memory limit. Option A is wrong because `CacheHits` shows successful reads. Option B is wrong because `CurrConnections` shows current connections.

Option C is wrong because `ReplicationLag` shows replication delay, not evictions.

Option A is correct because Performance Insights can identify high-load queries. Option B is correct because Enhanced Monitoring provides OS-level metrics like CPU, memory, and disk I/O. Option C is wrong because increasing instance class is a solution, not an investigation step.

Option D is wrong because enabling Multi-AZ does not help investigate CPU usage. Option E is wrong because disabling query cache is not a diagnostic action.

Option B is correct because CloudWatch metrics for ThrottledWriteEvents help identify if auto scaling is keeping up. Option A is wrong because disabling auto scaling may cause throttling. Option C is wrong because changing to on-demand may not be cost-effective and is not a troubleshooting step.

Option D is wrong because increasing read capacity does not affect write latency.

Aurora Auto Scaling for readers uses the 'AverageActiveConnections' metric to determine when to add readers. Option B is correct. Option A is incorrect because the auto scaling is based on connections, not CPU.

Option C is incorrect because ReaderNodeCPU is not a predefined metric for auto scaling. Option D is incorrect because ReadLatency is an effect, not a scaling trigger.

B and D are correct. SwapUsage indicates memory pressure when it exceeds 0. FreeableMemory shows available memory.

A is wrong because CPUUtilization is CPU, not memory. C is wrong because DatabaseConnections affects memory but is indirect. E is wrong because ReadIOPS is disk I/O.

Options B, D, and E are correct. Checking the replica's load (B), examining replication threads (D), and reviewing network latency (E) are common troubleshooting steps. Option A (increasing instance size) is a solution, not a troubleshooting step.

Option C (disabling binary logging) would break replication.

RDS supports modifying storage to increase allocated storage without downtime. You can also enable storage auto scaling to automatically increase storage. Option C (deleting data) may free space but is not recommended as a proactive measure.

Option D (snapshot and restore) involves downtime. Option E (conversion to Aurora) is a migration, not a direct fix.

Option A (FreeStorageSpace) and Option D (FreeStorageSpaceInPercent) are correct. Option B is wrong because BinLogDiskUsage is for MySQL. Option C is wrong because DatabaseConnections is connections.

Option E is wrong because ReadLatency is for read operations.

Option D is correct because GSIs have their own capacity and if the GSI is throttled, writes to the base table can be throttled, affecting read latency. Option A is wrong because throttled requests are zero. Option B is wrong because on-demand capacity does not have a provisioned limit.

Option C is wrong because hot partitions would cause throttling.

Option D is correct because increasing the instance size of the reader provides more memory for the buffer pool, improving read performance and reducing replica lag. Option A is wrong because increasing the writer instance size does not directly help read latency on the reader. Option B is wrong because adding another reader does not help if the existing reader is under-resourced.

Option C is wrong because increasing storage does not directly improve memory or query performance.

Option C is correct because DynamoDB Accelerator (DAX) reduces read latency and throttling by caching. Option A is wrong because increasing write capacity doesn't help reads. Option B is wrong because enabling Streams may add overhead.

Option D is wrong because switching to on-demand may increase cost but not necessarily fix throttling if workload is bursty.

Option A is correct because modifying the allocated storage for an RDS instance is a dynamic operation that can increase storage without downtime. Option B is wrong because deleting old snapshots does not free up storage on the instance. Option C is wrong because creating a read replica requires the source to have available storage.

Option D is wrong because rebooting does not free storage.

Option A: Enabling Performance Insights provides a performance schema and wait events to identify problematic queries. Option C: Enabling the slow query log captures queries that take longer than a specified time, helping to pinpoint slow queries. Option B: RDS Enhanced Monitoring is for OS-level metrics, not query-level.

Option D: Amazon DevOps Guru is for broader anomaly detection, not specific query identification. Option E: RDS Proxy is for connection pooling, not query analysis.

Option C is correct because distributing data across more nodes allows parallel processing, which reduces query execution time. Option A is incorrect because the issue is CPU, not storage. Option B is incorrect because WLM queues manage concurrency, not individual query performance.

Option D is incorrect because vacuum reclaims space but does not directly speed up queries.

Option A is correct because a Multi-AZ failover can cause a brief outage during the transition. The restart at 14:30 is likely the result of the failover. Option B is wrong because the restart occurred after the failover.

Option C is wrong because the events do not indicate a maintenance window. Option D is wrong because there is no evidence of insufficient storage.

Option A is correct because STL_QUERY provides query execution details. Option C is correct because SVL_QUERY_REPORT shows step-level execution. Option E is correct because STL_WLM_QUERY shows WLM queue waits.

Option B is wrong because STV_BLOCKLIST shows disk allocation, not performance. Option D is wrong because STL_LOAD_ERRORS is for load errors.

Option D is correct because if the reader instances are overloaded, they may not be able to apply changes quickly enough, causing replica lag. The failover could be due to the failover priority settings or a health check failure. Option A is incorrect because storage issues would affect both writer and readers.

Option B is incorrect because an oversized instance would reduce lag, not increase it. Option C is incorrect because long-running queries on the writer would not directly cause replica lag on readers.

Option A is correct because Evictions metric shows how many keys are evicted due to memory pressure. Option B is wrong because CacheHits shows hits, not evictions. Option C is wrong because ReplicationLag is for replication.

Option D is wrong because CPUUtilization does not directly show evictions.

Option C is correct because Provisioned IOPS (io2) provides consistent low latency. Option A is wrong because gp3 may not provide enough IOPS for the workload. Option B is wrong because increasing storage size increases baseline IOPS for gp2, but 500 GB already gives 1500 IOPS; more IOPS may be needed.

Option D is wrong because Multi-AZ does not improve write performance; it may increase latency.

Option D is correct because Aurora's Performance Insights with the 'Wait for lock' dimension can identify blocking sessions. Option A is wrong because Amazon Inspector is a security assessment tool. Option B is wrong because CloudWatch Logs Insights can analyze logs but does not directly identify blocking transactions.

Option C is wrong because Enhanced Monitoring provides OS-level metrics, not database lock information.

Option D is correct because a failover results in a brief downtime (DatabaseConnections drop to zero) followed by a switch from master to standby (ReplicaLag disappears). Option A is wrong because WriteIOPS and ReadIOPS may change but not specifically indicate failover. Option B is wrong because FreeStorageSpace does not change during failover.

Option C is wrong because NetworkThroughput may fluctuate but is not a direct indicator of failover.

Options A and D are correct. ReadIOPS measures read operations per second, and ReadLatency measures the time for read operations; both are I/O metrics. WriteIOPS and WriteLatency are also I/O metrics but given the question, the correct ones are ReadIOPS and ReadLatency.

Option C is wrong because CPUUtilization indicates CPU-bound, not I/O-bound. Option E is wrong because NetworkReceiveThroughput measures network throughput.

Checking ThrottledRequests helps identify if requests are being throttled, even with on-demand. Reviewing PartitionKey metrics helps detect hot partitions. ReadCapacityUnits and WriteCapacityUnits are not applicable for on-demand.

Latency metrics show end-to-end performance but are less diagnostic.

Option C is correct because the CreateDBSnapshot action requires permissions on both the DB instance and the snapshot resource. The policy only grants access to the snapshot resource, not the DB instance. Options A and B are wrong because the actions are allowed.

Option D is wrong because the resource is specified correctly.

Option B is correct because the 'db.load.avg' metric in Performance Insights shows the average number of active sessions, which helps identify queries causing high load. Option A is incorrect because 'db.cpu.avg' is not a standard Performance Insights metric. Option C is incorrect because 'db.sessions' is not a standard metric.

Option D is incorrect because 'db.bytes_sent' relates to network throughput, not CPU.

Options A, C, and D are correct. Option A: Reducing transaction duration minimizes the window for lock conflicts. Option C: Accessing tables in a consistent order avoids cycle dependencies.

Option D: Using READ COMMITTED reduces shared locks for reads. Option B is incorrect because increasing retries does not prevent deadlocks; it only delays failure. Option E is incorrect because lowering isolation level to READ UNCOMMITTED can cause dirty reads and is not recommended for e-commerce.

Options B and D are correct. ReadIOPS and WriteIOPS show actual I/O operations. Option A is wrong because CPUUtilization is about CPU.

Option C is wrong because DatabaseConnections is about connections. Option E is wrong because FreeableMemory is about memory.

Option D is correct because a multi-threaded application may exhaust connections; using a connection pooler like RDS Proxy reduces connection churn. Option A is wrong because increasing max_connections may overload the instance. Option B is wrong because read replicas do not accept writes.

Option C is wrong because it's not a configuration change.

Option C is correct because increasing the instance size or switching to a storage-optimized instance reduces I/O latency. Option A is wrong because Aurora is not the same as PostgreSQL. Option B is wrong because Multi-AZ does not improve performance.

Option D is wrong because enabling auto-scaling only scales storage, not performance.

Option B is correct because ThrottlingException occurs when the request rate exceeds the provisioned throughput capacity. Option A is wrong because eventually consistent reads consume half the read capacity units, so they are less likely to cause throttling. Option C is wrong because network issues would cause timeout errors, not throttling.

Option D is wrong because incorrect table name would cause ResourceNotFoundException.

Options B and C are correct. Option B: Network latency between regions directly affects replication time. Option C: Large write volumes can cause backpressure on the replication stream.

Option A is incorrect because on-demand capacity does not throttle writes; it scales automatically. Option D is incorrect because Global Tables replicate all data to all regions; us-west-2 must receive all writes. Option E is incorrect because read capacity on a replica does not affect replication; replication uses write capacity in the replica region.

Neptune's query profiler provides detailed information about query execution time. The slow query log only logs queries that exceed a threshold. The audit log is for security events.

The performance insights is for RDS, not Neptune.

Option A is correct because CloudWatch metrics like CPUUtilization and CacheHits are the first step to diagnose. Option B is wrong because increasing node size is a solution, not a diagnostic step. Option C is wrong because ElastiCache does not have a built-in query profiler.

Option D is wrong because disabling persistence is unrelated.

Option D is correct because replicas have independent capacity; if replication writes exceed provisioned capacity, throttling occurs. Option A is wrong because burst capacity is not relevant to throttling. Option B is wrong because auto-scaling doesn't cause throttling.

Option C is wrong because DAX is for reads.

Option C is correct because INNODB_LOCKS shows current locks and InnoDB lock conflicts. Option A is wrong because PERFORMANCE_SCHEMA provides performance data but not specifically locks. Option B is wrong because INFORMATION_SCHEMA.PROCESSLIST shows current processes but not lock details.

Option D is wrong because INFORMATION_SCHEMA.TABLES shows table metadata.

Option D is correct because Performance Insights provides a dashboard to monitor active sessions and the queries they are running. Option A is wrong because AWS CloudTrail logs API calls, not database sessions. Option B is wrong because RDS Enhanced Monitoring provides OS metrics, not database session details.

Option C is wrong because Amazon Inspector is a security assessment tool.

Option D is correct because CloudWatch Logs can monitor log file sizes and trigger alarms based on metric filters. Option A is wrong because CloudTrail does not monitor log file sizes. Option B is wrong because RDS Enhanced Monitoring provides OS-level metrics, not log file metrics.

Option C is wrong because S3 is a storage service, not a monitoring service.

Option A is correct because CloudWatch metrics can show throttling, latency, and utilization. Option D is correct because DynamoDB Accelerator (DAX) can reduce latency if the issue is high read volume. Option B is wrong because increasing capacity is not needed for on-demand.

Option C is wrong because partitions are managed automatically. Option E is wrong because disabling TTL is unrelated.

Option B is correct because BurstBalance dropping to 0 indicates that the instance has exhausted its I/O burst credits, causing I/O throttling. Scaling up storage to a larger size increases the baseline I/O performance and burst credit earning rate. Option A is wrong because read replicas do not increase I/O credits.

Option C is wrong because enabling Performance Insights only helps diagnose, not resolve. Option D is wrong because changing the instance class does not affect I/O credits.

Option B is correct because increasing the max_connections parameter allows more concurrent connections. Option A is wrong because rebooting does not increase the connection limit. Option C is wrong because modifying the security group does not affect database connections.

Option D is wrong because the error is not about authentication.

Option A is correct because DatabaseConnections can indicate high concurrent sessions causing CPU spikes. Option B is wrong because ReadIOPS may not directly cause CPU. Option C is wrong because SwapUsage is memory-related.

Option D is wrong because NetworkThroughput is network-related.

Option C is correct because auto-scaling only triggers when storage usage reaches the maximum allocated storage and there is at least 10% free space remaining. Option A is wrong because Maximum Storage Duration is the maximum time the instance can run on provisioned storage before scaling. Option B is wrong because there is no minimum storage size for scaling.

Option D is wrong because storage auto-scaling does not require modification to the DB instance class.

The metric FreeStorageSpace shows the available storage in bytes. The alarm should be set to when FreeStorageSpace is less than 10% of allocated storage. Option B is correct.

Option A is wrong because it's a percentage of total. Option C is wrong because it's a fixed value. Option D is wrong because BinaryLog usage is not relevant.

Option D is correct because when BurstBalance is 0%, the gp2 volume has exhausted its burst credits and is operating at baseline IOPS. To improve performance, you can increase the volume size (which increases baseline IOPS) or switch to Provisioned IOPS (io1/io2). Option A is wrong because enabling Multi-AZ does not increase I/O performance.

Option B is wrong because disabling backups reduces CPU but not IOPS. Option C is wrong because creating a read replica does not improve write performance.

Option A is correct because an open transaction can be rolled back using the ROLLBACK command. Option B is incorrect because Point-in-Time Recovery restores the entire DB instance to a previous timestamp, not just the transaction. Option C is incorrect because ROLLBACK will undo the DELETE.

Option D is incorrect because COMMIT would make the DELETE permanent.

Option D is correct because using an adaptive capacity add-on allows DynamoDB to automatically manage throughput for hot partitions. Option A is wrong because increasing capacity units may not help if throttling is due to hot partitions. Option B is wrong because GSI does not directly solve hot partition throttling.

Option C is wrong because changing partition key distribution would require table redesign and data migration.

Option C is correct because DAX caches eventually consistent reads, not strongly consistent reads. Option A is wrong because TTL affects item expiration but not the consistency model. Option B is wrong because write capacity does not directly affect cache hits.

Option D is wrong because the DAX cluster size affects performance but not the consistency mismatch.

Option D is correct because the RDS for MySQL engine-specific counter 'Queries' reflects the number of queries executed. Option A is wrong because DatabaseConnections shows connections, not query performance. Option B is wrong because ReadIOPS measures disk I/O, not query volume.

Option C is wrong because NetworkThroughput measures network traffic.

Options B, C, and E are correct. FreeableMemory shows available memory. SwapUsage indicates memory pressure.

DatabaseConnections can increase if connections are not closed properly. Option A is incorrect because CPUUtilization is not a memory metric. Option D is incorrect because ReadIOPS relates to I/O.

The db.r5.large instance type has 2 vCPUs and 16 GiB of memory. Sustained CPU utilization above 90% during peak hours indicates that the instance is compute-bound for the workload. Upgrading to db.r5.xlarge (4 vCPUs, 32 GiB memory) doubles the available CPU capacity, directly reducing CPU utilization for the same query load.

The on-premises server had similar specifications, but RDS instances may have different CPU architectures or hypervisor overhead, making the larger instance the most direct fix.

Option D is correct because using the cluster endpoint (CNAME) automatically redirects to the current primary, reducing downtime. Option A is wrong because enabling Multi-AZ is already done. Option B is wrong because increasing instance size does not reduce failover time.

Option C is wrong because read replicas are for read scaling, not failover.

Options A and B are correct. CloudWatch provides metrics like ConsumedReadCapacityUnits and can trigger alarms. CloudWatch Logs can capture log events but is not directly used for capacity alarms.

CloudTrail records API calls, not capacity metrics. DynamoDB Auto Scaling adjusts capacity but does not monitor. AWS Config tracks configuration changes, not capacity.

Options A and C are correct. STL_QUERY contains query execution details. SVV_QUERY_STATE shows currently running queries.

Option B is wrong because STV_TBL_PERM shows permissions. Option D is wrong because PG_TABLE_DEF shows table definitions. Option E is wrong because STL_LOAD_COMMITS shows load commit information.

Option A is correct because security groups are stateful; for inbound traffic, the target security group must allow the incoming connection. Option B is incorrect because the subnet routing does not block specific ports. Option C is incorrect because NACLs are stateless but the issue is with security groups.

Option D is incorrect because the Redis cluster is in a VPC, not on-premises.

Global tables replicate writes asynchronously; network latency between regions affects replication. The write capacity on the replica table must be sufficient to handle replication writes. Item size affects write latency.

Option D (read capacity) does not affect write latency. Option E (auto scaling) is not a direct cause of high latency.

Option A is correct because an under-provisioned GSI can throttle reads even if the base table uses on-demand. Option B is wrong because on-demand handles bursts. Option C is wrong because TTL does not affect read latency.

Option D is wrong because DynamoDB Accelerator (DAX) reduces latency, not increases.

High CPU on the primary could be due to the binary log (binlog) being enabled by default for Aurora MySQL. Option B is correct because disabling binlog reduces CPU overhead. Option A is wrong because increasing instance size would help but is not the most likely fix.

Option C is wrong because read replicas don't affect write latency. Option D is wrong because increasing storage doesn't reduce CPU.