CCNA Troubleshooting Questions

Option A is correct because moving read-intensive workloads to the read replica reduces CPU on the primary. Option B is wrong because Multi-AZ failover does not help with CPU load. Option C is wrong because increasing instance size would help but is not the best first step.

Option D is wrong because read replicas are asynchronous and cannot be used for writes.

Option D is correct because auto scaling uses a target utilization (default 70%) and cannot scale fast enough for sudden spikes. Option A is wrong because auto scaling is enabled. Option B is wrong because WCU are separate.

Option C is wrong because partition count doesn't directly cause throttling if RCU is sufficient.

Option B is correct because hot keys or uneven access patterns can cause throttling even if overall provisioned capacity appears sufficient. Option A is wrong because increasing capacity units should reduce throttling if the issue is simply capacity. Option C is wrong because Global Tables replication does not cause throttling on the source table.

Option D is wrong because DynamoDB Accelerator (DAX) reduces read load on the table, not write, and improper use might cause cache misses but not ProvisionedThroughputExceededException.

Option A is correct because the DatabaseMemoryUsagePercentage metric indicates the percentage of the node's memory used, which can cause latency if high. Option B is wrong because CPUUtilization indicates CPU load, not memory. Option C is wrong because NetworkBytesIn measures incoming network traffic.

Option D is wrong because ReplicationLag only applies to clusters with replicas.

Option A is correct because a low cache hit ratio indicates the cache is not being used effectively, often due to a low TTL. Option B is wrong because insufficient read replicas would not cause low cache hits. Option C is wrong because high CPU might cause latency but not necessarily low cache hits.

Option D is wrong because insufficient write capacity is not relevant for ElastiCache read behavior.

Option B is correct because Redshift performs better with fewer, larger files. Consolidating small files into larger ones reduces the overhead of opening many files. Option A is incorrect because the issue is not with the COPY command syntax.

Option C is incorrect because changing the distribution style does not affect COPY performance directly. Option D is incorrect because the issue is not with the column data type.

Option A is correct because running ANALYZE updates table statistics, which helps the query planner choose the best plan. Option B is wrong because increasing maintenance_work_mem helps with autovacuum but not query planning. Option C is wrong because disabling sequential scans is a brute-force approach that may not be optimal.

Option D is wrong because setting cpu_tuple_cost changes the cost model but might not be necessary.

ReadLatency and WriteLatency directly measure the time taken for I/O operations. High values indicate I/O bottlenecks. DatabaseConnections and CPUUtilization measure other resources.

FreeStorageSpace indicates storage capacity, not performance.

Option A is correct because throttling with on-demand capacity can be due to a single partition receiving too many requests. Option B is wrong because switching to provisioned capacity does not fix the root cause. Option C is wrong because CloudWatch does not have a metric for hot partitions.

Option D is wrong because enabling auto scaling is for provisioned capacity.

Option B is correct because 'ReplicationLatency' measures the time taken for updates to propagate. Option A is wrong because 'ThrottledRequests' is for throttling. Option C is wrong because 'PendingReplicationCount' shows number of pending items, not time.

Option D is wrong because 'ConsumedWriteCapacityUnits' is for capacity.

Option B is correct because deleting old automated backups reduces storage used by backup data. Option A is wrong because manual snapshots are separate and do not affect automated backup storage. Option C is wrong because disabling backups would remove point-in-time recovery capability.

Option D is wrong because modifying DB instance class does not reduce storage consumption.

Option B is correct because Evictions metric shows the number of keys evicted due to memory pressure. Option A is wrong because CacheHits shows successful retrievals. Option C is wrong because CurrItems shows the current number of items, not evictions.

Option D is wrong because SwapUsage shows swap usage, not evictions.

Queries that use the GSI can consume read capacity from the index. However, if the query uses the base table's primary key, it consumes from the base table. Option A is wrong because local secondary indexes don't help.

Option B is wrong because adjusting write capacity doesn't affect reads. Option C is correct because enabling DynamoDB Accelerator (DAX) caches reads, reducing read capacity consumption. Option D is wrong because changing the sort key is not a direct solution.

Option A is correct because the DMS replication instance may have insufficient memory. Increasing its size provides more memory. Option B is wrong because source database memory is not controlled by DMS.

Option C is wrong because target RDS instance memory may not be the bottleneck. Option D is wrong because S3 is not involved in the DMS process by default.

Option D is correct because the logs indicate table corruption and page corruption, typically due to underlying storage issues. Option A is wrong because connection issues would show different errors. Option B is wrong because backup restore would not cause corruption.

Option C is wrong because replication lag shows different symptoms.

Option C is correct because deleting old automated backups frees up storage space without affecting the current backup capability. Option A is wrong because modifying the backup retention period does not free space immediately. Option B is wrong because manual snapshots are stored in Amazon S3 and do not affect the DB instance storage.

Option D is wrong because disabling backups would stop future backups.

Option C is correct because auto scaling should increase capacity to meet demand, but if it is not scaling fast enough, adjusting the scaling policy can help. Option A is wrong because disabling auto scaling is not a solution. Option B is wrong because the pattern is uniform, so partitioning is not the issue.

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

Option B is correct because ReadThrottleEvents directly measures throttled read requests. Option A is wrong because ConsumedReadCapacityUnits shows actual usage, not throttled events. Option C is wrong because ThrottledRequests includes write throttles as well.

Option D is wrong because ReadLatency shows latency, not throttling.

Option C is correct because SVV_DISKUSAGE provides information about disk usage per slice, which helps identify skew. Option A is wrong because STL_SCAN is for scan operations. Option B is wrong because STV_TBL_PERM shows permissions.

Option D is wrong because PG_TABLE_DEF shows table definitions.

Option B is correct because CloudWatch Alarms can trigger on FreeStorageSpace metric. Option A is wrong because SNS is a notification service, not monitor. Option C is wrong because CloudTrail is for API activity.

Option D is wrong because Config tracks configuration changes.

Option B is correct because Performance Insights shows detailed query-level performance. Option A is wrong because CloudWatch Logs are for log files. Option C is wrong because CloudTrail is for API calls.

Option D is wrong because S3 does not monitor RDS memory.

Option A is correct because the Lambda function creates new connections, exhausting the pool. Using RDS Proxy manages connections efficiently. Option B is wrong because increasing max_connections may cause resource contention.

Option C is wrong because reducing provisioned concurrency may not solve the root cause. Option D is wrong because decreasing max_connections would worsen the issue.

Option C is correct because Performance Insights provides a database performance tuning and monitoring feature that allows you to assess the load on your database and determine when and where to take action. Option A is wrong because CloudWatch Logs does not capture query-level performance. Option B is wrong because CloudTrail tracks API activity, not database queries.

Option D is wrong because RDS Enhanced Monitoring provides OS-level metrics, not query-level metrics.

Option B is correct because increasing the DMS task's max LOB size or setting LOB mode to 'Limited' can reduce overhead when replicating large objects. Option A is wrong because increasing Aurora instance size may help but DMS task configuration is more direct. Option C is wrong because disabling Multi-AZ does not affect DMS performance.

Option D is wrong because using a smaller Aurora instance would worsen performance.

Option B is correct because AWS DMS provides metrics for replication lag. Option A is wrong because CloudWatch can monitor DMS metrics but is not specific to Data Guard. Option C is wrong because RDS Performance Insights does not monitor replication lag.

Option D is wrong because CloudTrail does not monitor replication lag.

Option B is correct because increasing storage is immediate and prevents the instance from going into storage-full state. Option A is wrong because deleting logs may free space temporarily but is not a scalable solution. Option C is wrong because archiving to S3 does not free space.

Option D is wrong because it does not solve the immediate issue.

Option D: On-demand mode has a per-table throughput limit (e.g., 40,000 read/write units per second). When sustained traffic exceeds that, requests are throttled. Option A: Global tables replicate writes to other regions but do not cause throttling.

Option B: DynamoDB Accelerator (DAX) reduces latency but can still see throttling if the table itself is throttled. Option C: Adaptive capacity automatically scales partitions but cannot exceed the on-demand limit.

Option C is correct because if the partition key is not chosen well, data can be skewed, causing hot partitions. Even if total consumed capacity is below provisioned, a single partition may receive more requests than its share of capacity, causing throttling on that partition. Option A is wrong because global tables do not cause slow queries on the source table.

Option B is wrong because DynamoDB Streams do not affect read latency. Option D is wrong because DAX is a caching layer that speeds up reads, not slows them.

FreeableMemory shows the amount of available RAM. Low FreeableMemory indicates memory pressure, which can cause swapping and slow queries. Option A is wrong because CPUUtilization measures CPU, not memory.

Option B is wrong because DatabaseConnections tracks connections, not memory. Option C is wrong because ReadIOPS measures I/O operations, not memory.

Option A is correct because Deadlocks is a specific metric in Amazon RDS for SQL Server that tracks deadlock occurrences. Option B is wrong because it is a generic metric. Option C is wrong because it is for blocking, not deadlocks.

Option D is wrong because it is for lock waits.

Option C is correct because ReplicationLatency measures the time between the last update on the source table and the last update on the replica table. Option A is wrong because ThrottledRequests indicates throttling. Option B is wrong because ConsumedWriteCapacityUnits measures usage.

Option D is wrong because SuccessfulRequestLatency measures request latency.

Options A, D, and E are correct. Enabling DynamoDB Accelerator (DAX) reduces read load. Implementing exponential backoff helps retry gracefully.

Using DynamoDB global tables can distribute write load. Option B is wrong because disabling auto scaling would not help. Option C is wrong because switching to on-demand capacity might help but is not always cost-effective and does not address hot keys.

Option A (DMLThroughput) measures the rate of write operations. Option D (CommitLatency) measures the time to commit transactions, which directly affects write latency. Option B is incorrect because it is a read metric.

Option C is incorrect because it measures storage. Option E is incorrect because it is a read latency metric.

Options B and D are correct. A high disk queue depth indicates I/O bottleneck. Increasing IOPS (Option B) or enabling Auto Scaling (Option D) can alleviate the bottleneck.

Option A is wrong because more connections may increase I/O contention. Option C is wrong because query caching is not a primary solution for I/O. Option E is wrong because Multi-AZ does not improve I/O performance.

Option B is correct because increasing the instance size provides more resources for replication. Option A is wrong because pausing replication would increase lag. Option C is wrong because switching to full load would cause downtime.

Option D is wrong because disabling logging would break CDC.

Options A and C are correct because CloudWatch Logs can monitor error logs in real time and set metric filters for specific error patterns. Option B (S3 export) is for archiving, not real-time monitoring. Option D (CloudTrail) is for API activity, not database logs.

Option E (X-Ray) is for tracing, not log monitoring.

Option B is correct because VACUUM reclaims space from deleted rows without adding cost. Option A is wrong because adding nodes increases cost. Option C is wrong because increasing slice count is not possible without changing node type.

Option D is wrong because compression reduces storage but may not immediately free space.

Option B is correct because enabling Enhanced Monitoring at the instance level provides OS-level metrics that can help identify resource bottlenecks causing high write latency. Option A is wrong because profiling queries in the slow query log may show slow queries but not necessarily latency at the OS level. Option C is wrong because creating additional replicas does not reduce write latency on the primary.

Option D is wrong because switching to Amazon DynamoDB is a major architectural change, not a troubleshooting step.

Option C is correct because a high volume of write operations can cause the primary instance to become overloaded, leading to failover. Option A is incorrect because DatabaseConnections alone does not directly cause failover. Option B is incorrect because ReadLatency is a symptom, not a direct cause of failover.

Option D is incorrect because FreeableMemory is not a typical cause of failover.

Option A is correct because scaling up the instance provides more CPU capacity. Option B is wrong because read replicas help with read scaling but not CPU on the writer. Option C is wrong because gp3 storage does not affect CPU.

Option D is wrong because changing the query cache is unlikely to significantly reduce CPU.

Amazon RDS events are stored in the RDS console under 'Events' and also in CloudTrail. But the best place to see failover reasons is the RDS console events. Option B is correct because the RDS console shows events like 'failover' with reasons.

Option A is wrong because CloudTrail records API calls, not internal failover reasons. Option C is wrong because CloudWatch Logs does not automatically log failover reasons. Option D is wrong because StatusInsis not a service.

Option A is correct because changing the isolation level to READ COMMITTED eliminates gap locks for non-unique indexes, reducing lock contention. Option B is wrong because increasing innodb_lock_wait_timeout only increases timeout, not reduces locks. Option C is wrong because adding an index on transaction_date may increase lock granularity? Actually, if the index is not unique, gap locks still occur.

Option D is wrong because partitioning does not affect gap locks.

Option B is correct because enabling slow query log helps identify long-running queries causing high CPU and connections. Option A is wrong because Multi-AZ failover does not resolve performance issues. Option C is wrong because enhancing CloudWatch metrics does not directly fix timeouts.

Option D is wrong because read replicas are for read scaling, not for reducing connection timeouts on the primary.

Option A is correct because DynamoDB charges for read capacity units based on item size, and using GetItem with strongly consistent reads consumes 1 RCU for items up to 4 KB. Option B is wrong because changing to provisioned capacity may not be cost-effective with unpredictable traffic. Option C is wrong because compression is not natively supported.

Option D is wrong because DAX adds cost and may not reduce RCU consumption if reads are strongly consistent.

Option B is correct because high disk usage on dc2 nodes indicates that the cluster is spilling to disk due to insufficient memory. Option A is wrong because network throughput is not indicated. Option C is wrong because WLM queue waits would show in other metrics.

Option D is wrong because high CPU alone does not explain disk usage.

Option A is correct because enabling Performance Insights helps identify top SQL queries. Option B is correct because reviewing slow query logs helps identify inefficient queries. Option C is wrong because increasing instance size is a fix, not investigation.

Option D is wrong because disabling Multi-AZ reduces availability, not CPU. Option E is wrong because modifying parameter groups without analysis is not investigative.

Option B is correct because changing the eviction policy to allkeys-lru allows Redis to evict less frequently used keys when memory is full, reducing cache misses. Option A is wrong because adding more nodes does not solve the eviction issue if the policy prevents eviction. Option C is wrong because increasing the memory of existing nodes only delays the problem.

Option D is wrong because enabling persistence does not help with memory management.

Option C is correct because TTL deletion can be delayed if the table has a high write rate; increasing provisioned write capacity can speed up the deletion process. Option A is wrong because DynamoDB does not have a 'ttl_deletion_lag' parameter. Option B is wrong because TTL handles deletion automatically; a Lambda function is unnecessary.

Option D is wrong because the attribute type is correct (Number).

Option C is correct because Amazon RDS Event Notifications for 'failover' events are delivered within seconds. Option A is wrong because CloudWatch metric 'DatabaseConnections' dropping to zero may indicate failover but the alarm evaluation period is typically 1 minute, causing delay. Option B is wrong because Enhanced Monitoring metrics have a granularity of 1 second but do not directly indicate failover.

Option D is wrong because CloudTrail logs API calls like FailoverDBCluster but these are not real-time, often with minutes delay.

Options A and C are correct because CloudWatch Logs can be used to stream RDS logs, and Performance Insights can help identify slow queries. Option B is wrong because CloudTrail records API calls, not database logs. Option D is wrong because S3 is a storage service, not a monitoring service.

Option E is wrong because Config records configuration changes.

Option B is correct because enabling Enhanced Monitoring provides OS-level metrics that can help identify resource bottlenecks. Option A is wrong because increasing instance size without diagnosis may not address the root cause. Option C is wrong because switching storage type does not reduce CPU load.

Option D is wrong because deleting slow query logs removes diagnostic data.

Options B and D are correct because they directly identify top CPU queries and resource-intensive databases. Option A is wrong because increasing CPU doesn't diagnose. Option C is wrong because read replicas don't help with CPU.

Option E is wrong because instance class change is a fix, not diagnosis.

Options B and D are correct. DatabaseConnections shows current connections, and BurstBalance helps understand if I/O is throttled. Option A is wrong because SwapUsage is not a standard RDS metric.

Option C is wrong because ReadLatency is about I/O latency, not connections. Option E is wrong because NetworkTransmitThroughput is about network, not connections.

Option C is correct because this error indicates the cache has reached its memory limit and eviction policy is not set or set to noeviction. Option A is wrong because that error is about keyspace notifications. Option B is wrong because cluster mode may affect sharding but not this specific error.

Option D is wrong because connection limit would show a different error.

Even with on-demand capacity, DynamoDB can throttle if you exceed the account-level provisioned throughput limits or if there is a sudden spike that exceeds the previous peak traffic by a large margin. Option C is correct; Option A is incorrect because on-demand does not have per-table limits. Option B is incorrect because adaptive capacity helps but does not prevent all throttling.

Option D is a possible reason but not the most likely.

Option B is correct because Amazon RDS provides the Multi-AZ DB cluster deployment option which includes two readable standby instances, allowing read traffic to be offloaded to the standby to reduce load on the primary. Option A is incorrect because increasing the instance size would increase compute costs. Option C is incorrect because enabling Performance Insights does not reduce read latency.

Option D is incorrect because increasing the allocated storage does not directly improve read latency for random reads.

Option C is correct because global tables replicate writes across regions, and each write generates a stream record. The Lambda function should filter on the 'awsRegion' attribute in the stream record to avoid processing the same update from different regions. Option A is wrong because disabling streams in eu-west-1 would break the replication.

Option B is wrong because using a dedicated stream for each region is not supported. Option D is wrong because increasing Lambda concurrency doesn't prevent duplicate processing.

Option A is correct because STL_QUERY contains query execution details including resource consumption. Option B is wrong because STV_TBL_PERM shows table permissions, not query performance. Option C is wrong because PG_TABLE_DEF shows table definitions.

Option D is wrong because SVV_VACUUM_PROGRESS shows vacuum progress, not query performance.

Option A is correct because high binlog replication lag on a reader indicates that the reader is falling behind in applying changes from the writer. Option B is wrong because instance size affects performance but not specifically lag. Option C is wrong because storage is shared across all instances in Aurora.

Option D is wrong because parameter groups are shared across the cluster.

Option A is correct because enabling RDS Proxy reduces connection churn and pools connections, reducing lock contention. Option B is wrong because increasing instance size adds CPU but doesn't directly address lock contention. Option C is wrong because reducing connections may increase contention per connection.

Option D is wrong because read replicas don't handle write locks.

Option A is correct because even if the security group allows traffic, the DB subnet group must be in a public subnet with an internet gateway. Option B is wrong because the issue is connectivity, not backups. Option C is wrong because CloudWatch Logs may not show connection attempts.

Option D is wrong because parameter groups don't control network access.

Option A is correct because creating an index on the queried fields avoids collection scans. Option B is wrong because increasing instance size helps but doesn't address the root cause. Option C is wrong because changing consistency level affects read freshness, not performance.

Option D is wrong because increasing storage doesn't improve query speed.

Option D is correct because Multi-AZ deployments with read replicas offload read traffic and reduce latency. Option A is wrong because increasing instance size may help but does not specifically address read latency. Option B is wrong because enabling Multi-AZ alone does not reduce read latency.

Option C is wrong because disabling Multi-AZ would remove the standby, reducing availability.