Task 3.1: Implement highly available solutions to meet resilience and business requirements.

• AConvert the DB instance to a Multi-AZ DB instance deployment

  Correct

  RDS then keeps a synchronous standby in a different Availability Zone and fails over to it automatically when the primary or its Availability Zone fails.

• BCreate a read replica in a second Availability Zone and promote it by hand if the primary DB instance becomes unavailable

  Incorrect

  Promotion is a manual step that turns the replica into a standalone instance, so it adds operational work during the outage.

• CReplace the zonal NAT gateway with a regional NAT gateway and route the private subnets in every Availability Zone to it

  Correct

  A regional NAT gateway expands across the Availability Zones where the workload runs, so outbound traffic no longer depends on one zone.

• DAdd a second zonal NAT gateway in another Availability Zone and keep all private route tables pointed at the first NAT gateway

  Incorrect

  Traffic still goes only through the first NAT gateway, so losing its Availability Zone still cuts off the other zones.

• EMove the DB instance to a larger instance class with Provisioned IOPS storage to absorb the extra load after a zone failure

  Incorrect

  More capacity in one Availability Zone does not help when that zone is the one that fails.

The two single points of failure are the single-AZ database and the NAT gateway in one zone. Multi-AZ gives the database automatic failover, and a regional NAT gateway removes the shared dependency on one Availability Zone.

• ACreate a separate table in each Region and have the application write every item to both tables in one TransactWriteItems call

  Incorrect

  A TransactWriteItems call can only target tables in the same AWS account and Region, so it cannot write to tables in two Regions atomically.

• BCreate an MRSC global table with replicas in us-east-1 and us-east-2 only, and no component in any other Region

  Incorrect

  An MRSC global table must be deployed in exactly three Regions, as three replicas or as two replicas and a witness.

• CCreate an MRSC global table with replicas in us-east-1 and us-east-2 and a witness in a third supported Region

  Correct

  MRSC replicates each write synchronously to another Region before it succeeds, strongly consistent reads return the latest version, and a witness holds no readable copy in the third Region.

• DKeep a single-Region table with point-in-time recovery and copy its backups to us-east-2 with AWS Backup every hour

  Incorrect

  Restoring a backup copy loses the writes made since the last copy, and us-east-2 would not serve traffic until the restore finishes.

Multi-Region strong consistency (MRSC) replicates each write synchronously to at least one other Region before it succeeds, so no acknowledged write is lost and a strongly consistent read in any replica Region returns the latest version. It needs three Regions, and one of them can be a witness that takes part in the replication without holding a readable copy of the data.

• ATurn on cross-zone load balancing for the Network Load Balancer or for its target group

  Correct

  With cross-zone load balancing each node spreads traffic across targets in all enabled zones. It is off by default for Network Load Balancers.

• BRaise the target group's deregistration delay so that the busy targets finish their in-flight connections before new ones arrive

  Incorrect

  The deregistration delay only controls how long a deregistering target stays in the draining state. It does not change where new traffic goes.

• CTurn on stickiness for the target group so that each client keeps using the same target for the length of its session

  Incorrect

  Stickiness keeps clients on targets they already use, which does not move load from the first zone to the second.

• DLower the DNS TTL of the load balancer's record so that clients resolve the node addresses again more often

  Incorrect

  Each node already gets about half of the traffic. The imbalance comes from each node routing only within its own zone.

When cross-zone load balancing is off, each load balancer node routes only to targets in its own Availability Zone, so 2 targets share half the traffic while 8 share the other half. Application Load Balancers always have it on at the load balancer level; Network Load Balancers do not by default.

• AAn S3 Event Notification on each PUT that invokes a Lambda function to copy the object and to publish an alert if the copy takes longer than 15 minutes

  Incorrect

  This is custom code that the team must run and monitor, and it gives no AWS-defined replication time.

• BA Cross-Region Replication rule with S3 Replication Time Control, and an event notification for OperationMissedThreshold events

  Correct

  S3 RTC copies most objects in seconds and 99.9% within 15 minutes, and it emits an OperationMissedThreshold event for objects that miss the threshold.

• CA Cross-Region Replication rule without Replication Time Control, and a CloudWatch alarm on the bucket's BucketSizeBytes metric in eu-central-1

  Incorrect

  Without RTC there is no defined replication time, and a storage size metric cannot tell which objects are late.

• DAn S3 Batch Replication job that an EventBridge Scheduler schedule starts every 15 minutes for objects that have not been replicated

  Incorrect

  Batch Replication is for existing objects. Running it on a schedule delays new copies and still needs custom tracking to prove timeliness.

S3 Replication Time Control adds a replication time target and visibility to replication: S3 Replication metrics for pending and late objects, and event notifications when an object misses or later meets the 15-minute threshold.

• AThe least outstanding requests algorithm with slow start mode turned on

  Incorrect

  Slow start ramps up new targets; it does not react to targets that return errors.

• BA health check with a lower unhealthy threshold and a shorter interval

  Incorrect

  The targets pass health checks, so stricter checks may still not catch them and would remove them fully.

• CThe weighted random algorithm with anomaly mitigation turned on for the group

  Correct

  Anomaly mitigation routes traffic away from targets whose error rates deviate from others, and it needs weighted random routing.

• DSticky sessions so that each client keeps using the same target over time

  Incorrect

  Stickiness keeps clients on a target even when it returns errors.

Automatic Target Weights detects targets with significantly higher error rates than others in the target group. With weighted random routing, mitigation reduces their traffic until they recover.

• AWeighted records with equal weights for the three Regions, each associated with a health check

  Incorrect

  Health checks remove a failed Region, but equal weights spread users without regard to their latency to each Region.

• BLatency records for the three Regions, each associated with a health check

  Correct

  Latency routing picks the Region with the lowest latency, and when that record is unhealthy Route 53 chooses another healthy record by the same criteria.

• CLatency records for the three Regions without health checks, with a low TTL on each record

  Incorrect

  Users get the lowest-latency Region, but records without a health check are always considered healthy, so users keep going to a failed Region.

• DFailover records with us-east-1 as primary and the other two Regions as secondary

  Incorrect

  Failover is active-passive: it sends all users to one Region at a time and leaves the others idle.

Latency-based routing picks the Region with the lowest latency for each user. Health checks on every record let Route 53 skip an unhealthy Region and answer with the next healthy one.

• AMove the broker to a larger instance type in the same Availability Zone

  Incorrect

  A larger broker in one zone still fails with that zone.

• BReplace the broker with an SQS queue and change the application to use the SQS API

  Incorrect

  This is an application redesign, which the team wants to avoid.

• CTake daily backups of the broker configuration so that a new broker can be created after a failure

  Incorrect

  Rebuilding a broker after a failure still stops messaging for a while.

• DUse an active/standby broker deployment with two brokers in different Availability Zones

  Correct

  Active/standby brokers run a redundant pair across two zones with storage that is redundant across zones.

A single-instance broker runs in one zone. An active/standby deployment places two brokers in different zones as a redundant pair, with one active and one on standby.

• AMove the content to an EFS Regional file system with mount targets in all three zones

  Correct

  Regional file systems store data redundantly across multiple zones and stay available when a zone is unavailable.

• BKeep One Zone and turn on its automatic backups so that content can be restored after an outage

  Incorrect

  Backups help recovery but do not keep the content available during the outage.

• CAdd mount targets for the One Zone file system in the other two zones

  Incorrect

  The data would still be stored in one zone.

• DCopy the content to an EBS volume on each instance every night

  Incorrect

  Nightly copies on each instance are stale and add manual work.

EFS offers Regional and One Zone file systems. Regional, the recommended type, keeps data available across zones; One Zone data lives in a single zone.

• AKeep the datasets only in S3 Express One Zone, because it stores data on multiple devices

  Incorrect

  Its devices are within a single zone, so losing the zone can affect the data.

• BMove the cluster's reads to S3 Standard and stop using S3 Express One Zone

  Incorrect

  This gives up the low latency that training needs.

• CKeep the source of truth in a general purpose bucket and stage working data in Express

  Correct

  General purpose buckets store data across zones, while the directory bucket keeps serving low-latency reads.

• DCreate a second directory bucket in the same Availability Zone as a backup copy

  Incorrect

  Both copies would be in the same zone.

S3 Express One Zone stores data in directory buckets within one zone for high performance. Data that must survive a zone loss belongs in a storage class that spans zones.