Design for high availability

Sable Robotics runs an Azure Kubernetes Service cluster named aks-telemetry in the North Europe region. The cluster has a single user node pool that was created without availability zones. The company requires that the workloads keep running if one availability zone in North Europe fails, and it wants to minimize cost. A regional outage does not need to be covered.

What should you recommend?

1. AAdd a new user node pool that spans zones 1, 2 and 3, move the workloads to it, and delete the original node pool.Zones are fixed when a node pool is created, so a new zonal node pool is required; spreading nodes across three zones keeps capacity available when one zone fails, at the same node cost.
2. BUpdate the existing user node pool to add availability zones 1, 2 and 3.The availability zone setting of an AKS node pool cannot be changed after creation.
3. CEnable the cluster autoscaler on the existing node pool with a minimum of three nodes.Autoscaling adds nodes within the same zone configuration, so all nodes would still be lost together in a zone outage.
4. DDeploy a second AKS cluster in West Europe and distribute traffic with Azure Traffic Manager.A second cluster in another region would survive a zone failure but doubles the cost to cover a regional outage that the company does not require.

Tidewater Energy runs a metering platform on an Azure Database for PostgreSQL flexible server named pg-meter in the Germany West Central region. The company requires a 99.99 percent availability SLA, automatic failover with zero data loss (RPO of 0) if a datacenter in the region fails, and no manual intervention during failover. Cross-region recovery is handled separately and cost must be minimized.

What should you recommend?

1. AEnable zone-redundant high availability on pg-meter.Zone-redundant HA keeps a synchronously replicated standby in a different availability zone with automatic failover, which gives the 99.99 percent SLA and zero data loss for a zone failure.
2. BEnable same-zone high availability on pg-meter.Same-zone HA also replicates synchronously with automatic failover, but both nodes share one zone, so a datacenter failure takes both down.
3. CCreate a read replica of pg-meter in Germany North and promote it when needed.Read replicas use asynchronous replication, so data loss is possible, and promotion is a manual operation.
4. DConfigure geo-redundant backup storage for pg-meter.Geo-redundant backups support a restore in the paired region after an outage; they do not provide automatic failover or a running standby.

Corvid Payments runs a settlement web application on four Azure virtual machines behind a public Standard Load Balancer in the East US 2 region. Clients connect through a DNS name. The company requires an active-passive disaster recovery design in Central US with an RTO of 1 hour and an RPO of 15 minutes, requires that client traffic is redirected to Central US automatically after a failover, and wants to minimize cost by not running the application continuously in both regions.

Which two actions should you recommend? Each correct answer presents part of the solution.

1. AReplicate the four VMs to Central US with Azure Site Recovery and create a recovery plan that fails them over behind a pre-created load balancer.Site Recovery keeps replicas in Central US without running VMs there until failover, delivering the 15-minute RPO and a failover well within 1 hour at replication-only cost.
2. BCreate an Azure Traffic Manager profile with priority routing that lists the East US 2 load balancer first and the Central US load balancer second, and point the application DNS name at the profile.Priority routing with endpoint monitoring moves DNS resolution to the Central US endpoint automatically once the primary fails health checks, with no manual DNS edits.
3. CDeploy a second set of four VMs in Central US that runs continuously and receives half of the client traffic.An active-active deployment doubles compute cost, which violates the requirement not to run the application continuously in both regions.
4. DBack up the VMs daily to a Recovery Services vault with Cross Region Restore enabled and restore them in Central US when needed.A daily backup gives an RPO of 24 hours, and a manual multi-VM restore is unlikely to meet a 1-hour RTO.
5. EPeer the virtual networks in East US 2 and Central US and place both sets of VMs behind one internal load balancer.A regional load balancer cannot front backends in another region, and peering does not provide public client redirection.

Quillon Games stores match state for players in Western Europe and Southeast Asia in an Azure Cosmos DB for NoSQL account. The company requires that players in both geographies can write with single-digit-millisecond latency to a nearby region, that writes continue automatically if either region is lost, and that the account qualifies for a 99.999 percent availability SLA for both reads and writes.

What should you recommend?

1. AAdd West Europe and Southeast Asia to the account and enable multi-region writes.With multi-region writes every region accepts writes locally, so each geography gets low-latency writes, no region depends on another to stay writable, and the 99.999 percent read and write SLA applies.
2. BKeep a single write region in West Europe, add Southeast Asia as a read region, and enable service-managed failover.Players in Southeast Asia must write across the globe to West Europe, and the write SLA for a single-write-region account is 99.99 percent, not 99.999 percent.
3. CCreate a separate account in each region and synchronize them by using the change feed and an Azure Functions app.Two independent accounts do not form one highly available dataset; conflict handling and failover would be custom code and no platform SLA covers the combination.
4. DEnable availability zone redundancy for the account in West Europe.Zone redundancy improves resilience within West Europe but does nothing for write latency in Asia or for the loss of the whole region.