7.5 Knowledge Check - Operations, Resilience, and Performance

1.

An architect must give FinCorp's payments-adjacent service an automated failover between two availability zones. The team asks which managed IONOS product orchestrates the failover by watching the primary, declaring it dead, and promoting the secondary across the whole stack. What is the correct answer, and what is the platform's actual automated failover mechanism?

A IONOS sells no managed failover product; the native automated pattern composes load-balancer health checks, which determine endpoint health, with a low-TTL Cloud DNS record that is re-pointed to steer traffic toward healthy endpoints by changing which address a name resolves to. Cloud DNS itself is not health-aware. B The Managed Network Load Balancer is the orchestrator; once two zones are registered as targets it automatically promotes the standby zone and fails the whole stack over, including stateful connections. C A dedicated Cloud DNS "health-check failover record" type is a point-and-click console feature that watches the primary and promotes the secondary with no further design work. D VM Auto Scaling performs the failover by detecting the dead primary and launching a replacement stack in the second zone automatically.

2.

A team places a primary database node and its standby in the same VDC and lets the platform assign availability zones automatically, assuming that an automatically assigned zone gives them multi-zone redundancy. The architect flags this as the auto-zone trap. Why is the assumption wrong, and what is the correct discipline?

A Automatic zone assignment always spreads a redundant pair across different zones, so the design is already multi-zone and no change is needed. B Multi-zone redundancy is impossible within a single VDC, so the standby must be moved to a separate contract before it can sit in a different zone. C Letting the platform choose zones can co-locate a redundant pair in the same zone, defeating the redundancy; every member of a redundant pair must be placed into an explicitly named, distinct availability zone at design time. D Auto-zone is a managed multi-AZ feature that synchronously replicates the pair across all zones, so explicit placement would actually reduce resilience.

3.

FinCorp splits production, non-production, and a compliance-isolated workload across separate contracts, and also runs Managed Kubernetes clusters. The operations team expects a single managed pane that unions all telemetry and expects Kubernetes control-plane events to arrive in the Logging Service alongside application logs. Which statement correctly describes the observability boundaries they must design around?

A Both expectations are met out of the box: the Logging Service aggregates across all contracts automatically, and control-plane events are delivered to it once a cluster is created. B Cross-contract aggregation works automatically, but control-plane events require enabling the cluster's "Logging to S3" toggle to appear in the Logging Service. C Control-plane events flow into the Logging Service by default, but cross-contract aggregation is only available for production-tagged contracts. D There is no cross-contract aggregation, since Monitoring and Logging pipelines are per-contract and per-region, and Kubernetes control-plane events never reach the Logging Service, so both a unified pane and control-plane visibility must be composed externally.

4.

FinCorp's regulated transaction database holds only a few tens of gigabytes of live data, and an engineer proposes provisioning it on a 40 GB SSD volume to match the small data footprint. The architect overrules this. Which reasoning is correct for sizing the volume?

A SSD performance is flat and independent of volume size, so 40 GB performs identically to 400 GB; the volume should be kept at 40 GB to save cost. B SSD throughput and IOPS accrue per gigabyte, so a volume below roughly 100 GB sits under the performance floor and degrades a database tier even when the data set is tiny; the volume is provisioned at or above 100 GB regardless of data footprint, sizing for performance first. C HDD scales its performance with volume size while SSD does not, so the database should be moved to a larger HDD volume rather than a small SSD one. D The Data Center Designer guarantees full SSD performance at any size, so the 40 GB volume will perform at the per-VM ceiling without any floor concern.

5.

FinCorp must migrate a large VMware estate into IONOS. The project plan assumes a native OVF/OVA import wizard for the VMs and a replication-based cutover for the databases moving onto IONOS Managed PostgreSQL. The architect rejects both assumptions. Which description of the correct engineered approach is accurate?

A There is a native OVF/OVA import wizard, so the VMs import directly, and the databases follow them through the same wizard with zero downtime. B Databases migrate by a native replication-based cutover into Managed PostgreSQL, while the VMs are moved cross-site live with vMotion from on-premises into IONOS. C The whole estate is moved with a single backup/restore pass, and running VMs are relocated from on-premises into IONOS live, so no maintenance window is needed for any wave. D There is no native OVF/OVA import, so migration is engineered per workload across three paths: image conversion and upload to the KVM Public Cloud surface, VMware-native replication and failover into dedicated Private Cloud with a Layer 2 VPN for address continuity and intra-cluster-only live mobility, or backup/restore; and the database wave is a dump-and-restore hard cutover, not replication.