Task 6.1: Understanding the principles and recommendations of the operational excellence pillar of the Google Cloud Well-Architected Framework

• ARestrict console access to one senior operator who makes every change

  Incorrect

  Changes stay manual and undocumented, and one person becomes a bottleneck and a single point of failure.

• BExport Cloud Audit Logs to BigQuery and review the changes each month

  Incorrect

  Reviewing logs afterward detects changes; it does not prevent bad ones or make them repeatable.

• CManage changes as code in version control with reviewed pipelines

  Correct

  The pillar's automate-and-manage-change principle calls for infrastructure as code and a structured change process, which makes every change reviewed, recorded and repeatable.

• DOversize every instance so that operators rarely need to resize them

  Incorrect

  Larger instances do not stop undocumented console edits, and firewall rules would still be changed by hand.

Operational excellence treats change as code: proposals are reviewed, applied by automation and recorded, so outages from ad hoc console edits stop.

• ASpecific, measurable SLOs for the critical user journeys, with monitoring and alerts

  Correct

  The pillar asks teams to define SMART SLOs and SLAs for critical workloads and back them with monitoring and alerting.

• BA 100% availability target, so that every single failure is treated as an SLO breach

  Incorrect

  A 100% target is unrealistic and leaves no error budget for change.

• CSLOs written after the first major incident shows what users care about

  Incorrect

  Waiting for an incident means launching without agreed expectations.

• DA CPU utilization threshold for each VM as the main measure of service

  Incorrect

  Resource metrics do not reflect what users experience of the service.

Operational readiness means agreeing on measurable performance expectations and monitoring them from day one, not discovering them in production.

• AA rule that the most senior engineer alone may take any action during an outage

  Incorrect

  One person cannot investigate, communicate and fix at once, and this creates a bottleneck.

• BDefined incident roles, such as an incident commander, with documented runbooks

  Correct

  Clear roles, communication protocols and runbooks or playbooks give a coordinated, repeatable response to incidents.

• CMore alerting policies so that more engineers join every incident

  Incorrect

  Paging more people adds to the uncoordinated activity that caused the problem.

• DRoot cause reviews that begin once the next quarter's planning ends

  Incorrect

  Delayed reviews do not fix how incidents are run while they happen.

Managing incidents well depends on preparation: defined roles, communication and documented procedures, which the pillar lists under clear incident response procedures.

• AWrite a more detailed manual checklist for each of the routine operations tasks

  Incorrect

  Better checklists still rely on people repeating manual steps correctly.

• BHire additional operators so that each person repeats fewer tasks

  Incorrect

  Adding people scales the toil instead of removing it.

• CAutomate the routine tasks with tested tooling and configuration management

  Correct

  Automating configuration and routine management reduces the risk of human error that manual processes carry.

• DSchedule the routine tasks for nights, when traffic is at its lowest

  Incorrect

  Moving manual work to nights changes the timing, not the error risk.

Automation is a core operational excellence practice: it removes repetitive manual work and the errors that come with it.

• ARely on autoscaling within the current quotas during the sale

  Incorrect

  Autoscaling stops at the quota limit, so it cannot reach three times the usual capacity.

• BMove the sale's frontend VMs to Spot VMs to free up capacity

  Incorrect

  Spot VMs can be preempted at the busiest moment, which is a risk for a sale.

• CEstimate the quota needs and request quota adjustments in advance

  Correct

  The pillar advises estimating quota requirements and requesting adjustments before the resources are needed.

• DUse future reservations to assure capacity for the GPU machines

  Correct

  For specialized hardware such as GPUs, future reservations give assurance that the capacity will be there.

• ESet the VM groups to their peak size permanently from today

  Incorrect

  Paying for peak capacity for two months is wasteful compared with planning for the event.

Capacity planning means knowing the limits and securing resources before a known peak, rather than finding the ceiling during the event.