Managing implementation and operations
Managing implementation and operations
This domain tests operating deployed solutions: monitoring, logging, incident response, release management, and day-to-day operations.
Core rules
- Monitoring and logging: Cloud Monitoring dashboards and alerting on SLIs; Cloud Logging with log sinks to BigQuery/Cloud Storage; log-based metrics; uptime checks; Ops Agent for VM metrics/logs.
- Incident management: route alerts to PagerDuty/Slack; Error Reporting aggregates exceptions; Cloud Trace for distributed tracing; define runbooks; conduct postmortems.
- Release management: Cloud Deploy for GKE; traffic splitting (Cloud Run revisions, GKE services) for canary; Binary Authorization; rollback to previous revisions.
- Operations: Cloud Asset Inventory for visibility; Config Controller for policy; Cloud Scheduler for cron; Cloud Tasks for async work; manage quotas via console/API.
- Security operations: Security Command Center findings; Cloud Audit Logs; Access Transparency; VPC Flow Logs for network forensics.
Worked example
A GKE app has intermittent 5xx errors. Reason: set up Cloud Monitoring uptime checks and alerting on error rate; use Cloud Logging to search logs; Cloud Trace to find slow spans; Error Reporting to group exceptions; deploy the fix via Cloud Deploy canary with traffic splitting; monitor error-budget burn. This reduces MTTR.
Traps
Monitoring only infrastructure, not user-facing SLIs; not centralizing logs; ignoring log retention costs; no rollback plan; alert fatigue from low-severity alerts; forgetting to monitor quotas; not using log-based metrics.