Infrastructure Monitoring
What the domain tests
Infrastructure Monitoring tests host, cloud, Kubernetes, and custom telemetry: the infrastructure agent, integrations, entity synthesis, tagging, and saturation alerts. The exam focuses on choosing the right metric and avoiding false alarms.
Core rules
- The infrastructure agent collects host data such as SystemSample, ProcessSample, StorageSample, and NetworkSample. Integrations add service-specific metrics.
- Cloud integrations poll provider APIs (for example CloudWatch, Azure Monitor, and GCP) and synthesize entities; polling can lag and may miss short spikes.
- Kubernetes monitoring uses the Kubernetes integration and cluster explorer; node, kubelet, and state metrics answer different questions.
- Prometheus/OpenMetrics, StatsD, and Flex bring custom metrics. Tags drive entity synthesis, dashboards, and alert scoping.
- Alert on saturation and pressure-available memory, disk queue, CPU steal-rather than raw utilization alone.
Worked example
A host memory alert fires at 92% used, but the app is healthy. Linux page cache inflates "used" while memory is reclaimable. Check available memory and process RSS, then alert on available memory percentage or a baseline condition. Reason: utilization without pressure does not prove a user-impacting problem.
Traps
Treating high CPU as automatically bad; confusing agent and integration coverage; missing tags that break entity links; ignoring cloud polling delay; high-cardinality custom metrics; static thresholds on bursty workloads.