Monitoring, Performance, and Troubleshooting
What this domain tests
Monitoring, Performance, and Troubleshooting tests health checks, alerting, metric interpretation, log collection, and layered isolation. You must find the bottleneck and prove it, not restart services until the symptom disappears.
Core rules
- Start with health: NCC checks, cluster alerts, capacity, and failed components. A storage rebuild or network error can look like an application problem.
- Use a baseline. Compare current latency, IOPS, CPU ready, and network drops with normal values for the same time of day.
- Isolate by layer: guest OS, VM, host, storage, network, and upstream dependency. Change one variable at a time.
- Averages hide spikes. Check percentiles, queue depth, and retransmits.
Worked example
Users report a slow VM. Check the guest first for CPU saturation and application errors. Then compare host CPU ready and memory pressure; if other VMs on the host are healthy, inspect storage latency and network drops for this VM. If storage latency rises only during backup windows, the backup job is the likely cause. Confirm with time-correlated metrics before moving the VM.
Traps
Clearing alerts without root cause; checking only one layer; trusting averages; ignoring NCC; making multiple changes at once; and forgetting time synchronization when correlating logs. Collect logs and note the timeline before escalation.