| Host CPU | Hypervisor run time (real physical load), root partition, sockets, cores, NUMA, vCPU:LP ratio, clock throttling, individual overloaded cores | Host overloaded, vCPU over-commitment, root partition saturated (check VMQ), CPU throttled |
| VM CPU | Load per VM and vCPU, CPU wait per dispatch (≈ CPU Ready), limits and reserves | Too many or too few vCPUs with target value, high CPU wait, vCPUs larger than one NUMA node |
| Memory | Free/used, assignment per VM, guest memory demand, Dynamic Memory, host paging, Smart Paging | Memory demand too high or too low, static or dynamic by role, memory recommendation per host |
| Storage | Volumes incl. CSV, media type (SSD/HDD/NVMe/SAN/Storage Spaces), throughput, IOPS, latency, queue, VHDX, checkpoints, SMB latency, Storage QoS | Volume nearly full, high latency, old checkpoints, VHD → VHDX, orphaned differencing disks, QoS limit reached |
| Network | Physical and virtual adapters, utilisation, errors, VMQ/RSS, vSwitches, SET and LBFO teams, VLANs, TCP connections | Adapter at its limit, team without redundancy, LBFO → SET, high retransmit rate, VMs without their own queue |
| VM state | Heartbeat, Hyper-V state, integration services, NUMA distribution, Hyper-V Replica | VM not responding, outdated integration services, memory on a remote NUMA node, replication overdue |
| Causes | Plain-language diagnosis per overload, processes and paging in the guest, Windows event logs of the hosts, known culprits | “Storage bottleneck on volume X – affects 6 VMs”, “Host overloaded – the VM is waiting for CPU” |
| Trends | Capacity forecast, baseline per weekday/hour, idle VMs, performance profile per volume | “Volume below 10 % free in about 23 days”, “VM idle for 30 days – frees 4 vCPU and 8 GB” |
| Load balancing | Headroom of all hosts, hardware, configuration version, CPU generation, space for VHDX | Up to three VM moves with load before/after |
| Clusters & S2D | Totals across all nodes, state of nodes, roles and cluster networks, Storage Spaces Direct | Cluster not N+1 capable, node paused, virtual disk degraded |
| Drives & hardware | SSD wear, temperature, errors, iLO/iDRAC via Redfish, MPIO, clock drift, power plan | Drive near end of life, power supply critical, only one MPIO path, clock drifting |
| Security & best practices | Defender, firewall, Secure Boot, VBS, SMB1, live migration, generation 1, vTPM, time synchronisation | “Firewall off: domain”, “Generation 1 – no Secure Boot possible” |
| Backup & patch level | Last backup per VM (Veeam), Windows Server Backup, patch level, pending reboot | “Backup: last run failed”, “Different patch levels in the cluster” |
| Alerting | E-mail and Teams on overload, escalation, end and critical Windows events; maintenance windows | Quiet hours, flap suppression, reminder for ongoing problems |
| Reports & docs | HTML, PDF, CSV, JSON, monthly report with logo, inventory as Excel, change log, licence estimate | “Standard: 5 × 16 cores for 9 VMs”, “Datacenter cheaper from about 11 VMs” |
| Planning | Plan a new VM, savings potential, tickets by e-mail or webhook | “VM with 8 vCPU/64 GB fits only on HV03”, “16 vCPUs unused” |
| Benchmark | Storage benchmark (throughput, IOPS, latency P50/P95/P99), CPU and memory, network path, before/after | “A single write takes 8 ms on average – check the RAID controller’s write cache” |
| User experience | Input delay per session, RDP encoding time, dropped frames, round-trip time to the client, GPU in the guest | “Input delay avg 420 ms – 3 of 4 sessions above 200 ms” |
| Operations & integration | Self-monitoring, interface for PRTG, Zabbix and Prometheus/Grafana, Active Directory search | “Dashboard certificate expires in 12 days”, “HV03 unreachable for 20 min” |
| Extensible | Any Windows performance counter and any WQL query via configuration | Your own thresholds for warning and critical, above or below |