DPM Host Power-Down Eligibility for Low-Traffic Workloads: An Unresolved Mode Decision

Determining Host Power-Down Eligibility

A host is eligible for power-down only when it meets three concurrent criteria: it must be evacuable (no pinned VMs, local storage dependencies, or hardware passthrough), under-utilized (CPU and memory usage below the DPM threshold), and stable (remaining idle for the duration of the configured standby timer).

The signal that distinguishes a truly eligible host from one that merely appears idle is the DRS evacuation feasibility. DPM does not simply turn off a host because it is quiet; it first queries DRS to ensure all resident VMs can be migrated to other active hosts without violating cluster resource constraints or affinity rules. If DRS cannot find a valid placement for every VM on that host, the host remains powered on regardless of its idle state.

Selecting Aggressiveness for High Variability and Low SLA Tolerance

When workload variability is high but SLA tolerance is low, the Moderate setting is the recommended baseline. The selection logic is as follows:

• Conservative/Moderate: Prioritizes SLA by maintaining a larger resource buffer and requiring a longer period of sustained idleness before powering down. This reduces the frequency of "power-on spikes" where a sudden traffic burst forces a host to boot, causing latency during VM migrations.
• Aggressive: Prioritizes power savings. This is unsuitable for low-SLA-tolerance environments because it triggers power-downs more quickly, increasing the risk that a sudden workload spike will hit a cluster with insufficient active capacity, leading to CPU contention while a host boots.

Automated vs. Manual Mode Across vSphere Releases

The choice between automated and manual mode is driven by operational risk rather than version-specific defaults. While integration with DRS has matured across releases, the fundamental behavior remains: Automated mode is for environments with predictable growth and high confidence in their DRS rules. Manual mode is for environments where capacity planning is tied to strict licensing (per-powered-host) or where the boot time of physical hardware is too slow to meet recovery SLAs.

Verification Steps

To verify current eligibility and behavior, perform the following:

1. Check the cluster's DPM configuration via the vSphere Client to confirm the current aggressiveness level.
2. Review the vCenter Task Console for "Power Off Host" events to correlate power-downs with traffic dips.
3. Verify that no VMs on candidate hosts have "Local Storage" dependencies, as these block DPM eligibility.

Missing Diagnostic: To refine this recommendation, please provide the average boot-to-ready time for your physical hosts; if boot times exceed 5–10 minutes, Manual mode is strongly advised regardless of traffic levels.