VMware ESXi and vSphere Cluster Management

Configure VMware DRS Automation Levels and Migration Thresholds

Learn how VMware vSphere DRS automation modes control VM power-on placement, vMotion migrations, administrator approval, and migration thresholds.

VMware vSphere Distributed Resource Scheduler (DRS) is a vCenter Server capability that evaluates compute-resource use across a group of ESXi hosts called a DRS cluster. DRS helps decide where virtual machines should run and whether workloads should move between hosts to improve balance.

DRS makes placement decisions at two important points:

  • Initial placement: selecting an ESXi host when a VM is powered on.
  • Load balancing: evaluating running workloads and proposing or performing VM migrations when host utilization becomes uneven.

The cluster's automation level determines whether DRS only recommends these actions or performs some of them automatically. The three modes are Manual, Partially Automated, and Fully Automated.

Compare DRS Automation Levels

Automation level | VM power-on placement | Response to imbalance | Administrator approval required | Typical use case

Manual | DRS proposes a destination host | DRS generates migration recommendations | Yes, for placement and migrations | Strictly controlled production environments

Partially Automated | DRS selects a host automatically | DRS generates migration recommendations | Yes, for later balancing migrations | Automatic startup placement with formal approval for vMotion

Fully Automated | DRS selects a host automatically | DRS can initiate qualifying migrations | No, for qualifying DRS actions | Large or dynamic clusters with mature monitoring

Manual DRS Automation

In Manual mode, DRS provides recommendations but does not independently change VM placement. This gives the administrator direct control over both startup placement and later migrations.

When a VM powers on

DRS evaluates the cluster and proposes an appropriate destination host. The VM is not placed by DRS until an administrator accepts the recommendation or otherwise selects a host.

When the cluster becomes imbalanced

DRS creates migration recommendations. A migration recommendation is a proposed VM move intended to satisfy a cluster constraint or improve resource balance. An administrator must review and accept the recommendation before the vMotion migration occurs.

Manual mode is appropriate when every placement and workload move must pass through an explicit operational change process. It can also be useful while validating rules, host capacity, application behavior, or vMotion readiness.

Partially Automated DRS

Partially Automated mode is a hybrid policy. DRS automatically selects a suitable host when a VM is powered on, but it does not automatically perform later load-balancing migrations.

  • At power-on: DRS chooses the initial host according to available resources and applicable cluster rules.
  • During rebalancing: DRS presents migration recommendations for administrator review.
  • Before a vMotion: an administrator must accept the recommendation.

This mode combines reliable initial placement with human approval for ongoing workload movement. It is useful when administrators want DRS to avoid poor startup placement but still require formal approval before moving running applications.

Fully Automated DRS

In Fully Automated mode, DRS handles both major placement decisions automatically:

  • It selects an initial host when a VM powers on.
  • It can initiate qualifying VM migrations without waiting for administrator approval.

When DRS identifies a valid opportunity, it can move workloads from more heavily used hosts toward less heavily used hosts. DRS applies the configured migration threshold and considers resource availability, placement rules, host compatibility, and other constraints before acting.

The operational benefit is fast, continuous balancing as demand changes. The trade-off is reduced explicit change control: workloads may move without an administrator approving each vMotion. Before enabling this mode, evaluate application sensitivity, migration capacity, monitoring, and the organization's change policy.

Migration Threshold and Recommendation Priority

The migration threshold controls how readily Fully Automated DRS applies migration recommendations. A recommendation priority ranks the importance or expected value of a proposed migration.

Lower threshold settings are conservative. They allow only higher-priority actions, which usually means fewer migrations and a larger expected benefit per migration. Higher settings are more aggressive and allow progressively smaller balancing improvements, potentially increasing migration frequency.

Level 3 was the historical default in the traditional five-level DRS threshold model. Confirm the behavior and interface in the vSphere release you administer because DRS algorithms and client controls can differ between releases.

DRS Migration Threshold Levels

Threshold level | Label | Recommendation priorities applied | Expected balancing benefit | Migration frequency tendency

Level 1 | Conservative | Priority 1 only | Mandatory or very important actions, such as rule compliance or maintenance-related constraints | Lowest

Level 2 | — | Priorities 1 and 2 | Includes substantial balancing gains | Low

Level 3 | — | Priorities 1 through 3 | Includes actions with a solid expected improvement; historical default | Moderate

Level 4 | — | Priorities 1 through 4 | Includes moderate expected improvement | High

Level 5 | Aggressive | Priorities 1 through 5 | Permits all qualifying recommendations, including minor predicted improvements | Highest

How to interpret the levels

  • Level 1: Choose this when automatic movement should be limited to mandatory cluster requirements, rule compliance, or conditions associated with host maintenance.
  • Level 2: Use this when only substantial balancing gains should justify automatic migration.
  • Level 3: Use this as a balanced starting point when the cluster can tolerate normal DRS movement and the historical default is appropriate for the deployed release.
  • Level 4: Select this when the cluster has sustained imbalance and can support more frequent vMotion activity.
  • Level 5: Select this when the organization accepts more workload movement in exchange for pursuing smaller optimization opportunities.

Configure the Automation Level in the vSphere Client

  1. Select the target cluster in vCenter Server.
  2. Open the cluster configuration area for vSphere DRS.
  3. Enable DRS if it is not already enabled.
  4. Choose Manual, Partially Automated, or Fully Automated as the cluster automation policy.
  5. If Fully Automated is selected, set the migration threshold from Level 1 through Level 5.
  6. Review DRS recommendations, cluster rules, events, and VM-specific settings after changing the policy.

For an example of the configuration target, see Set DRS Automation Level.

Configuration Checks Before Increasing Automation

  • Ensure vMotion is configured and operational before expecting automatic live migrations.
  • Confirm that hosts have compatible CPU capabilities. Configure Enhanced vMotion Compatibility (EVC) where required to support vMotion between hosts with different CPU generations.
  • Review VM/host affinity and anti-affinity rules. An affinity rule keeps selected VMs or VM/host combinations together or apart and can restrict eligible destinations.
  • Verify management, vMotion, and storage networking capacity for the expected migration volume.
  • Check for per-VM DRS automation overrides. A VM-specific setting can differ from the cluster default.
  • Review application performance and migration history before selecting a more aggressive threshold.

Choosing an Automation Level

Operational requirement | Recommended starting point | Reason

Every change requires explicit approval | Manual | Administrators review startup placement and each migration.

Automatic startup placement, approved workload movement | Partially Automated | DRS chooses an initial host while administrators approve balancing vMotion operations.

Continuous balancing in a mature, monitored cluster | Fully Automated, often Level 3 initially | DRS can respond automatically while avoiding every marginal optimization opportunity.

Migration traffic and application movement must be minimized | Fully Automated with Level 1 or Level 2 | Only mandatory or higher-value recommendations are applied automatically.

Frequent sustained imbalance with sufficient vMotion capacity | Fully Automated with Level 4 or Level 5 | More recommendations, including modest improvements, are allowed.

Before increasing automation, assess application performance, vMotion bandwidth, storage and network readiness, host capacity, migration frequency, and change-control requirements. A higher threshold is not automatically better: it can improve balance while also creating more workload movement.

Practical Configuration Examples

Strictly controlled production environment

Use Manual DRS when the operations team must review every startup destination and every suggested migration before making a change.

Automatic placement with controlled migrations

Use Partially Automated DRS when the team wants reliable power-on placement but requires formal approval before moving running workloads.

Dynamic large cluster

Use Fully Automated DRS with Level 3 as a measured starting point when demand changes continuously and monitoring is mature. DRS can respond to meaningful imbalance without acting on every marginal opportunity.

Minimize migration traffic

Use Fully Automated with Level 1 or Level 2 when migration traffic and application movement should be minimized. These settings reserve automatic actions for mandatory or high-value recommendations.

Persistent imbalance with migration capacity

Use Fully Automated with Level 4 or Level 5 when the cluster has sustained imbalance, compatible hosts, sufficient vMotion capacity, and an operational policy that accepts more frequent moves.

Troubleshooting DRS Automation

DRS displays recommendations but does not move VMs automatically

Likely causes include Manual or Partially Automated cluster mode, a recommendation priority below the configured Fully Automated threshold, or a VM-specific automation override.

  • Confirm the cluster automation level.
  • Inspect the migration threshold and recommendation priority.
  • Review VM-specific DRS settings and overrides.

DRS cannot rebalance a cluster even though one host appears heavily loaded

High apparent utilization does not guarantee that a move is possible. Affinity or anti-affinity rules may limit destinations, eligible hosts may lack CPU or memory capacity, host CPU compatibility may prevent vMotion, or required networking and storage connectivity may be unavailable.

  • Review DRS rules and constraint violations.
  • Check CPU and memory headroom on eligible hosts.
  • Validate EVC and CPU compatibility.
  • Test vMotion prerequisites and network paths.

The cluster performs more migrations than expected

A high threshold can permit low-priority recommendations. Fluctuating demand can also change DRS calculations repeatedly, while resource contention or poorly designed policy constraints can cause corrective moves.

  • Lower the threshold after assessing workload behavior.
  • Review DRS migration history and recommendation priorities.
  • Investigate noisy workloads, host-capacity imbalance, and affinity or anti-affinity rule design.

A VM is not placed on the preferred host at power-on

DRS may have selected another host because it offered better available resources or better compliance with applicable rules. The preferred host may lack capacity, or placement rules and VM-specific settings may restrict the choice.

  • Review DRS placement reasoning and cluster events.
  • Check host utilization and rule compliance.
  • Confirm whether the cluster or VM uses an automation override.

Exam-Relevant Notes

  • Manual: DRS recommends both initial placement and migrations; the administrator applies them.
  • Partially Automated: DRS performs initial placement automatically, but later balancing migrations require approval.
  • Fully Automated: DRS performs initial placement and qualifying balancing migrations automatically.
  • The migration threshold determines which recommendation priorities Fully Automated DRS may apply.
  • Level 1 accepts only priority 1 actions; Level 5 accepts priorities 1 through 5.
  • Level 3 is the historical default for the five-level threshold model, but release-specific behavior should be verified.
  • DRS balancing is constrained by rules, host capacity, CPU compatibility, vMotion configuration, and available network and storage paths.
  • DRS and vSphere HA have different purposes: DRS focuses on resource placement and balancing, while HA focuses on restarting VMs after host failures.

Summary

DRS evaluates resource balance in a vSphere cluster and makes decisions during VM power-on and while workloads are running. Manual mode leaves both decisions to administrators after presenting recommendations. Partially Automated mode automates initial placement but requires approval for rebalancing migrations. Fully Automated mode performs both functions automatically when recommendations meet the configured policy.

The migration threshold controls how many recommendation priorities Fully Automated DRS can apply. Lower levels favor fewer, higher-value moves; higher levels allow more frequent migrations and smaller predicted improvements. Select the automation level and threshold according to workload sensitivity, change-control policy, vMotion readiness, cluster capacity, and the behavior of the vSphere release in use.