VMware ESXi and vSphere Cluster Management
Set the VMware DRS Automation Level
Learn how Manual, Partially Automated, and Fully Automated VMware DRS levels control VM placement, rebalancing, vMotion, and migration thresholds.
VMware vSphere Distributed Resource Scheduler (DRS) helps a cluster place virtual machines (VMs) on suitable ESXi hosts and maintain a useful balance of resource demand. The DRS automation level determines whether vCenter Server only suggests an action or performs it automatically.
This setting affects two separate decision points:
- Initial placement: selecting an ESXi host when a VM is powered on.
- Ongoing load balancing: moving running VMs when resource demand becomes uneven across the cluster.
A DRS recommendation is a proposed placement or migration action. A recommendation is not the same as an executed migration. Depending on the automation level, an administrator may need to review and apply the recommendation, or DRS may perform the action automatically.
How DRS automation works
A DRS cluster is a group of ESXi hosts managed by vCenter Server as a shared resource pool with DRS enabled. DRS evaluates host suitability, resource demand, reservations, placement constraints, and cluster balance. It can identify a better host for a newly powered-on VM or identify a migration that would improve the distribution of workloads.
DRS does not treat every possible move as equally valuable. It assigns recommendation priorities based on factors such as constraints that must be satisfied and the expected improvement in balance. The selected automation level controls who acts on those recommendations.
DRS automation levels
| Automation level | VM power-on placement | Response to cluster imbalance | Administrator involvement | Typical use case |
|---|---|---|---|---|
| Manual | DRS recommends suitable destination hosts. | DRS presents migration recommendations. | The administrator reviews and applies placement or migration recommendations. | Tightly controlled environments where every move requires approval. |
| Partially Automated | DRS automatically selects a suitable host when a VM powers on. | DRS presents migration recommendations but does not perform rebalance migrations automatically. | The administrator reviews and applies migration recommendations. | Environments that want automatic startup placement but controlled rebalancing. |
| Fully Automated | DRS automatically selects a suitable host when a VM powers on. | DRS automatically migrates eligible running VMs to improve cluster balance, normally using vMotion. | The administrator defines policy and monitors the resulting actions. | Mature clusters where automated balancing is acceptable. |
Manual
In Manual mode, DRS recommends a suitable host when a VM is powered on. The administrator chooses whether and where to place the VM from the available recommendation.
When the cluster becomes unbalanced, DRS presents migration recommendations instead of moving VMs. An administrator reviews the expected effect, operational timing, rules, and application impact, then applies selected recommendations.
For example, a production team may use Manual mode when application ownership, a maintenance window, or a current operational event must be considered before any workload moves.
Partially Automated
In Partially Automated mode, DRS automatically chooses an appropriate host for a VM at power-on. This removes routine placement decisions while preserving administrator control over live migrations.
When host utilization becomes uneven, DRS creates migration recommendations. It does not automatically perform those rebalance migrations. The operations team reviews and applies recommendations during an approved period.
Fully Automated
In Fully Automated mode, DRS automatically chooses a destination host when a VM powers on and automatically performs eligible migrations that improve cluster balance.
The normal mechanism for moving a running VM between compatible ESXi hosts is vMotion. vMotion transfers a running VM with minimal interruption, but it still depends on appropriate networking, storage access, CPU compatibility, permissions, licensing, and placement constraints.
Use Fully Automated mode only when you have confidence in the cluster design and operating procedures. Validate vMotion networking, shared resources, host compatibility, EVC, affinity rules, maintenance processes, and workload behavior before allowing DRS to initiate migrations without approval.
Migration threshold for Fully Automated DRS
The migration threshold is the sensitivity setting used by Fully Automated DRS. It determines how readily DRS acts on its recommendations by selecting which recommendation priority levels can be executed automatically.
A lower threshold requires stronger justification for a migration. A higher threshold permits migrations for smaller projected gains. Raising the threshold can make the cluster respond more readily to imbalance, but it can also increase vMotion activity, migration overhead, and operational change.
Level 3 is the default threshold. The exact wording and presentation can vary between vSphere releases, but the five-level behavior is commonly described as follows:
| Threshold level | Label | Recommendation priorities applied | Expected balance-improvement threshold | Operational implication |
|---|---|---|---|---|
| 1 | Conservative | Priority 1 only | Generally actions required to meet cluster constraints, such as affinity requirements or host maintenance. | Minimizes nonessential migrations. |
| 2 | — | Priorities 1 and 2 | Includes changes expected to provide a substantial balance improvement. | Allows important improvements while remaining cautious. |
| 3 | Default | Priorities 1 through 3 | Includes changes expected to provide a good balance improvement. | Provides a general-purpose balance between responsiveness and migration activity. |
| 4 | — | Priorities 1 through 4 | Includes changes expected to provide a moderate balance improvement. | Permits more frequent balancing actions. |
| 5 | Aggressive | All recommendation priorities | Includes moves with only a small expected improvement. | Can generate substantial migration activity; use only when that behavior is acceptable. |
Threshold levels apply to Fully Automated DRS. In Manual and Partially Automated modes, DRS may still produce recommendations, but the administrator remains responsible for applying rebalance migrations.
Choosing an automation level
| Operational requirement | Suggested automation level | Suggested threshold where applicable | Reasoning |
|---|---|---|---|
| Every placement and migration must be reviewed by an administrator. | Manual | Not applicable to automatic execution | Provides maximum administrative control and visibility. |
| Automatic startup placement is useful, but live migrations need approval. | Partially Automated | Not applicable to automatic execution | Removes routine power-on placement work without permitting automatic rebalance moves. |
| The cluster is mature, compatible, and designed for automated balancing. | Fully Automated | Usually start at Level 3 | Allows DRS to respond without waiting for an administrator. |
| Avoiding nonessential movement is more important than fine-grained balance. | Fully Automated | Level 1 or 2 | Restricts automatic actions to higher-priority recommendations. |
| Small balance improvements justify additional vMotion activity. | Fully Automated | Level 4 or 5 only after validation | Permits more recommendations to be executed automatically, with greater migration overhead. |
Choose Manual when change control, observation, or workload-specific decisions are required. Choose Partially Automated when automatic power-on placement is desirable but migration approval must remain with operations. Choose Fully Automated when the cluster is operationally mature and automated balancing fits the organization’s risk model.
Before raising the threshold, measure vMotion capacity, application sensitivity, host utilization patterns, and the frequency of DRS recommendations. Validate behavior in a controlled environment before using aggressive automation in production.
Configure the DRS automation level in vCenter Server
- Sign in to vCenter Server and select the target vSphere cluster.
- Open the cluster’s configuration area and locate the vSphere DRS settings.
- Edit the DRS configuration. Confirm that DRS is enabled for the cluster.
- Select Manual, Partially Automated, or Fully Automated.
- If you selected Fully Automated, locate the migration threshold and choose Level 1 through Level 5.
- Save the configuration.
- Return to the effective cluster settings and verify the selected automation level and threshold.
- Review DRS recommendations, recent tasks, and events after the change.
Validate behavior after changing the setting
- Review current DRS recommendations and confirm that their priorities match your selected policy.
- Check recent tasks and events for placement actions, recommendation applications, and automated vMotion activity.
- Test initial placement with a suitable noncritical VM or workload. Confirm whether DRS recommends placement, performs placement automatically, or waits for administrator action as expected.
- After enabling Fully Automated mode, monitor vMotion activity, host CPU and memory utilization, cluster balance, and application performance.
- Confirm that DRS actions respect VM-VM and VM-host affinity or anti-affinity rules.
- Review actions taken during host maintenance and verify that workload evacuation matches operational expectations. For background, see ESXi maintenance mode.
Troubleshooting DRS automation
DRS shows recommendations but does not move VMs automatically
- Confirm that the cluster is not configured for Manual or Partially Automated mode.
- If Fully Automated is enabled, confirm that the migration threshold includes the recommendation’s priority.
- Review recommendation details and related events for compatibility, rule, resource, or maintenance constraints.
A powered-on VM is not placed as expected
- In Manual mode, the administrator must choose from the placement recommendations.
- Check host connectivity, available CPU and memory, reservations, and current demand.
- Review VM-host and VM-VM affinity or anti-affinity rules and confirm that DRS prerequisites are complete.
Fully Automated DRS performs more migrations than expected
- Check whether the threshold is Level 4 or Level 5.
- Review DRS events and the priorities of executed recommendations.
- Assess recurring workload patterns. If migration overhead is too high, test a less aggressive threshold such as Level 2 or Level 3.
DRS cannot complete a recommended migration
- Review the migration task error and the DRS recommendation details.
- Validate vMotion networking, destination storage access, CPU compatibility, and applicable licensing requirements.
- Review EVC status and confirm that it provides the CPU compatibility required by the migration path. See CPU compatibility for vMotion and EVC requirements.
- Check affinity rules and other cluster constraints that may prevent an eligible destination.
Exam-relevant distinctions
- Manual: DRS recommends both power-on placement and rebalance migrations; the administrator applies them.
- Partially Automated: DRS automatically handles power-on placement but only recommends rebalance migrations.
- Fully Automated: DRS automatically handles power-on placement and eligible rebalance migrations, normally through vMotion.
- Migration threshold: controls which priority levels Fully Automated DRS executes automatically.
- Level 3: the default migration threshold.
- Level 1: conservative and limited to priority 1 recommendations.
- Level 5: aggressive and permits all recommendation priorities.
The automation level is therefore a policy choice about control versus automation, while the migration threshold is a sensitivity choice within Fully Automated DRS. Configure both at the cluster in vCenter Server, then verify actual recommendations and tasks rather than assuming that a saved setting guarantees every proposed migration can proceed.