Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

Proactive HA: '<host> has an unknown health state' and VMs Moved Unexpectedly

Technical article
VMware

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

If vCenter lists "<host> has an unknown health state." under a Proactive HA cluster and virtual machines have moved between hosts without anyone asking, the usual cause is timing. Proactive HA starts every host in an unknown state. Until the hardware vendor's health provider has posted a first update for a host, DRS treats that host as quarantined and avoids it. If DRS rebalances before every host has reported, VMs move to the hosts that have.

Broadcom documents this as expected behaviour in KB 313410. The practical fix is to put DRS into manual mode before turning Proactive HA on or restarting vCenter services, and to return it to fully automated once every host has had a health update.

What the error means

Proactive HA reacts to a hardware component failing before the host itself fails. Broadcom's 8.0 documentation (Last Updated October 28, 2026) describes a Proactive HA failure as a host component failing and causing a loss of redundancy, with a power supply as the example. VMs can then be evacuated and the host is placed in Quarantine mode or Maintenance mode. The same page says the cluster must use vSphere DRS for Proactive HA failure monitoring to work.

The health information comes from a provider, normally a plugin from the server vendor. The same Configure Proactive HA page describes the quarantine remediation as one that "balances performance and availability, by avoiding the usage of partially degraded hosts". It also says the Automation Level setting decides whether quarantine, maintenance mode and VM migrations are recommendations or automatic, and that with automation on, VMs are migrated to healthy hosts and degraded hosts are put into quarantine or maintenance mode. KB 313410 adds that DRS treats unknown hosts as quarantined and avoids placing VMs on them.

KB 313410 describes what you see: after Proactive HA is enabled, or after vCenter services restart while it is enabled, hosts show a configuration issue in All Issues. The issue cannot be acknowledged or reset to green, and VMs may migrate. The cause given is that hosts start in an "unknown" health state until the provider sends its first updates.

Broadcom does not publish the wording of an event or alarm raised when a host is actually quarantined for a real hardware fault. Use the host's health state in the vSphere Client and the provider's own console.

Likely causes

Cause What you see How to confirm
Provider has not reported since Proactive HA was enabled or vCenter restarted "<host> has an unknown health state." in All Issues; VMs moved to other hosts The message appeared right after enabling Proactive HA or restarting vCenter services (KB 313410)
DRS rebalanced before all hosts reported VMs migrated off some hosts, onto hosts with a first update DRS was fully automated at the time; check the cluster's recent tasks and DRS history
An unmonitored host is in the cluster "Reconfigure cluster" fails with A specified parameter was not correct: InfraUpdateHaConfig.providers, or no Health Update provider appears The provider's unmonitored host list (KB 323322, below)
Two-host cluster Proactive HA cannot put a host into maintenance mode when a VM has a soft affinity rule Cluster has two hosts; KB 398787 says this is by design
Remediation setting does what you chose Hosts quarantined or placed in maintenance for a real, reported fault Cluster > Configure > vSphere Availability > Proactive HA settings

Prove the cause before changing anything

These checks are read-only. These commands have not been run or lab-tested.

  1. Read the issue text. In the vSphere Client, find the issue under All Issues, where KB 313410 says it appears. The exact text is "<host> has an unknown health state."
  2. Read the Proactive HA settings. The vSphere Client path in the 8.0 page (Last Updated October 28, 2026) is the cluster's Configure > vSphere Availability > Edit, with "Turn on Proactive HA" and then "Proactive HA Failures and Responses". The remediation choices are "Quarantine mode for all failures", "Quarantine mode for moderate and Maintenance mode for severe failure (Mixed)" and "Maintenance mode for all failures". The page also says it is editable only if vSphere DRS is enabled.
  3. Check the timing. Was Proactive HA enabled, or vCenter restarted, shortly before the migrations? Compare the time of the first "unknown health state" message with the migration tasks. If the migrations follow within a DRS cycle, you are looking at KB 313410.
  4. Check the DRS automation level. In the cluster's vSphere DRS settings, read the automation level. Fully automated means DRS acts on its own.
  5. If the reconfigure task failed, list the unmonitored hosts. KB 323322 describes this with the Managed Object Browser (MOB). Sign in to https://<vCenter>/mob with SSO administrator credentials; open Content, then the datacenter entry (group-d1), then the datacenter and its host folder (group-h); read the cluster ID (of the form domain-cxx) from childEntity. Back on Content, open HealthUpdateManager, run QueryProviderList and note the provider UUID, then run QueryUnmonitoredHosts with that UUID as providerId and the cluster ID in the cluster field, once for each provider. A host returned here is one the provider is not monitoring.

Example (illustrative): cluster prod-a has hosts esx01 to esx04, DRS fully automated. After a vc01 service restart, All Issues lists "esx03 has an unknown health state." and "esx04 has an unknown health state." Within minutes DRS moves a dozen VMs onto esx01 and esx02. The provider's console shows esx03 and esx04 healthy, but vCenter has not yet received updates for them. This matches KB 313410. No hardware has failed.

Fix

Stop the unwanted moves

Per KB 313410, set DRS to manual before enabling Proactive HA or restarting vCenter services, and return it to fully automated once every host has received at least one health update.

Risk: in manual mode DRS no longer balances the cluster by itself. Put it back to fully automated afterwards; if you forget, load imbalance builds up.

Clear the message

KB 313410 says to disable and re-enable Proactive HA on the cluster to clear the "unknown health state" message. Disruptive step: turning Proactive HA off removes protection for hardware faults until it is back on. Set DRS to manual first, as above, so that the re-enable does not trigger another round of moves.

Provider will not report

KB 313410 describes the steps for the Cisco Intersight plugin: set DRS to manual, toggle the Intersight provider off and on under the cluster's Proactive HA Providers tab, and watch whether host health updates. If that does not work, unregister the provider in the Intersight plugin's Proactive HA Registration tab, re-register it and confirm the alarms map correctly, then return DRS to fully automated after the unknown state clears. Other vendors have their own providers; use that vendor's guide for the registration steps.

Reconfigure fails with InfraUpdateHaConfig.providers

Per KB 323322, the cause is an unmonitored host in the cluster, and the fix is to remove the unmonitored host or hosts. Use the MOB procedure above to identify them.

Disruptive step: removing a host from a cluster takes its capacity out of the cluster and means its VMs must be moved first. Do it in a planned window, and check HA admission control still holds without that host.

Two-host cluster

KB 398787 states that in a two-node cluster, a VM with a soft affinity rule cannot be moved off the host, so Proactive HA cannot automatically put the host into maintenance mode. The page says this is by design and that there is no resolution other than adding a third host.

Confirm the fix

  • "has an unknown health state" no longer appears under All Issues for any host.
  • Each host shows a health state reported by the provider.
  • DRS is back to fully automated, and a few DRS cycles pass without unexpected migrations.
  • The Proactive HA remediation setting is the one you chose.

What to watch afterwards

  • Make "DRS to manual" a step in your procedure for vCenter service restarts and upgrades on any cluster with Proactive HA.
  • After a restart, check All Issues once the providers have had time to report, before returning DRS to automated.
  • Recheck the remediation setting if quarantine or maintenance mode moves more VMs than you expect.
  • Make sure every host in the cluster is covered by a provider.

Risks and limits

  • Broadcom does not document the wording of any event raised when a host is quarantined for a genuine fault; the only documented message text is the "unknown health state" wording.
  • No vendor-documented command to read a host's quarantine state was found; the vSphere Client is the documented place.
  • KB 313410 gives steps for one provider (Cisco Intersight). Other providers differ.
  • The MOB procedure in KB 323322 uses an internal browser interface that exposes vCenter objects. Use read-only methods only (QueryProviderList, QueryUnmonitoredHosts) and do not call other methods from it.
  • Version scope is as each Broadcom KB page states it: vCenter Server 7.x and 8.x for KB 313410 and KB 323322, and ESXi 7.x and 8.x for KB 398787.

NetLeverage runs its own hosting on VMware vSphere, vCenter and ESXi. Talk to an engineer.

Sources