Skip to content
English
  • There are no suggestions because the search field is empty.

Overcommitment monitors in the VMware MP

The nine overcommitment monitors in the VMware Management Pack, their thresholds and how they're calculated.

Applies to: OpsLogix VMware Management Pack (all versions) with the VMware Capacity Analysis Management Pack · SCOM 2019, 2022, 2025 · vCenter 7.0, 8.0, 9.0
Last updated: October 2026 · Reading time: 4 minutes

The overcommitment monitors show how much CPU, memory and storage your virtual machines are configured to use, compared with what the cluster physically has. They come with the OpsLogix IMP - VMware Capacity Analysis management pack and target VMware Cluster objects. The OpsLogix IMP - VMware Capacity Analysis Dashboards management pack adds dashboards for the same data, such as Cluster Commitment ALL Dashboard, in the Monitoring workspace under OpsLogix > VMware > Capacity Analysis Dashboards.

Monitors and defaults

There are nine monitors: CPU, Memory and Datastore, each for all VMs, for powered-on VMs and for powered-off VMs.

Monitor On by default
Cluster (ALL) CPU Overcommitment Monitor Yes
Cluster (ALL) Memory Overcommitment Monitor Yes
Cluster (ALL) Datastore Overcommitment Monitor Yes
Cluster (Powered On VMs) CPU / Memory / Datastore Overcommitment Monitor No
Cluster (Powered Off VMs) CPU / Memory / Datastore Overcommitment Monitor No

All monitors run every 5 minutes and use the same thresholds:

Commitment percentage State
90% or lower Healthy
Above 90% and below 120% Warning
120% or higher Critical

The monitors raise an alert (priority Medium) when the state is Warning or Critical, with a severity that matches the state, and close it automatically when the cluster is healthy again. The alert shows the cluster, its commitment percentage and the most overcommitted hosts. Nine performance collection rules collect the same percentages; all of them are on by default.

How the percentage is calculated

The management pack first calculates a percentage per ESXi host or datastore. 100% means the VMs are configured to use exactly what is physically there.

Resource Per host or datastore
CPU The number of virtual CPUs of the VMs on the host, divided by the number of physical CPU cores of the host. When hyperthreading is on, the cores count twice.
Memory The configured memory of the VMs on the host, divided by the physical memory of the host.
Datastore The provisioned storage of the VMs on the datastore (used space plus space that thin-provisioned disks can still grow into), divided by the capacity of the datastore.

The cluster percentage is the average of its hosts (CPU and memory) or of its datastores (datastore). A cluster can therefore look healthy while one host or datastore is heavily overcommitted. The alert lists the most overcommitted hosts to help you find them.

The (ALL) monitors count all VMs. The Powered On and Powered Off monitors count only VMs in that power state; suspended VMs count as powered off.

Tune the monitors

You can override these parameters on each monitor:

Parameter Default What it does
HealthyValue 90 Highest percentage that is still Healthy.
CriticalValue 120 Percentage from which the state is Critical.
ClusterFilter (.*) Regular expression on the cluster name. Only matching clusters are evaluated.
HostFilter (.*) Regular expression on the host name. Only matching hosts are counted.
DatastoreFilter (.*) Regular expression on the datastore name. Only matching datastores are counted.
VirtualMachineFilter (.*) Regular expression on the VM name. Only matching VMs are counted.
IntervalSeconds 300 How often the monitor runs.

The filters aren't case-sensitive and match anywhere in the name. For example, set VirtualMachineFilter to ^(?!TEMPLATE-) to leave out VMs whose name starts with TEMPLATE-.

To change a threshold or filter, go to Authoring > Management Pack Objects > Monitors, search for Overcommitment, right-click the monitor and select Overrides > Override the Monitor. Use the same filter values on all overcommitment monitors and rules: they share one data collection (cookdown), and different values add a separate collection for each combination.

See also