Determining the actual cluster resource capacity when the cluster is in an overcommitted state

Products

• VMware vCenter Server
• VMware vSphere ESXi

Issue/Introduction

When a vSphere cluster is in an overcommitted state (for example, due to ESXi hosts entering Maintenance Mode), the Total Reservation Capacity displayed in the vSphere Client might significantly exceed the Cluster Total Capacity.

For details on why the Total Reservation Capacity displays a value larger than the Cluster Total Capacity in this state, please refer to the following Knowledge Base article: The Total Reservation Capacity value is larger than the Cluster Total Capacity value in cluster Resource Reservation page in Web Client(313895)

Since the displayed Total Reservation Capacity does not reflect the physical limit in this state, administrators cannot rely on this value to determine how much resource capacity is actually available. This article provides methods to determine the actual resource limit of the cluster when it is overcommitted.

Environment

VMware vCenter Server
VMware vSphere ESXi

Cause

The cluster is in an overcommitted state (internally known as the YELLOW state). In this state, the calculation logic for available resources changes, causing the vSphere Client to display a calculated value that ignores the physical boundaries of the root resource pool.

Resolution

To ascertain the valid physical resource limit of the cluster while in an overcommitted state, use one of the following methods.

Method 1: Resolve the overcommitted state The most straightforward way to view the correct capacity is to return the cluster to a normal (undercommitted) state.

1. Temporarily reduce the configured reservations on Resource Pools or Virtual Machines, or exit Maintenance Mode on ESXi hosts to restore the cluster's physical capacity.

2. Once the warnings clear, check the Monitor > Resource Allocation tab in the vSphere Client.

3. The Total Reservation Capacity and Available Reservation values will now accurately reflect the physical limits (Cluster Total Capacity minus system overhead).

Method 2: Verify capacity via vCenter Server logs If you cannot immediately change the configuration to resolve the overcommitted state, you can verify the actual root capacity recognized by the system by checking the vpxd logs.

1. Increase the vCenter Server logging level to verbose.

2. Connect to the vCenter Server Appliance via SSH.

3. Search for the string root memory capacity (or root cpu capacity) in the vpxd.log file.

Example command: grep "root memory capacity" /var/log/vmware/vpxd/vpxd.log

Sample output: YYYY-MM-DDTHH:MM:SS verbose vpxd[xxxxx] [Originator@6876 sub=ResMgr] root memory capacity = 256761659392, root memory allocated = 272407461888 => memory is OVERCOMMITTED

In the example above:

• root memory capacity: Represents the actual available capacity in bytes.

• root memory allocated: Represents the total reservations currently requested by Virtual Machines and Resource Pools.

You can convert the capacity value to GB (bytes / 1024 / 1024 / 1024) to determine the actual limit.

Additional Information

Japanese version: クラスタがオーバーコミット状態のときに実際のクラスタリソースキャパシティを確認する方法(426507)