Overcommitment

Overview

Operations such as migrations, demand uplifts and configuration changes can leave hosts or clusters under provisioned and therefore unable to serve all the workloads you want it to serve. The aggregate utilisation that is larger than the total capacity that can be served is known as overcommit.

Where can I see overcommitment?

Overcommitment is reported in Capacity Plannerin four places, each in a slightly different way:

  • Operations panel
  • Operations list
  • Timeline
  • Watch list

Operations panel

The Operations configuration panel is the main location for all information about the total overcommitment in your Forward Thinking scenario. It lists the VMs, datastores, and hosts where the capacity demand is breached, along with details of the total CPU demand, memory demand, storage, and Disk IO demand of all overcommitted VMs.

To see the panel, follow these steps:

  • From the configuration panel list on the left, click Operations.
    The Operations & Overcommitment panel opens.

All operations list

The All Operations list details every operation you perform in your Forward Thinking scenario model. Numbers in the Successful, No Action and Watchlist columns indicate how many files were successfully moved, how many caused overcommitment, and for how many files no action was taken. Clicking the numbers in the columns opens a new dialog window with more details about each operation.

To view the operations list, follow these steps:

  1. From the configuration panel list on the left, click Operations.
    The Operations & Overcommitment panel opens.
  2. Click Operations to see the list of all operations, their statuses and other details.

Timeline

Warning and error indicators on the timeline indicate operations that have resulted in some overcommitment. Click the indicator to see that day's list of operations, and which among them created the demand breach. The list looks the same as the All Operations list but it only shows the operations for that selected day.

For more information, see Timeline.

Watch list

Any server where utilisation is above the configured capacity thresholds appears in the watch list. If, as a result of an operation, a server becomes overcommitted then it is placed on the watch list. The metric that is breaching its threshold is shown there.

Remove entities from the overcommitment list

Hosts are removed from the overcommitment list as soon as there is enough headroom to serve the total projected utilisation at that point in the timeline. VMs are removed from the overcommit list when their configuration is changed to ensure it has sufficient capacity. This is often resolved by using the right-sizing operation.

For example, after a migration, a collection of hosts in a cluster may be overcommitted. It is likely that the cluster itself is overcommitted as well. Observing the watch list, you can notice that all hosts in that cluster are now in the watch list. The quickest way to resolve this is to use host templates to create a new host in that cluster and then distribute workloads across the cluster to optimise use of the available hardware. You may need to add further hosts in the cluster if one is not sufficient.

Overcommitment in scenario reports

Scenario reports contain a subsection called Storage Overcommitment within the operations section. A summary of the total overcommitment is given along with details of each overcommitted VM, including its name and original host.

For more information, see Scenario Reports.