vSphere High Availability (HA) ensures availability of the virtual machines in your SDDC. If a host fails, vSphere HA restarts its VMs on a different host. All clusters in a VMware Cloud on AWS SDDC are configured to use vSphere HA. This setting cannot be reconfigured.
To ensure availability of all workload and management VMs in your SDDC,
VMware Cloud on AWS must maintain sufficient capacity to power them on in the event of host failure. HA admission control is the primary mechanism for capacity maintenance. Admission control imposes constraints on resource usage, and can prevent any action that consumes more resources than the cluster can support during a failover. These constraints apply to actions like powering on or migrating a VM, or reserving additional CPU or memory resources for a VM, and effectively limit the availability of host resources as shown here:
- In a two-host i3.metal SDDC cluster, admission control limits the number of workload VMs that you can power on. These limits vary depending on the SDDC software version used:
- For SDDC version 1.22 and later, admission control prevents you from powering on more than 25 VMs or assigning more than 806 MHz CPU reservation to a single VM.
- For SDDC version 1.18, admission control prevents you from powering on more than 35 VMs or assigning more than 1152 MHz CPU reservation to a single VM.
- In SDDC clusters with three to five hosts, admission control reserves one host for failover.
For a detailed discussion of vSphere HA and other features that ensure vSphere availability, see How vSphere HA Works in the VMware vSphere Product Documentation. Although most HA settings in VMware Cloud on AWS are managed for you by VMware and cannot be changed in your SDDC's vCenter, it's important to understand the fundamental concepts of vSphere HA and how they apply to workload deployment in your SDDC.
