VMware vSphere Clustering Service (vCLS) was introduced in the vSphere 7 Update 1 release with the idea of creating a decoupled and distributed control plane for clustering services in vSphere.
In the first release, a subset of DRS capabilities was using the new vCLS feature and starting with vCenter 7.0.1, VMware vSphere Distributed Resource Scheduler (vSphere DRS) cannot function if vSphere Cluster Services VMs are not present in each vSphere cluster.
But with VMware vSphere 8.0.3, the vSphere Clustering Service (vCLS) will be “migrated” with a new Embedded vCLS model. The entire vSphere Cluster Service (vCLS) is rearchitected to use fewer resources, remove storage footprint, and eliminate issues associated with vCLS deployment.
The ESXi host spins up the Embedded vCLS VM(s) directly. There is no OVF deployment pushed from vCenter and EAM (ESX Agent Manager) is no longer involved.
The number of resources has also reduced: only two VMs when using Embedded vCLS, no storage footprint and run entirely in host memory.
The new Embedded vCLS uses the CRX runtime technology first developed for the vSphere IaaS control plane (formerly vSphere with Tanzu).
Embedded vCLS VMs are system managed and typical VM operations are disallowed. Note that vSphere vMotion and Storage vMotion are not supported when using Embedded vCLS (they where not recommend with previous implementation). VM management is totally automatically: when placing a host currently running an Embedded vCLS VM into maintenance mode, the host will automatically spin-down the Embedded vCLS VM and another host in the cluster will automatically spin-up a new Embedded vCLS VM. Only power-off and hard stop operations are permitted for troubleshooting purposes.
The original version of vCLS will be referred to as External vCLS, as opposite with the Embedded version.
You can check the cluster type by looking at the cluster properites:
Embedded vCLS is automatically activated on vSphere clusters that contain at least one ESXi 8 Update 3 host and is managed by vCenter 8 Update 3. Embedded vCLS will only revert to vCLS if there are no ESXi 8 Update 3 or later hosts available to run virtual machines in the cluster
Like with the previous implementation, anti-affinity with vCLS VMs, if needed, is achieved using a Compute Policy. And Retreat Mode is still available for troubleshooting purposes.
Other products and solutions that have defined business logic around External vCLS might not work with Embedded vCLS. See individual product documentation to understand the interoperability support and impact. For more information, see vSphere Cluster Services and content available from vSphere Technical Marketing on the Broadcom Support Portal.
For more information on Embedded vSphere Cluster Service, see these articles:
