With vSphere 7 U1, VMware introduced vSphere Clustering Service (vCLS), which helps in a situation when vCenter Server becomes unavailable. As you know, if you lose vCenter Server, you also lose the Distributed Resource Scheduler (DRS), so your VMs are no longer balanced across your cluster.

While this might be OK for smaller, more static environments, larger infrastructures that rely on the DRS to optimize the loads can struggle or even become very unbalanced with unresponsive VMs. Note that this can only occur if the vCenter is lost/unavailable.

You can still increase the availability of vCenter Server via vCenter High Availability (VCHA), which was introduced in vSphere 6.0. However, the architecture is pretty heavy and resource hungry since you create additional nodes (Passive and Witness) that consume valuable resources.

Additionally, when patching VCSA, you have to follow the exact order for patching, which complicates the workflows.

vCLS Architecture ^

The architecture of vCLS comprises small footprint VMs running on each ESXi host in the cluster. You don't have to worry about those VMs at all. No shutdown, no backups.

Those VMs are also called Agent VMs and form a cluster quorum. The management is assured by the ESXi Agent manager. It's the same mechanism that manages agent VMs for HA. When you activate HA, the Agent VMs are deployed on the host and activated. It's those VMs that communicate and maintain the vSphere HA within a cluster.

For vCLS, this is the same principle. You activate DRS on your cluster, and vCLS VMs are automatically deployed and started. Absolutely no configuration from your end is needed.

VMware vSphere Cluster Service architecture

VMware vSphere Cluster Service architecture

When you look at the vSphere client and click the view where you can see all the VMs, you'll find there is a new folder created called vCLS that contains the vCLS VMs. You should not rename the vCLS folder or rename the vCLS VM(s).

vSphere Cluster VMs

vSphere Cluster VMs

When the new DRS is freshly enabled, the cluster will not be available until the first vCLS VM is deployed and powered on in that cluster. The DRS service is strictly dependent on the vCLS starting vSphere 7 U1.

The vSphere Cluster Service VMs are managed by vSphere Cluster Services, which maintain the resources, power state, and availability of these VMs. These VMs are required for maintaining the health of vSphere Cluster Services. Any impact on the power state or resources of these VMs could impact the DRS functionality of the cluster.

Host maintenance mode ^

What happens when the host is in maintenance mode? When patching the host or upgrading its hardware, you'll need to put the host into maintenance mode before shutting down. In this case, the vCLS agent VMs are migrated by vMotion to other hosts within the cluster.

Maintenance mode of host 2

Maintenance mode of host 2

Let's have a look. I put host 2 (esxi02) in maintenance mode. I receive a prompt about evacuating VMs to other hosts, which I validate. Then, when I check for vCLS VMs, I see that I have two vCLS VMs running on host 3 (esxi03). So yes, the system takes care of it. Once you exit maintenance mode, the VMs return automatically to the same host, as with other infrastructure VMs.

How many VMs are deployed? ^

In my example, I have a lab cluster with three hosts, and the system deploys three vCLS VMs. This is a hard maximum. If you have a cluster with, let's say, 12 hosts, only three vCLS VMs will be deployed in total. They use the cluster's shared storage so vMotion can be leveraged if a particular VM needs to be vMotioned.

What happens if I accidentally delete vCLS VM? ^

Well, it's simple. The VM is automatically recreated by the Agent Manager and powered on. The system is clever enough to maintain itself in case of human error. If you power off the agent VM, the system detects it and powers it on automatically as well.

Host 2 in maintenance mode and vCLS migrated to another host

Host 2 in maintenance mode and vCLS migrated to another host

Is there any resource consumption of those VMs?

No, not really. In fact, when looking at the UI, you notice that those lightweight VMs do not consume almost any resources—no CPU or memory resources and very few storage resources (2.91 Gb). When we look at the details of the virtual hardware, those VMs are configured with 128 Mb of RAM, 1 vCPU, and 2 Gb hard disk.

Final words ^

This is the first release of VMware vSphere Cluster Service, which is responsible for maintaining DRS operations in the event of vCenter Server unavailability. There will be more services added to future releases. I imagine that vSphere would be capable of managing not only vSphere services, but probably also some networking services, storage, or application services.

Subscribe to 4sysops newsletter!

Currently, VMware vSAN remains operational following vCenter failure; however, it's impossible to manage through vCenter because you can't connect. It would be great if we could manage vSAN when vCenter is not available, through a host client UI.

1 Comment
  1. Aped Brewsader 12 months ago

    Would you know what would happen if say, the datastore where agents are installed suddenly disappeared forever? 

Leave a reply

Please enclose code in pre tags

Your email address will not be published. Required fields are marked *


© 4sysops 2006 - 2021


Please ask IT administration questions in the forums. Any other messages are welcome.


Log in with your credentials


Forgot your details?

Create Account