vmb01
Enthusiast
Enthusiast

Streched cluster and Affinity Rules

Jump to solution

Hi

in a vSAN streched cluster it's possible to use DRS soft affinity rules  to set a Preferred Fault Domain for a bounch of VMs.

If I need to move (vmotion) a VM or some VMs to the Secondary Fault Domain (without having any failure) what happens?

DRS will enforce the soft affinity rule and vmotion back the VMs to the preferred site? 

And the same question for a simple, not streched, cluster. If I have a soft rule to prefer a couple of hosts for some VM and I need to move a VM to an host 'not preferred' what does DRS do?

 

2 Solutions

Accepted Solutions
DurhamNeil
Enthusiast
Enthusiast

We have a stretched cluster (not vSan, but Nutanix hardware with ESXi hypervisor).

VM/Host Rules manage which side of the cluster the VM resides in. It works fine.

During patching or host reboots, it's possible for a VM to end up on the wrong side of the cluster and the DRS rules just send it back to the right side. VMs work find on the wrong side, it's just their compute/storage aren't physically local to each other any more.

VCP-DVC 2022 & CCNA

View solution in original post

depping
Leadership
Leadership

DRS will definitely try to move it back if you are violating rules. The VMs can of course only move back if the host is available, that is why you use maintenance mode in these situations 🙂

View solution in original post

2 Replies
DurhamNeil
Enthusiast
Enthusiast

We have a stretched cluster (not vSan, but Nutanix hardware with ESXi hypervisor).

VM/Host Rules manage which side of the cluster the VM resides in. It works fine.

During patching or host reboots, it's possible for a VM to end up on the wrong side of the cluster and the DRS rules just send it back to the right side. VMs work find on the wrong side, it's just their compute/storage aren't physically local to each other any more.

VCP-DVC 2022 & CCNA
depping
Leadership
Leadership

DRS will definitely try to move it back if you are violating rules. The VMs can of course only move back if the host is available, that is why you use maintenance mode in these situations 🙂