Unmounting or detaching a VMFS, NFS and vVols datastore fails

Products

• VMware vSphere ESXi

Issue/Introduction

This article provides required steps to successfully complete datastore detach or unmount workflows which might be impacted due to the placement of vCLS VMs.

Symptoms:
Unmounting a VMFS, NFS or vVols datastore fails

Environment

VMware vSphere ESXi 7.0.0

Cause

This issue occurs as due to presence of running vCLS VMs on these datastores.

Resolution

To resolve this issue:

Prior to unmount or detach a datastore, check if there are any vCLS VMs deployed in that datastore. If there are any, migrate those VMs to another datastore within the cluster if there is another datastore attached to the hosts within the cluster. 

If the datastore that is being considered for "unmount" or "detach" is the only datastore connected to the host and this datastore has vCLS VM, then the only option is to power-off this vCLS VM is by using Retreat Mode on the cluster. Note: This datastore might be connected to different hosts in different clusters and vCLS VMs from all these different clusters might be stored in this datastore. In such a case, Retreat Mode has to configured on all the clusters connected to this datastore, so this datastore can be unmounted or detached. 

• Use Retreat Mode to remove vSphere Cluster Service VMs from a cluster.  First, enable Retreat Mode in the cluster so that all the vCLS VMs in the cluster gets deleted. If in case the same datastore is shared between multiple clusters and the vCLS VMs from different clusters are placed in this same datastore then in all the clusters where the corresponding cluster vCLS VMs are placed in this datastore should be enabled with Retreat Mode. Also when this Retreat Mode is enabled, the DRS in that particular clusters will be non-functional until Retreat Mode is disabled back. Once the datastore Maintenance Mode is successful, Retreat Mode should be disabled in all clusters connected to this datastore where its enabled in order to get back the vCLS VMs in all the cluster for DRS functionality.