What is the right way to do it to ensure a vSAN cluster is not holding any meta data about the removed host?
How to make sure that all objects held by the floating node is migrated to another node or disk group in the old cluster before removing the ESXi host?
Would that have an impact on the capacity of the cluster?

It is important to follow the correct procedure when you are looking to permanently decommission or migrate a vSAN node. If the disk groups are left behind or the node is not disassociated with the cluster, residual components can be left behind which can have an impact on cluster stability.

Below are the steps I follow to ensure removing ESXi host from one vSAN cluster without any impacts on the capacity and without compromising vSAN objects health.

•  Ensure there is sufficient capacity in the vSAN disk groups to decommission a node. You can do that by running “Data Migration Pre-check” from the “Monitor” tab on the vSAN cluster.
  
  • Select the candidate host which you are planning to move \ decommission.
    
  • Ensure “Full Data Migration” option is selected to check if the remaining vSAN nodes have enough storage to hold the migrated vSAN objects.

If the test passed successfully, you could place the host in “maintenance mode” within the test.

• Or you can put the host into maintenance mode with full data migration selected by right clicking the host and putting in into maintenance mode.

• Wait for resync traffic to complete and the host to enter maintenance mode. You can check the progress for data resync from the “Monitor” tab of the vSAN cluster by checking “Resyncing Objects” under “vSAN”
  • This screenshot shows that resync is done successfully and at that stage the host should have successfully entered in maintenance mode.

Now you can safely delete the disk group(s) that reside on the host you want to move \ decommission by selecting the vSAN cluster>Configure>Disk Management> Select the disk group >Remove the Disk group. Repeat this step as necessary to delete all disk groups under the vSAN node.

Click “Remove” in the confirmation page. Repeat this action as many as disk groups you have under the vSAN node.

• Move the ESXi host out of the cluster to disassociate it from vSAN. You may also run esxcli vsan cluster leave command from the command line to leave the cluster. I do both to ensure the host is no longer seeing the old vSAN cluster.
  • You can move the vSAN node out of the cluster by dragging it out of the cluster.
  • Then enable SSH on the host and run the command esxcli vsan cluster leave

Now the host has been removed successfully from the vSAN cluster.

Finally to ensure that the Node is successfully no longer part of the vSAN cluster, you can use
esxcli vsan cluster get

The result of the above command should give you vSAN disabled.

Happy vSAN every one 🙂

Written By,
Mohamed Basha