Snapshot vmdk files of the Parent image used to create IC pools, not deleted from Datastore

[paulhackett]

Environment:

• Host - VMware ESXi, 7.0.1, 17551050
• Vcenter - Version: 7.0.2 Build: 17694817
• Horizon running on Windows Server 2019 (no other applications running on this server)
• Horizon - 8.3.0 build - 18294467,
• Pool - Instant Clone, Floating Assignment

We are having the same issue as mentioned in this article https://kb.vmware.com/s/article/79869 but we do NOT have multiple pools using the same image.

We have 20 Master VMs each with a Horizon Pool attached to them.

What we have been doing for snapshot maintenance has been the following:

1. Make changes to the Master Image
2. Shut down Master Image
3. Take a snapshot
4. Publish the pool
5. Clean up oldest snapshot from vcenter

What this looks like is:

Master VM

• Snapshot 1 - 10/1/21
• Snapshot 2 - 11/1/21
• Snapshot 3 - 12/1/21

In this example I would remove Snapshot1 - 10/1/21 from vcenter.

When I look on the Datastore this snapshot is NOT removed.

The only way that I have found to effectively remove the old snapshots from the datastore is to do a storage Vmotion and then do a consolidate snapshots.  Of course this removes all of the snapshots which is not something we want to do moving forward as we like to keep at least a couple of snapshots so we can go back if we find an issue from the changes we just made.

The tests that I have done so far:

1. Created a new VM, create multiple snapshots, delete oldest snapshot through vcenter and found that it was deleted from the datastore.
2. Created a new VM, create a snapshot, Create new Instant Clone pool from this VM and published the pool. 
1. Make changes to the VM, shut it down, create a snapshot, publish the pool (I did this a few times)
2. Delete the oldest snapshot through vcenter.  It was NOT deleted from the datastore. 

I've run these same tests a few times and have found that each time I have an instant clone pool attached to the VM, the snapshots are NOT deleted from the datastore.

Any help would be greatly appreciated.

[LukaszDziwisz]

We have the same situation where snapshot is deleted however it leaves behind a lot of vmdks, over time provisioning of pools becomes slower and slower until it reaches timeout and those pools would fail. We had  a ticket opened with vSphere team and they blame it on outdated fnic driver for our storage and recommended updating it. I would check if you are on compatible storage driver. For now our workaround is to clone parent image (simply add something to its name and publish new clone to the pool.

[paulhackett]

 LukaszDziwisz, thank you for the reply and the suggestion.  This is certainly a pain.  I should have mentioned that we are using local SSD on the ESXi host for this and that there is no SAN attached.