vSphere Clustering Service VMs on Veeam NFS datastore


Userlevel 7
Badge +12
  • On the path to Greatness
  • 1261 comments

As you may know, with vSphere 7 Update 1 VMware introduced the vSphere Clustering Service.
The plan is to get more independent of vCenter Server for certain cluster features like for example DRS.
Where you certainly notice the difference, compared to older releases, are the vCLS agent VMs shown in vCenter.

https://blogs.vmware.com/vsphere/2020/09/vsphere-7-update-1-vsphere-clustering-service-vcls.html

There's no need to backup them and if you have a current Veeam version (v10a Patch 20201202 or higher), they're are automatically excluded from your jobs.
The agent VMs are manged by vCenter and normally you should not need to look after them.
Normally…yesterday we've had the case were some of the vCLS VMs were shown as disconnected; like in this screenshot:

Checking the datastore we have noticed that those agents VM had been deployed to the Veeam vPower NFS datastore.
As this datastore is managed by the Veeam server and can connect/disconnect at any time, it's a very bad location for the vCLS VMs. Unfortunately you can't control the placement of the VMs or hide/mask a datastore, so this could happen anytime.
Searching for similar issues I've only found this post in the VMware Community; so this issue is known but probably also rare:

https://communities.vmware.com/t5/VMware-vCenter-Discussions/vCLS-VMs-datastore-selection/td-p/2837202

 

If this happens to you and you find the VMs disconnected, the solution or workaround is rather simple. Just remove the VMs from the inventory in vCenter with a right-click. A few seconds later a new VM is deployed.


 


20 comments

Userlevel 7
Badge +20

Yes this new feature for v7 is interesting especially how vCenter deploys them.  They are not marked for backup for my lab nor would I back them up since a new one is deployed anytime anything happens.

Great article links and some further reading now. :grinning:

Userlevel 7
Badge +4

@regnor : Interesting, thanks !

Userlevel 7
Badge +13

Thanks @regnor for sharing your finding!

I have to say, VMware has not done itself any favors by enabling this feature in vCenter 7 U1 without an easy option to disable it. I would have preferred to see a opt-in, rather than a complicated opt-out.

Userlevel 7
Badge +12

That’s right @vNote42 Such a change should be well thought and perhaps needs more attention or a transition phase.

By the way it looks like I’ve posted this as a question instead of “content”… @Kseniya Is it possible to change the topic type?:neutral_face:

 

Userlevel 7
Badge +3

Just converted it to ‘content’, @regnor :blush:

Userlevel 7
Badge +12

Just converted it to ‘content’, @regnor :blush:

Thanks Kseniya 😀

Userlevel 7
Badge +11

Great post @regnor !!! Thx for sharing! Love this ;-)

Userlevel 7
Badge +16

Something to be aware of the rare time I need to do recovery processes. Eek! VMware strikes again 🙄

Thanks for sharing @regnor !

I came across this scenario in my setup also, vCLS VM’s on Veeam NFS Datastores. I could not just delete the VM’s.

I think it came about due to an unscheduled power outage and when the hosts came up vCLS chose those datastores inadvertantly.

Anyway FIX was….

I had to add the Advanced Setting to my vCenter Server to perform the RETREAT method of getting rid of the vCLS VM’s.
Set to false.
Then unmounting the Veeam NFS Datastores.
Set back to true.
This then recreated the vCLS VM’s on more preferred permanent datastore.
 

Putting a Cluster in Retreat Mode

Userlevel 7
Badge +16

@MacGee - that’s an option, but VMware allows you to storage vmotion the vCLS VMs to a different datastore so you don’t have to go through that process, as noted on pg. 84 of the vSphere Resource Mgmt Guide. At least, I think you can storage vmotion the VMs out of the Veeam NFS Datastore...

Userlevel 7
Badge +22

I only figured what these little things were a week ago during the vsphere course :) . Before that it was like I have no clue what these are, no time to google so JUST don’t touch them.

It pays to take courses.. :) 

Userlevel 7
Badge +16

I’ve read through both the v7 and v8 Guides talking about them over the past few yrs, but still don’t know what they actually “do”. VMW doesn’t actually state what they do except say they’re “used to maintain the cluster services health”. Umm...ok, but what do they actually *DO* is my question. 🤣😂

Userlevel 7
Badge +22

I think it was to get DRS out of vcenter. At least one of the reasons. So that if vcenter goes down DRS and HA will still work?

Userlevel 7
Badge +22

Have to look at my course notes… this could lead to an epic exam fail… although I think you said the course is useless for the exam :) . Just paying the toll :) 

Userlevel 7
Badge +16

Not entirely sure to be honest. Again, the Guides don’t go into that detail...you know, actually relevant info detail 😊

Userlevel 7
Badge +16

Well, the Course was useless 14yrs ago. Could’ve changed since then 😊

Userlevel 7
Badge +22

Google was faster than my course notes and this fellow has a nice explanation:

 

https://www.stephenwagner.com/2020/10/15/vcls-virtual-machine-vmware/

Userlevel 7
Badge +16

Yeah..same info that’s in the Guide, minus the part about vCenter being down.

Userlevel 7
Badge +7

Hi all, doesn't seem to have been mentioned, but you can specify the datastores allowed for vCLS:

https://docs.vmware.com/en/VMware-vSphere/7.0/com.vmware.vsphere.resmgmt.doc/GUID-6C11D7F9-4E92-4EA8-AA63-AABAD4B299E7.html

https://www.provirtualzone.com/move-vsphere-cluster-services-vcls-vms-datastore/

Userlevel 7
Badge +20

This is what I do in my homelab.

Comment