Over the last month or so I have had two customers inquire about vROPS’s ability to perform cross-cluster storage vmotions in order to enforce business intent or multi-cluster workload optimization.

Being the eager wet-behind the ear SE that I am, I set it up in my lab with some vSAN datastores and it worked! Success! Then I reviewed my notes of the use case the customer described:

  • NFS based storage
  • Datastore per cluster
  • No shared datastores between clusters

no-shared-storage

I then set out to create that scenario in my lab so that I could let my customers know that vROPS would solve their problems and slice their bread, to my dismay I was met with the following helpful message from vROPS.

error

I turned to a bit of RTFM at this point, which called out that I needed to use Datastore Clusters for non vSAN workloads.  I proceeded to rush into a configuration where I created a common datastore cluster for all my clusters, but non-shared datastores.

oops

I’m not sure why I thought this would work

You should not be surprised that this did not work as I would expect. So after hammering at it for a bit, I realized that the needed configuration was no-shared storage, no-shared datastore clusters – which in hindsight makes complete sense.

The winning receipe is what makes logical sense.  The key here is that the datastores need to be fully shared between all clusters you want to balance between or the datastores need to not be shared, but a part of a “dedicated” datastore cluster for that vSphere cluster. See the diagram below for two different valid configurations.

thereciepe2

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.