Subscribe

Vmware vSphere 5 VAAI Unmap feature working with Ontap 8.0.x?

Hi.

I'm testing the new features in vSphere 5 VMFS 5 and created an 400GB large VM on a newly created thin provisioned LUN and deleted the VM.

It worked as expected and the LUN returned to occupying almost nothing.

This behavior is one of the big reasons I have been sticking to NFS so far.

For reasons I won't go into we will probably switch to FC/VMFS for a while, possibly FCoE later on.

Now I noticed that especially the EMC camp has been shooting flares that there are timeout problems with the UNMAP feature and Vmware has KB describing how to disable it.

So my question is : is the VAAI accelerated delete (unmap) safe on Ontap 8.0.1 or do we have to go down the same route as the others and disable the feature until further?

Re: Vmware vSphere 5 VAAI Unmap feature working with Ontap 8.0.x?

Hi dejan,

i have already upgraded some of my customers to vsphere 5 and have now the problem with the timeouts.

so i have opened cases at vmware and netapp for assistance.

i'm looking forward to hear from them asap and will give you a response with the solution.

Kind regards.

Stefan

Vmware vSphere 5 VAAI Unmap feature working with Ontap 8.0.x?

Hi again,

after some analysis of my problem we see that there is a issue with vaai UNMAP feature and OnTap 8.0.x.

After disabling the features the problem is solved.

Whether the issue persits in 8.1 i can't say at this time.

See VMware KB2007723

Stefan

Re: Vmware vSphere 5 VAAI Unmap feature working with Ontap 8.0.x?

I think I have been lucky as I haven't seen any of these timeouts yet.

Probably a matter of time.

I would have liked to get Netapps (or in my case IBMs) official stand in this.

I will have to open a case it seems.

Message was edited by: Dejan Ilic

Re: Vmware vSphere 5 VAAI Unmap feature working with Ontap 8.0.x?

I had a customer with timeout and performance issues with 5, moving to the current patch set of ESXi500-201112001 the issues look to have been resolved.

Also there is a fix in 8.02P4 via Bug 488909 which might be related.