BlakeReed
Contributor
Contributor

Recompose/Delete Immediately on Logout is leaving a trail of storage concerns

We recently migrated to View Horizon 5.3 from 5.2 on all new infrastructure and we're seeing a behavior with our floating pool that we didn't see before.  We have a large floating pool based on Linked Clones and whenever vCenter/vComposer tears down and rebuilds a machine it isn't re-using the old machine's datastore folder, it's creating a new one _1, _2, _3, etc.  The first machine in the pool has been rebuild 9 times since Friday and it's leaving behind a folder with ~200MB of data in the from of two .vmdk files.

We're not seeing any errors on the composer tasks during the rebuild, no datastore based errors (locked files, etc.).  I've searched and come up with zip so far.  Any obvious ideas before I open a case?

image.png

image2.png

0 Kudos
6 Replies
nzorn
Expert
Expert

Is disk caching enabled?  Sounds like the same problem I am having so we aren't using disk caching currently:  https://communities.vmware.com/thread/436283

BlakeReed
Contributor
Contributor

We are using View Storage Accelerator with NetApp VAAI support.  Our NFS Datastores show 'Hardware Acceleration' supported and we're running NetApp's VSC 4.2.1 which supports vCenter 5.5.  All of this was configured the same way, on the same NFS Datastore under 5.2 and we didn't have this issue but maybe an old bug is back.

I'll turn them off and see if the behavior changes, thanks for the suggestion.

0 Kudos
BlakeReed
Contributor
Contributor

Just an update, disabling View Storage Accelerator and VAAI support did not resolve the situation and new folders are still being created.  I'm opening a ticket and will share the fix here later.

0 Kudos
CCSD
Contributor
Contributor

We are having the exact same issue with our NetApp 3240 MetroCluster.  We have a ticket open with NetApp and is being escalated but have not found a resolution.  Were you able to find a fix?

0 Kudos
WheatonCollege
Contributor
Contributor

We ran into this with NetApp and VAAI using composer linked clones and NFS. The fix for us was to turn off VAAI in all the pools and turn on Storage Acceleration instead. We had tickets open with VMware and NetApp for months and they battled it out and in the end shutting off VAAI fixed the issue.

0 Kudos
SanderGNL
Contributor
Contributor

After an extensive session with VMware Support, it appears that the current version of Horizon View has very limited support for VAAI. Enabling VAAI on unsupported backends can lead to this behavior, as well as other issues. To name one, we had DLL corruptions ( mstscax.dll, d3d8thk.dll and others) that occurred only on the linked-clone desktops and not in the original image.

For more information see:

VMware Compatibility Guide - View Composer Array Integration (VCAI) - Horizon

View Composer API for Array Integration (VCAI) support in VMware Horizon View (2061611)

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=206161...

0 Kudos