Highlighted
8 Krypton

VAAI support information

This post is meant to summarize and serve as a central location for the operational support of VAAI under VMware vSphere 4.1.

Prerequisite reading

See the below link for a great explanation from Chad Sakac around VAAI, what it is, and what benefits can be seen.  In summary VAAI is implemented by way of SCSI commands that allow for certain predictable tasks directed toward storage to be offloaded thus minimizing cpu cycles across the whole infrastructure stack.

http://virtualgeek.typepad.com/virtual_geek/2010/07/vsphere-41---what-do-the-vstorage-apis-for-array...

http://www.yellow-bricks.com/2010/11/23/vstorage-apis-for-array-integration-aka-vaai/

VAAI FAQ VMware KB

See the below link for VMware’s official KB article around FAQ’s for VAAI.  Some information in this article is from the link below.

http://bit.ly/eypxK1

What arrays support VAAI right now?

http://v-reality.info/2010/10/list-of-vaai-capable-storage-arrays/

Interesting takeaways

You cannot use VAAI when source and destination VMFS volumes have different block sizes or the VMDK is eagerzeroedthick going to thin.  Check with the VAAI FAQ VMware KB above to see other caveats to when VAAI would not be used.

Why is the status value showing as unknown?

I’ve ran into a few situations where datastores are listed as “Unknown” for hardware acceleration status.  This can be checked from Configuration -> Storage, or from the console with “esxcfg-scsidevs -l | egrep "Display Name:|VAAI Status:”.  In order to show the datastore as “Supported” for hardware acceleration, an offload operation must first take place.  The link above highlights a few things to verify to ensure that the VAAI features are enabled at the kernel level.  From there the article also highlights how to perform a basic operation that would utilize the hardware acceleration and would then flip the status to “Supported”.

Where to look for configuration settings via VI Client?

VI Client -> ESX/i Host -> Configuration -> Advanced Settings -> DataMover/HardwareAcceleratedMove, DataMover/HardwareAcceleratedInit, VMFS3/HardwareAcceleratedLocking

Other VAAI troubleshooting commands

“esxcfg-advcfg –g /DataMover/HardwareAcceleratedMove”

                Value of HardwareAcceleratedMove is 1

“esxcfg-advcfg –g /DataMover/HardwareAcceleratedInit”

                Value of HardwareAcceleratedInit is 1

“esxcfg-advcfg –g /VMFS3/HardwareAcceleratedLocking”

                Value of HardwareAcceleratedLocking is 1

“esxcli vaai device list” – Lists VAAI devices that are supported and what plugins are used

            naa.600601600e00220028efb534e8cfdf11

Device Display Name: DGC Fibre Channel Disk (naa.600601600e00220028efb534e8cfdf11)

            VAAI Plugin Name: VMW_VAAIP_CX

naa.60000970000194900548533030303338

Device Display Name: EMC Fibre Channel Disk (naa.60000970000194900548533030303338)

            VAAI Plugin Name: VMW_VAAIP_SYMM
            

Enabling VAAI by way of scripts

http://vghetto.svn.sourceforge.net/viewvc/vghetto/scripts/vaaiHWAccelerationMgmt.pl?revision=81&view...

http://www.virtuallyghetto.com/2010/08/hidden-esxcli-apis.html

Labels (1)
0 Kudos
6 Replies
8 Krypton

Re: VAAI support information

Hi Clint -

Great post! I do have a specific one I haven't had clearly answered just yet, though I've read blogs by / been in contact with Bas. Here's the scenario:

On a NS-960 with an iSCSI LUN presented to an ESX host, does the VAAI offload the additional IOPS to the DataMover or the Storage Processor directly?

My confusion is that the type of I/O that is bound to be generated by the three features is inherently a block level task, but the terms are named

"DataMover/HardwareAcceleratedMove" and "DataMover/HardwareAcceleratedInit." Logically my mind leads me to think the load is on the SP / backend, but by naming convention I think the task is handled by the DMs. Can you clarify for me?

Thanks - Keep the knowledge coming!

mjbrender

0 Kudos
8 Krypton

Re: VAAI support information

Matt,

Thanks for the comment.  VAAI is integration on the native block side of the array at the storage processor level.  I can see that it is a bit confusing due to VMware labeling the settings with DataMover.  To clarify, in the EMC world Datamover refers to a NAS head that serves us iSCSI/NFS/CIFS volumes.  Although VMware labels the option with DataMover, it is actually integration to the native block side of our array by way of the storage processors.

With that being said, VAAI will only work with on the native block side (Clariion) of our array by way of iSCSI/FC/FCoE.  So to confirm yes the offload will happen at the storage processor level.

0 Kudos
8 Krypton

Re: VAAI support information

Thank you sir! That's the level of clarity I wanted to ensure before I provided a Technical Consultant with a definitive answer. I appreciate the reply.

Take care,

mjbrender

0 Kudos
8 Krypton

Re: VAAI support information

Any idea when VAAI will be available on the VPLEX? In the list of arrays it says Q3/2011 but I didn't think it was available yet in the latest  GeoSynchrony code.

0 Kudos
Cihl
6 Indium

Re: VAAI support information

Hi! I'd also like to know when VPLEX will be supported with VAAI. Thanks.

0 Kudos
8 Krypton

Re: VAAI support information

Cihl wrote:

Hi! I'd also like to know when VPLEX will be supported with VAAI. Thanks.

Cihl, there is a plan, it is on the roadmap,  customers can contact the VPLEX sales overlay team or Vplex champions to get an NDA presentation to communicate this to customers.

0 Kudos