Application Integration
cancel
Showing results for 
Search instead for 
Did you mean: 

VVOL with Nimble - why ?

SOLVED
Go to solution
csi-laser125
Occasional Visitor

VVOL with Nimble - why ?

Hi All ,  I hope it's the correct place to ask :

I'm now learning & deploying (or testing, to be more accurate) VVOL in our environment .

vSphere 6.5 , 40 ESXis and around 850 VMs .

 

From my short experience here is some points i could see :

 

  1. There is a mess in the Nimble's GUI . a VM is not stored and shown as a folder with the files in it, but as bunch of files . [and if you got hundreds of VMs, it means hundreds of files on the same "Folder"]

 

  1. You need to deal with each VM separately instead of in the LUN level .

Of course it has some advantages like giving more IO per VM instead of the whole LUN, but, it's also an disadvantage because you need to "edit settings" on each VM you want to apply the storage policy .

am I right ?

 

  1. To conclusion , the VVOL technology is recommended for specific environment, and not for all kind of environments generally

right ?

 

thanks in advance ,

Ron

11 REPLIES
ndyer39
Honored Contributor
Solution

Re: VVOL with Nimble - why ?

Hi,

I'm not quite sure if you're asking questions, or if you're making statements on your post...

1 - Yes, the Nimble array displays the VMs as individual volumes. Therefore what I personally would do is to organise by using more folders to assist with management.

2 - Your comment is also the reason why VVols were created - to manage them on the VM level, rather than the array side.

3 - I don't agree, there's nothing wrong with using VVols, and can be used generically in any type of VM estate. It does require it to be managed in a different way than traditional "LUN" based operations, but that's not necessarily a bad thing.

csi-laser125
Occasional Visitor

Re: VVOL with Nimble - why ?

Hi Nick,

You right, it was half question and half statement

I just wanted to write some facts and see if i got the VVOL logically in a right way .

Thanks for your reply .

Of course VVOL isn't a bad thing, but it's a different way of working, which to my opinion, now fits to all kind of PODs.

Maybe to most of it, but still, not to any situation [like all other technologies]

We'll hold the VVOL project for now, and maybe in the future we'll use it if our POD structure will change 

Thanks again for your reply Nick, and for the great videos . it really helpful 

Re: VVOL with Nimble - why ?

I have to agree with Nick Dyer's statement: "there's nothing wrong with using VVols, and can be used generically in any type of VM estate". It can be difficult to embrace new ways of doing things.  This is especially true with things like VVols that have turned the traditional approach of data storage within the hypervisor on its head.  Hang in there though.  Once you get your head around it ti will become apparent what VMware is shooting for and how awesome VVols can be.

You should take a look at these posts and the resources linked in them:

Whats New in Virtual Volumes 2.0

https://blogs.vmware.com/virtualblocks/2016/10/18/whats-new-in-virtual-volumes-2-0/?utm_content=bufferdea24&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer

Re: VVOL with Nimble - why ?

> There is a mess in the Nimble's GUI . a VM is not stored and shown as a folder with the files in it, but as bunch of files . [and if you got hundreds of VMs, it means hundreds of files on the same "Folder"]

The idea is to use only the vCenter UI to manage VVol VMs.  The only thing needed from Nimble GUI is the first step : that of registering the Vasa Provider.

In fact, many operations on VVol volumes are not allowed/recommended from the Nimble GUI.  The reason is this : the philosophy of VVols is to perform all management operations through the Vasa Provider - calls to the Vasa Provider are made only from ESXi and vCenter.

Please do not move VVol volumes into folders from the Nimble GUI.  Every folder created is a "storage container" for the vCenter.

csi-laser125
Occasional Visitor

Re: VVOL with Nimble - why ?

Thanks Mamata, that's make more sense .

We started now with one VVOL [keeping the rest of datastores as usual], and we'll see if this way of working fits to us .

Thanks again !

brent_irwin
Occasional Visitor

Re: VVOL with Nimble - why ?

Hello,

I just wanted to add my 2 cents to this discussion. We jumped into VVOLs a year ago at the same time we purchased an AF1000 array. After some issues totally related to me not understanding the whole, only work in vCenter and not in the Nimble GUI, thing, we have been doing great with VVOLs. I have migrated our entire environment to VVOLs, around 150 Guests and 8 Hosts. My number one win in using VVOLs was that I can maintain a large container, instead of many small LUNs. I used to maintain 4TB LUNs and found myself Storage vMotioning too often to keep everything even.

Now, I provision to the VVOL container and thats it. No need to balance the Guests across my LUNs. There is one hickup however. Nimble does not take deduplication into account when sizing your VVOL container. At least not in the version I am using. I have not upgraded to 4.x as of yet. This mean if you have 15TB RAW storage, you can only create a container size of a maximum of 15TB. Even if you get 30TB of storage, with dedup and compression, out of the 15TB.

We learned that you need to create two contains using the same available space. So, we have two 15TB containers using the same 15TB RAW storage. This gives me access to the entire 30TBs after dedup and compression. You do need to pay attention to your raw space filling up however. If you don't get 2x dedup/compression then your volumes in VMware may look like they have room available, but in reality they do not.

Other than the quirk in the dedup/compression calculations, I love my VVOLs and am glad I jumped in.

Brent

Re: VVOL with Nimble - why ?

Brent, thank you for sharing.  The container (folder) size maps to the datastore size, there should be no need to change that since it is just a holding object for a group of volumes.  When you create a VM Storage Profile with Deduplication set to 'Yes', the volumes of the VM will be deduped with other volumes that belong to the same dedupe domain.  If they are all part of this same datastore, and there are multiple copies (e.g. 10 copies of a Windows 2016 VM in one VVol datastore), you should see reduced datastore usage compared to no dedupe.  It would be great to get some more information from you about this - please let us know.

Thanks,

Mamata

brent_irwin
Occasional Visitor

Re: VVOL with Nimble - why ?

Hi Mamata,

I think I do need to clarify a bit more. The AF1000 I own currently has 15TB raw storage capacity. Regardless of the dedup/compress achieved, the biggest VVOL container I can create is 15TB. When I create a new VM in this new container, Nimble subtracts the original raw size of the VM from the available container size, even though, I am getting 2x compression/dedup.

So, if my new VM includes a 1TB C: drive and I turn on thick provisioning, the end result will be that I only have 14TB free space left on my VVOL container, from vCenters point of view. This is true, even though the actual storage footprint is 500GB with compression and dedup. In vCenter, if I fill up the 15TB VVOL contain and then look as the storage consumed on the Nimble array, I will actually only have used 7.5TB, with 7.5TB free space available. I have to create a second 15TB VVOL contain to use up the remaining 7.5TB of usable space on the AF1000.

The second VVOL container should also be created at the very beginning or you will not be able to create the full 15TB sizing. So, if you create your first 15TB container and fill it up so there is only 10TB free space left, you will only be able to create a second VVOL container that is 10TB max size. When I was faced with this issue, I ultimately had to migrate everything back to my old EqualLogic storage array and start from scratch on the Nimble and created two 15TB VVOL containers right off the bat.

An additional benefit to having two containers, is that I can now do Storage vMotion between them. This allows me to shrink vmdk’s after spaces has been reclaimed or I want to change a system from thick to thin provisioning.

Does that help explain what I was trying to get across?

Brent

BRENT IRWIN

Director, Information Technology

T: +1.858.281.7301

Ajinomoto Althea, Inc. | www.AltheaCMO.com<http://www.altheacmo.com> | 11040 Roselle St., San Diego, CA 92121

<http://altheacmo.com>

Re: VVOL with Nimble - why ?

Brent> The AF1000 I own currently has 15TB raw storage capacity. Regardless of the dedup/compress achieved, the biggest VVOL container I can create is 15TB.

Brent, thanks for adding details.  On my AF array (running 4.x), I could create a folder from Nimble GUI or CLI with limit more than raw capacity of the array.  e.g. on my array with 7.3TB raw capacity, I could create a 15TB folder.  Could you share your NimOS version please?

brent_irwin
Occasional Visitor

Re: VVOL with Nimble - why ?

They must have fixed that issue in 4.x then. I am still at 3.6.

If the extra management overhead is acceptable to you, I would still recommend creating 2 containers so you can Storage vMotion between the 2, unless you have 2 arrays to vMotion between.

BRENT IRWIN

Director, Information Technology

T: +1.858.281.7301

Ajinomoto Althea, Inc. | www.AltheaCMO.com<http://www.altheacmo.com> | 11040 Roselle St., San Diego, CA 92121

<http://altheacmo.com>

Dennis Handly
Acclaimed Contributor

Re: VVol with Nimble - why?

> The idea is to use only the vCenter UI to manage VVol VMs.

Similarly on HPE 3PAR we hide the VVols from a typical VV display.

The admins do need to know these "dark" storage objects take up space, so appear on usage summaries.