VMDK thin provisioning

[different-think]

Hi,

on our ESX farm, which is connected over NFS to a NetApp filer, we see that VMDK thin provisioning is on by default.

Is there a way to disable this for the entire host / cluster? The reason is that it causes confusion on our customer sides

because they have deployed 10 VMs with 20 GB each but have only 100 GB on the share.

Thanks in advance!

Robert

[davidjerwood]

I dont think you can turn it off as the default setting on NFS. But you can convert the thin provissioned disks to the thick disks if required at anytime. Let me know if you require the commands.

[different-think]

David,

thank you for getting back to me!

Is this behaviour NFS or Netapp specific. Someone from VMware told me that on his NFS share the vmdks are zerothick.

To be honest: converting to thick is no option because our customers are doing "self service" within their ressource pools.

Best,

Robert

[RParker]

Part of support is YOUR job is to EDUCATE your people.

If thin provisiong is causing a problem, then you need to tell them what they need to do so they do it right. Don't be afraid to tell them HOW do to things, especially if it's only going to make your job easier and minimal impact on them.

Explain to them the reasons WHY they should use THICK instead of THIN. Then show them how they can make work. If they have self service, then they should be able to do it themselves, with a little assitance from you or your group.

Educate and Encourage. Preventative maintenance is the best way to help people.

[Cowboy94]

Robert,

We were looking at using NFS at our organization recently but there was some concern with using the thin provisioning disk format, which is as you know the default disk format for vmdk files stored on NFS data stores. At first I thought it would be no big problem as we could simply go to the command line and convert the disk format to thick or eager zeroed in our case. However, if you look at the documentation for ESX 3.x you will find the following note:

"The only disk formats that can be used for NFS are thin and 2gbsparse. By default, files and virtual disks are stored in thin format on NFS servers, with blocks allocated on demand as needed. No other options, including RDMs, are supported on NFS."

The whole "no other options, including RDMs, are supported on NFS" statement resulted in a lot of internal debate. A SR was even put into VMware and the result was that the statement is still true, only thin disks are supported on NFS. Because our management does not like to do anything that may violate support, we are going to keep the disks on the NFS volume in the default thin format.

But to answer your initial question, you can not change the default disk format. If you are going to use a self servicing NFS datastore, you will end up with thin vmdk files. In our case, the NFS administrator just increases the share when the amount of free space gets under 30%.

Travis

[dlm1975]

Just as a note, they have rewritten this in the 3.5 Configuration Guide:

"NOTE The only disk formats you can use for NFS are thin, thick, zerodthick and 2gbsparse. Thick, zeroedthick and thin usually mean the same because the NFS server and not

the ESX Server host decides the allocation policy. The default allocation policy on most NFS servers is thin."

[andreazauli]

HI david,

I don't know the command... i'm not shure about the -j option or if it could be possible to do the conversion by cloning (-i), because during the cloning, the -d option doesn't accept

the thick value. Could you post the command to convert thin disks to thick disks ?

thanks

andrea

[Amir_vm]

Hi , I got the same error when i tried to convert an exiting thin Virtual disk to thick, it doen't accept -d thick, Any suggetion?/

[vhii81]

Does anyone know why the disk still shows as thin even though i've inflated and used vmfktools to convert it? thanks

[Igor_The_Great]

Here is some additional info that I found - this is true for ESX3.5/4.0 and a few things are applicable going forward as well.

Question: What type of Disk Provisioning is supported in ESX?
Answer: Disk Provisioning:

• thick 
  • A thick disk has all space allocated at creation time. This space may contain stale data as it exists on the physical media.
• thin
  • Space required for thin-provisioned virtual disk is allocated and zeroed on demand as opposed to upon creation.

Question: Why is Disk Provisioning different for NFS vs VMFS volume?
Answer: On the client side we have no control over how the server is going to allocate blocks for a file.

Question: What is the factor on NFS Server that causes it to be Thin or Thick provisioned?
Answer: It depends on how the filer is configured, but on almost all the filers we know of - the files will be thin provisioned.

Question:Why isn't there a common default standard?
Answer: Block allocation is not part of the NFS spec since NFS is a file level protocol. All servers will allocate blocks on demand (thin provision) but some can be configured to pre-allocate space for files based on the size of the file.

Question:Assuming that NFS Server has Thick provisioning by default will I be able to do Thin provisioning of vms?

Answer: No.