File Server Datastore Provisioning assistance

[vipclubber]

I am building a Win2012R2 File Server.

Datastore1 = 800GB

Datastore2 = 12TB

Datastore1 is taken. Datastore2 will have about 2TB space used for storage of copies of VMs. The rest is for File Server.

What provision is recommended to use for a growing File Server? Thick or Thin?

Is it ok to OVER allocate on Datastore2 space?

If other data on Datastore2 (other than File Server) begins to grow how will the File Server make up for the difference?

Thanks in advance.

[vipclubber]

See attaached. Why is the Free space smaller than Provisioned?

I provisioned 11tb but C drive only showing 2tb???

[a_p_]

What provision is recommended to use for a growing File Server? Thick or Thin?

Is it ok to OVER allocate on Datastore2 space?

There's no general rule of thumb regarding thin or thick provisioning. It actually depends on your needs and - that's important - whether you can increase the storage size fast enough if required. With thin provisioning in use, it's your responsibility to monitor datastore usage!

If other data on Datastore2 (other than File Server) begins to grow how will the File Server make up for the difference?

Basically the same as above. You need to monitor space usage to avoid issue. The file server itself - i.e. the virtual machine - is not aware of the physical disk space used on the datastore. If the daatstore runs out of disk space, your VMs will likely crash.

Why is the Free space smaller than Provisioned?

I provisioned 11tb but C drive only showing 2tb???

It looks like the guest OS partition uses an MBR partition table rather than GPT. With MBR you are limited to ~2TGB.

André

[vipclubber]

How to convert a SYSTEM OS C: disk from MBR to GPT without a data loss? Is it possible or do I have to reinstall everything using diskpart tool?

[a_p_ · Accepted solution]

Unless any 3rd party tool can do this, you need re-initialize the system disk. In addition to this you'll have to edit the VM's boot options, and switch the boot firmware from BIOS to EFI. Another option is to shrink the guest's current system partition to e.g. 100GB, and then create a second partition with a GPT partition table.

Please keep in mind that backing up and - even more important - recovering such large partitions will consume a considerable time, and it may be worth planning for multiple smaller partitions if possible.

André

[vipclubber]

Unfortunately, the 3rd party tools do not allow to convert Bootable OS partition to GPT. In my case, I found out that HP BL460c blade has Legacy BIOS and can not boot in UEFI. So the remaining option is to go with shrinking the C: drive to 200gb and using 10.8 TB (the max size for SAS drives RAID array allow) for data partition drive 😧

I wonder if I had LINUX OS, I had a better luck in solving this issue?

[vipclubber]

I was able to shrink the C: drive to 200gb desired space. Now I have two unallocated partitions that I need to combine into one and convert to GPT.

Any ideas how to best handle this? See image attached.

[a_p_]

Sorry, my bad. This doesn't work, since the partition table on a single disk is either MBR or GPT.

In my case, I found out that HP BL460c blade has Legacy BIOS and can not boot in UEFI.

It's not the host, but the guest OS which requires EFI boot in order to boot from a >2TB system disk. So you will unfortunately not be able to workaround this, but need to reinstall the guest OS. Even with GPT in place, you may still consider to create separate OS and data partitions, or - even better - create two virtual disks, one for the OS and one for the data.

André

[vipclubber]

Yeap, Reinstall sounds like the way to go here. Couple of questions to get this right this time around:

1. Can I delete the OS part without having to delete the entire VM or I should delete the entire VM and redo it from scratch?

2. Should I create two virtual disks right away or first create 200gb for OS then 10.8TB GPT for data?

Thanks in advance.

[a_p_]

I'd actually go with Option 2. There's no need to delete the whole VM, you can also just delete the VM's virtual disk, and then create 2 new ones. Please make sure you close the VM's settings after deleting the virtual disk, and before creating the new ones. This will ensure that the disks are named properly.

Rembmer that if you want to have both disks with GPT, you need to set the virtual boot firmware to EFI.

André

[vipclubber]

Are you referring to VMDK file(s) to be deleted in current virtual machine?

In my case there are two vmdk files. See attached,

Because my blade does not have EFI, I can only have data drive set to GPT.

I would have to create the first Drive as eager zero for Win OS. Then create thin provisioned for data in GPT.

Should I create both drives right away? First OS then Data?

[a_p_]

The .vmdk files shown in the screenshot belong to a single virtual disk with an active snapshot. Anyway, do NOT delete the files from the datastore browser. Delete it from the VM's settings hardware dialog.

Because my blade does not have EFI, I can only have data drive set to GPT.

The host's hardware doesn't matter. It's only the VM's setting and the guest OS partitioning which matter.

I would have to create the first Drive as eager zero for Win OS. Then create thin provisioned for data in GPT.

Don't bother with eager zeroed disks for operating systems. This won't improve performance very much.

Should I create both drives right away? First OS then Data?

Yes, after deleting the VM's current virtual disk, you can create both new virtual disks. This way the OS disk will be attached to SCSI 0:0 and the data disk will be SCSI 0:1.

André

[vipclubber]

Oops, I already deleted them from the datastore browser first (after consolidating snapshots) then I deleted the drive from VMs settings. Hope its not a problem. It did allow me to create new drive without any problems.

Question regarding creating second GPT drive (Thin) for 10.8TB:

At which point do I indicate it is in GPT? In the OS?

[a_p_]

Yes, the guest partitioning is done from the guest OS level.

André

[vipclubber]

I gave 😧 drive 10.8TB space i.e. 10800GB

So not sure why 😧 drive is only showing 10.5TB

I have a question regarding the mode of 😧 drive in vmware:

Do I want this large D drive be affected by snapshots or not? MODE: Independent Persistent/Non or not at all?

[a_p_]

I gave 😧 drive 10.8TB space i.e. 10800GB

So not sure why 😧 drive is only showing 10.5TB

That's correct. One TB equals 1,024GB, so 10,800 / 1,024 = ~10.56TB

Do I want this large D drive be affected by snapshots or not? ...

It depends on whether you need snapshots, e.g. for VM based backups.

... MODE: Independent Persistent/Non or not at all?

You definitely don't want to set the mode to "Non persistent"! In this mode all changes are discarded when you shut down the VM.

André

[vipclubber]

I guess i would have to recreate it again for 10.8tb. You mentioned not to delete vmdk drive file from datastore browser but do it in the settings. Why? what is the proper way to delete?

Do you know what would happen if I create a drive larger than what my RAID controller can support. My RAID support 10.8 TB sas volume but what if I create new 11tb HD in VM settings?

[douglasarcidino]

Basically, you do it from the VM settings because it removes the path to the VMDK from the config file and then also gives you the option to remove it from the datastore. Deleting VMDK and other VM files from the datastore browser is reserved for cleaning up after a disaster of some kind that left extra files behind.

[vipclubber]

In my case, with previous VMDK delete I did both but in the wrong order. First I deleted VMDK from Datastore browser then I deleted it from VM settings.

I am not sure if the method above still did it cleanly. If not, what file should I be cleaning up and what to clean up?

[douglasarcidino]

Browse the datastore and look at what big files are still there.You can edit the settings of your VM and get the path to the disk that is still there. Remove the other VMDK that isn't still in the folder. If you post a screen shot of your datastore, we can probably point you to the specific file.