continuum
Immortal
Immortal

Why does ESXi 6.5 GUI creates VMFS-volumes that are not alligned to 1MB blocks ?

Jump to solution

Why does ESXi 6.5 GUI creates VMFS-volumes that are not alligned to 1MB blocks ?

VMFS 6 - 100gb datastore

partedUtil getptbl mpx.vmhba3\:C0\:T0\:L0

gpt

13054 255 63 209715200

1 128 209715160 AA31E02A400F11DB9590000C2911D1B8 vmfs 0

VMFS 6 - 2222gb datastore

partedUtil getptbl mpx.vmhba3\:C0\:T1\:L0

gpt

290063 255 63 4659871744

1 128 4659871704 AA31E02A400F11DB9590000C2911D1B8 vmfs 0

VMFS 5 - 100gb datastore

partedUtil getptbl mpx.vmhba2\:C0\:T0\:L0

gpt

13054 255 63 209715200

1 128 209715160 AA31E02A400F11DB9590000C2911D1B8 vmfs 0

VMFS 5 - 2222gb datastore

partedUtil getptbl mpx.vmhba2\:C0\:T1\:L0

gpt

290063 255 63 4659871744

1 128 4659871704 AA31E02A400F11DB9590000C2911D1B8 vmfs 0


________________________________________________
Do you need support with a VMFS recovery problem ? - send a message via skype "sanbarrow"
I do not support Workstation 16 at this time ...

1 Solution

Accepted Solutions
elesueur
VMware Employee
VMware Employee

Hi all,

If the host client is not doing The Right Thing, then it's because I did it wrong Smiley Sad

Can someone explain to me the consequences of not aligning to 1MB?

View solution in original post

0 Kudos
11 Replies
a_p_
Leadership
Leadership

Maybe CHogan can provide some information about this!?

André

0 Kudos
CHogan
VMware Employee
VMware Employee

Not noticed this to be honest.

Does the same thing happen when you create a datastore with vmkfstools?

Which UI are you using? Its not the old C# client, is it? If so, try the same task with the web client, and see if the behaviour is the same.

If it is, then I'd suggest opening an SR as that is not expected behaviour imo.

http://cormachogan.com
0 Kudos
a_p_
Leadership
Leadership

Thanks for the reply.

In my case the datastore was created from the HTML5 Host Client (version 1.15 which came with the latest v6.5 patch bundle).

I didn't use vmkfstools yet, but I guess there will be no issue with this, because it requires an already created partition.

Just to put everything together what I discovered so far:

- ESXi 6.5 creates a VMFS5 datastore on the installation drive (using the interactive CD/ISO installation)

- VMFS not 1MB aligned by default (as mentioned)

- with a preexisting partition on the target disk/volume - even with the option to use all storage space - the existing partition gets deleted (as expected), and the new VMFS partition is created as partition 2 (instead of 1). Deleting an existing partition in the first step, and then creating a VMFS volume in a second works as expected (partition 1).

I'll do some more testing.

André

0 Kudos
elesueur
VMware Employee
VMware Employee

Hi all,

If the host client is not doing The Right Thing, then it's because I did it wrong Smiley Sad

Can someone explain to me the consequences of not aligning to 1MB?

0 Kudos
jhboricua
Enthusiast
Enthusiast

The Impact of Misalignment | StorageReview.com - Storage Reviews

in a nutshell, bad performance due to unnecessary disk IO operations.

a_p_
Leadership
Leadership

Thanks for your response.

Just to see whether this is an issue with an API, or the Host Client itself, I added two additional (virtual) disks on my lab ESXi 5.5 host, and created datstores on them.

I used ESXi 5.5 because it still supports the "old" Windows based vSphere Client. The ESXi Host Client version used is 1.17.

Results:

VMFS Datastore created using the old vSphere Client:

~ # partedUtil getptbl /vmfs/devices/disks/vml.0000000000766d686261313a323a30

gpt

1305 255 63 20971520

1 2048 20971486 AA31E02A400F11DB9590000C2911D1B8 vmfs 0

VMFS Datastore created using the ESXi Host Client:

~ # partedUtil getptbl /vmfs/devices/disks/vml.0000000000766d686261313a333a30

gpt

2610 255 63 41943040

1 128 41943000 AA31E02A400F11DB9590000C2911D1B8 vmfs 0

André

elesueur
VMware Employee
VMware Employee

Thanks everyone. I have fixed the issue in our code.

continuum
Immortal
Immortal

Hi elesueur
Thanks a lot for your answer.
This helps me a lot.
My job is to recover vmdks from corrupt VMFS-volumes.
So I often have to scan raw volumes to find the lost vmdks.
For ESXi-versions 5.0 and later I assume that the VMFS-partition starts with 1MB offset.
If I came across a ESXi 6.5 installation I would assume that every VMDK-beginning - which does not start at an offset of a  multiple of 1MB relative to the start of the disk - is a false positive detection.
I find those VMDK-beginning with the Linux-carving-tool scalpel.
Now that I know that some ESXi 6.5 setups used the unexpected offset of 128 sectors I can regard those cases as exceptions from the rule and can ask the customers about the exact version they used.
Without that knowledge I would have assumed that the volume has been upgraded or that an adminustrator had fixed the partitiontable incorrectly.
Thanks a lot for fixing the code so quickly Smiley Happy
Regards Ulli


________________________________________________
Do you need support with a VMFS recovery problem ? - send a message via skype "sanbarrow"
I do not support Workstation 16 at this time ...

0 Kudos
vrangler
Contributor
Contributor

Hi,

I installed the latest esxi 6.5 with the ISO installer to a local 1TB ssd the other day and the partitions aren't aligned right either. Seems the same issue as with the host/web client.

GPT partition table:

Number  Start (sector)End (sector)  Size  Code  Name
   1         64       8191  4064K
   2    7086080   15472639  4095M

.

.

Partition #3 was the large one, intended as the local datastore of 924GB.

I've had to redo it with partedUtil to have it start on a multiple of 2048. I forgot the actual number it started on 😉

            Regards

0 Kudos
a_p_
Leadership
Leadership

The ISO image doesn't contain the fixed ESX-UI version, i.e. you either need to update your host with the latest patch bundle, or update the Host Client individually.

You can find the download links for either the patch bundle, or the Host Client  at e..g. https://esxi-patches.v-front.de/ESXi-6.5.0.html

André

0 Kudos
vrangler
Contributor
Contributor

Hi,

Yeah, I had fixed it from the cli.. But I had already patched the host with the latest patch set.. It is running 6.5.0 (Build 5310538) right now. Is that new enough?

It doesn't matter much though if you had already deployed with the current installer in the ISO and you don't notice the issue. But it's useful if you add more storage or when the latest iso/installer gets updated. So I'm glad this has some focus Smiley Happy

Maybe you can give some specific info (KB?) about how to diagnose (and fix maybe) a miss-aligned partition on an esx host?

Thx!

0 Kudos