VMware Cloud Community
RHinder
Contributor
Contributor
Jump to solution

Thick Disk Lazy Zero vs Eager Zero

I have been researching this subject and understand the differences between the types and that ESX 4.0 will use Lazy zeroing by default for anything from creating a new disk to cloning etc..

What I haven't been able to find the answer to is that if you create a new virtual disk as per normal "lazy zeroed" and format it in Windows NOT using the Quick Format tick box, wouldn't this force the zeroing anyway on the VMDK level, because the blocks are actively being written to for the first time by the format operation?

Thanks

R Hinder

Tags (4)
Reply
0 Kudos
1 Solution

Accepted Solutions
AntonVZhbankov
Immortal
Immortal
Jump to solution

It would. All blocks in "Lazy zero" are zeroed on first access.


---

MCSA, MCTS, VCP, VMware vExpert '2009

http://blog.vadmin.ru

EMCCAe, HPE ASE, MCITP: SA+VA, VCP 3/4/5, VMware vExpert XO (14 stars)
VMUG Russia Leader
http://t.me/beerpanda

View solution in original post

Reply
0 Kudos
8 Replies
AntonVZhbankov
Immortal
Immortal
Jump to solution

It would. All blocks in "Lazy zero" are zeroed on first access.


---

MCSA, MCTS, VCP, VMware vExpert '2009

http://blog.vadmin.ru

EMCCAe, HPE ASE, MCITP: SA+VA, VCP 3/4/5, VMware vExpert XO (14 stars)
VMUG Russia Leader
http://t.me/beerpanda
Reply
0 Kudos
a_p_
Leadership
Leadership
Jump to solution

There are however configurations where eagerzeroedthick is a must.

e.g. Fault Tolerance, MSCS,...

André

Reply
0 Kudos
rickardnobel
Champion
Champion
Jump to solution

It would. All blocks in "Lazy zero" are zeroed on first access.

Sorry for updating an older thread, but I am wondering if the blocks are being zeroed even when just read?

I did some testing with thin disks and that thin vmdk disk were not extended (zeroed) by reading all blocks, only when writing to them.

My VMware blog: www.rickardnobel.se
Reply
0 Kudos
AntonVZhbankov
Immortal
Immortal
Jump to solution

What's the point of allocating and zeroing blocks on read? Unlike thick disks thin ones do not contain any data underneath, so ESX can simply return block of zeroes.


---

MCSA, MCTS Hyper-V, VCP 3/4, VMware vExpert

http://blog.vadmin.ru

EMCCAe, HPE ASE, MCITP: SA+VA, VCP 3/4/5, VMware vExpert XO (14 stars)
VMUG Russia Leader
http://t.me/beerpanda
Reply
0 Kudos
rickardnobel
Champion
Champion
Jump to solution

What's the point of allocating and zeroing blocks on read? Unlike thick disks thin ones do not contain any data underneath, so ESX can simply return block of zeroes.

That was somewhat what I was wondering. That is, if there is a difference between a read to an empty block a disk on a lazy thick and a thin disk? It seems logical that ESX can "fool" the guest more when using a thin disk.

The reason for my question was the original posters question and your answer to that:

What I haven't been able to find the answer to is that if you create a new virtual disk as per normal "lazy zeroed" and format it in Windows NOT using the Quick Format tick box, wouldn't this force the zeroing anyway on the VMDK level, because the blocks are actively being written to for the first time by the format operation?

Since the formating in Windows has changed between Server 2003 and server 2008 it is a bit interesting how the zeroing works. When doing a full format in 2003 all blocks are only read and doing a full format in 2008 all blocks are written too.

My VMware blog: www.rickardnobel.se
Reply
0 Kudos
AntonVZhbankov
Immortal
Immortal
Jump to solution

>That is, if there is a difference between a read to an empty block a disk on a lazy thick and a thin disk?

Lazy thick is already fully allocated, so for any block on VMDK is already underlying physical block, and there is security problem. If you read this block there could be data from previous VMs. Thin disks have no underlying blocks unless you write to them.


---

MCSA, MCTS Hyper-V, VCP 3/4, VMware vExpert

http://blog.vadmin.ru

EMCCAe, HPE ASE, MCITP: SA+VA, VCP 3/4/5, VMware vExpert XO (14 stars)
VMUG Russia Leader
http://t.me/beerpanda
Reply
0 Kudos
rickardnobel
Champion
Champion
Jump to solution

>That is, if there is a difference between a read to an empty block a disk on a lazy thick and a thin disk?

Lazy thick is already fully allocated, so for any block on VMDK is already underlying physical block, and there is security problem. If you read this block there could be data from previous VMs. Thin disks have no underlying blocks unless you write to them.

Thanks for taking the time to reply.

So a 2003/2008 full format on a thick disk causes the whole disk (or partition) to be zeroed,

and a 2003 full format on a thin disk causes not extending/zeroing,

but a 2008 full format on a thin disk will extend/zero it. Smiley Happy

My VMware blog: www.rickardnobel.se
Reply
0 Kudos
kgottleib
Enthusiast
Enthusiast
Jump to solution

Hi Anton - I don't want to open another thread because this is related to this discussion:

When in the GUI, the migration option presents you an option to change disk formats to thick, but it doesn't indicate whether changing from thin to thick changes the disk to eager-zeroed or lazy-zeroed.. I realize I can make the change on the command line with vmkfstools, but this requires a power down on the VM. Do you know which disk format is used when this is done in the GUI? VMware does not have this mentioned in any of its documentation..

Reply
0 Kudos