Tuesday, May 15, 2012

Estimating space for Hyper-V snapshots

A forum post inspired me to not directly answer a question this morning, but rather explain something so that folks can think about it instead.

The original post asked the following:

3 nodes Windows server 2008 R2
I have 3 nodes Hyper-V Cluster CSV in fiber channel with a multiple volumes.
1.What the amount of free space, I should keep the volume csv not to fail for lack of space.
2. Can I create a new volume and configure snapshots to be saved only in this new volume?
3. What is the recommended size for the other volume to store the snapshots.

I see two ways to answer this -

  1. directly answer question #2 and wait for the individual to fail
  2. Attempt to answer #1 and #3 (which estimation here is nearly impossible – due to chaos and randomness).

I jumped straight to a description of the technology – which I have not blogged about for some time, but it just seems to linger.

Here is my response:

You as asking questions about how to size a snapshot volume as if you have a background with ESX / VMware.

A snapshot is a moment in time - a differencing disk and configuration. Not a location where copies of VMs are saved off to.

The disk portion of a snapshot uses a differencing disk. Each differencing disk has the potential to grow to the full size of its parent. If its parent is a fixed disk, it will get that big - if it is a dynamic disk it can reach the maximum.

So, in theory - one snapshot can double the amount of storage required for a VM.

Also, a snapshot location is a runtime location. Where the snapshot is located is where the VM is now running from - until that snapshots is deleted and the snapshot is merged back in.

Knowing this, don't plan on using a slower or different class of storage - unless you are very specific and consistent about how you use snapshots and take this into consideration.

By default snapshots are in the same location as the VHD (and as the VM) - Hyper-V (the product) implemented this model with the R2 release because folks were getting into all kinds of trouble by separating the VM snapshots from configuration from VHD and when it came time to upgrade storage or migrate the VMs wound up broken.

So, you can separate the locations, but you need to be clear on how that benefits you. It is more about your process and procedure than the technology.

You might find this useful: http://itproctology.blogspot.com/search?q=snapshot

On a side note:

There is a growing swell in the Hyper-V forums of questions where folks just want to do something but are not taking the time to understand the implementation, they just push on and have problems.  Back in the day we used to refer to this as RTFM.  But you can’t say that in a forum.  But I plead with you, read the manual.

No comments: