Basic or Dynamic disks

[localzuk]

So, I discovered something interesting last night - VSS does not allow on the fly backup of live virtual machines using, say, BackupAssist (which uses Windows Backup) if you use Dynamic Disks on virtual machines.

 

So, the question is - what do people do in this situation then? I've got a script that supposedly can make a clone of a live machine, which can then be backed up as it is 'off' but this isn't exactly an efficient way to do it.

 

Do you use dynamic disks or basic disks, or do you use something which has its own way of backing up live machines?

[synaesthesia]

Using basic here. Also about to get rid of BackupAssist which is absolutely smegging diabolical on server 2012. No proper backup schemes, no ability to use incrementals or rolling backup schemes on anything but removable discs. I'm swearing at the useless gits by email now.

[localzuk]

I'm using incremental backups on 2012, with a proper backup scheme too (daily incremental, weekly/monthly/yearly full). You have to manually create the scheme IIRC.

 

With basic disks, isn't there the issue that you then need a heck of a lot more disk space straight away? IE. We use 1TB of space for all our servers. We'd need to double that, minimum, if we used basic disks!

Edited February 24, 2014 by localzuk

[synaesthesia]

7.4.2 wont let us create any schemes outside of it's current scope and BA support have told us to like it or lump it

[localzuk]

That's bonkers!

[synaesthesia]

Right, I told a small fib - whilst BA is crashing in the background as I clearly make it sweat, taking a look at disk management shows our C drive and shadow copies drives are Basic and the actual file stores are Dynamic.

[john]

Basic disks for everything, not used Dynamic for years now.

[sparkeh]

Dynamic disks here with Veeam to back them up, no issues with backing up live machines.

[tmcd35]

I only ever use Dynamic disks. But then I arrange downtime during holidays to do offsite backups of the VHD's. All onsite data backups is done using traditional backup software running inside the VM (Yosemite).

[sparkeh]

I only ever use Dynamic disks. But then I arrange downtime during holidays to do offsite backups of the VHD's. All onsite data backups is done using traditional backup software running inside the VM (Yosemite).

Oh yeah, Veeam can create a replica of a live VM so I get VHDs copied across to a standby Hyper-V server. I can copy these off for off-site backup at my leisure.

[SYNACK]

With basic disks, isn't there the issue that you then need a heck of a lot more disk space straight away? IE. We use 1TB of space for all our servers. We'd need to double that, minimum, if we used basic disks!

@localzuk - possibly dumb question time, why would it increase your disk usage, we use all basic disks to? I'm assuming you mean partition type rather than static (preallocated) or expanding VHDs/VHDX/VMDKs?

[localzuk]

@localzuk - possibly dumb question time, why would it increase your disk usage, we use all basic disks to? I'm assuming you mean partition type rather than static or expanding VHDs/VHDX/VMDKs?

 

I'm on about static/expanding VHDs. The terminology Microsoft uses is 'basic' and 'dynamic' disks for the VHDs...

[SYNACK]

I'm on about static/expanding VHDs. The terminology Microsoft uses is 'basic' and 'dynamic' disks for the VHDs...

 

Ah, yes I have been guilty of using expanding VHDs on occasion but it is TERRIBLE practice, if the drives grow too big for the drive you can bring down all the vms on that volume and end up in a nasty, possibly server loosing condition where they are all paused because of disk space and you have nothing to delete so can't unpause them to shut them down.

 

The other thing is speed, dynamic/expanding VHDs are shockingly slow in comparison to everything. Static VHDs re very close to as quick as native.

 

EDIT: looked it up, still slower but they fixed much of the horror in 2k8R2 for dynamic speed Hyper-V pass-through disk performance vs. fixed size VHD files and dynamic VHD files in Windows Server 2008 R2 | Clustering For Mere Mortals

 

Still static is better practice, I would at the least have a nice big place holder file or two like a couple of ISOs on the drive with the dynamic VHDs so you have something to delete to allow a hasty shutdown and move incase one server goes off into the weeds with storage.

[localzuk]

That sort of problem is easily dealt with by having a suitable server monitoring/alert system in place. If our disks that hold the VHDX files hit 90% full (meaning we have 150GB left) we get notifications.

 

Performance is not a key driver for us - especially at the levels seen in that article.

[tmcd35]

Ah, yes I have been guilty of using expanding VHDs on occasion but it is TERRIBLE practice,

 

Got to say I seriously disagree with this! IMHO, static drives should only be used if a) you can 100% guarrantee the maximum storage requirements, and/or b) performance is of primary concern (SQL servers?). @localzuk is right, expanding VHD's should be managed by monitoring the capacity of the storage pool. In deed, it's what Storage Pools in Server 2012 where designed to cover.

 

I thought hardware decoupling was a key benefit of virtualisation.

[SYNACK]

Got to say I seriously disagree with this! IMHO, static drives should only be used if a) you can 100% guarrantee the maximum storage requirements, and/or b) performance is of primary concern (SQL servers?). @localzuk is right, expanding VHD's should be managed by monitoring the capacity of the storage pool. In deed, it's what Storage Pools in Server 2012 where designed to cover.

 

I thought hardware decoupling was a key benefit of virtualisation.

 

Only in the new versions, the old ones had those issues and none of the benefits. That kind of space management is fine as long as there is always someone around to check, an email is good but a once a week tech may not be able to manage things so closely. It also depends on the subscription o your hadware. I have seen schenarios where thereis only 300GB to go around on the drives with no money to update so tight management and control was required. There are resons why such things can be required and I had not read up on that particular bit since Hyperv 2k8.