Jump to content

Recommended Posts

Posted

Hi all, well, them that are still reading despite the holidays!

 

I'm doing some server jiggery over the break, converting a physical server into a Hyper-V host that will host 1 x Spare DC and 1 x Virtualised backup location, for resilience needs. Discussed here - http://www.edugeek.net/forums/windows-server-2012/207910-adding-physical-dc-2.html

 

I'm trying to work out how to best use my space, without losing too much performance. I have available 1 x 1TB RAID5 array, and 1 x 2.7TB RAID5 array.

 

My plan at the moment is:

 

1TB RAID5

  • 100GB Hyper V host OS partition
  • 900GB Hyper V guest storage partition, containing:
    [*=2]Spare DC full VM
    [*=2]Backup Host C drive
    [*=2]All VM config files

 

2.7TB RAID5

  • Backup Host D drive, aka, backup destination

 

I know in an ideal world, you'd just have one big array containing all Hyper V disks and configs (and on RAID10 not 5), but this is the physical server I have available. The key point is to maximise the space available for backups on the backup location server.

 

I've been reading around, but I can't find an argument one way or another about whether it's a bad idea to have VMs with disks on different arrays. It feels kinda untidy, but also seems a shame to use up valuable backup space on the second array when I've got spare capacity on the first. Alternatively I guess I could put the DC on the 1TB array and stick all of the Backup VM materials on the 2.7TB array, and expand that array sooner rather than later, but that would involve spending money....

 

Alternatively, is something like Storage Spaces an option, where I feed it my two arrays and it will give me all the space in one place? Or is that a recipe for failure...

 

Any other Hyper V gotchas also welcome, I'm learning on the fly on this one (nothing new there...)!

Posted

I know you said this is the physical server you have, but is there anything that stops you using another raid config on it still? (Like do the controllers only have 5?) I'd much rather use something like 6/10 and have the spare drives for failing, than the extra space especially if you're running a DC/hyperV host etc on it

 

Generally I wouldn't split a VM files across 2 controllers (assuming that is 2?) as you've got 2 points to fail etc, if you got multiple backup jobs split them across the 2 disks/arrays but keep 1 VM to 1 side

 

What I'd have done though is look at attaching the 2.7TB and 800GB~ to the HOST as drives, and then just use the backup VM to point to those 2 locations (Not attached as VHDX to the VMs etc), assuming whichever software you're using allows it. (Something like Veeam you could add both locations in to a single pool and let it backup as it wants to each)

 

If that makes sense :p

 

(As a disclaimer, I'm assuming when you say backups you do mean backups rather than replicas?)

 

Steve

  • Thanks 1
Posted

Nothing to stop me using a different RAID config, but it wouldn't get me anything better due to the disks available. 3 x 500GB OS and 4 x 1TB for Data. Wouldn't have been my spec, I'd have gone for one big RAID 5/10 array with two partitions, but I can see where they were coming from.

 

Both arrays are on the same controller, but yeah, it does add extra points of failure.

 

Utilising the rest of the storage directly on the physical host is an idea, and would be the most space efficient I guess. It's a bit against the spirit of virtualisation though, shouldn't have anything bar the hypervisor running directly on the hardware and all that.

Posted
(As a disclaimer, I'm assuming when you say backups you do mean backups rather than replicas?)

 

Oh yes, totally. That server exists purely for file storage for my Veeam backup files. I think it installs a Veeam agent on there too, but nothing too heavy.

Posted

Ah assumed it'd be multiple lower sized disks :p

 

If it's a backup location though I'd argue there's no benefit to virtualising it (think like a NAS etc), else you're effectively going to put a backup of a VM onto another VM? What's that give you extra? (I'd still run the actual backup VM as a VM, just not it's storage space). Can't think of one benefit really doing that. There's no benefits with backups because this is the backup, no migrating/redundancy as again this is the backup. No need to extend/shrink VHDs. Unless I'm having a dumb day? :p

 

Veeam wise, Aye in that case I'd suggest using either a scale-out option and just push both C/D into the one pool, or mapping 2 jobs to the individual drives then. While you could go storage spaces route it's adding in more bits for no real benefit.

 

Steve

  • Thanks 1
Posted
If it's a backup location though I'd argue there's no benefit to virtualising it (think like a NAS etc), else you're effectively going to put a backup of a VM onto another VM? What's that give you extra? (I'd still run the actual backup VM as a VM, just not it's storage space). Can't think of one benefit really doing that. There's no benefits with backups because this is the backup, no migrating/redundancy as again this is the backup. No need to extend/shrink VHDs. Unless I'm having a dumb day? :p

 

Haha, no, you're making perfect sense. I guess using the disk space directly rather than through a VM would be more pragmatic, and a better use of limited disk space, I was just getting into the spirit of the thing really, all the arguments I read tend to be along the lines not of "why would you virtualise", more "why wouldn't you?".

 

Yes, the actual server in charge of running backups remains a VM, on a different server.

 

I'd be interested to hear if any virtualisation purists have any arguments against using the space on the Hyper-V host as direct storage!

  • 2 weeks later...
Posted
The only issue with RAID 5 on the 1TB drives is the rebuild time if one failed. It does take some time to do on spinning rust and it another drive goes than you have lost the RAID. I would said RAID 10 on them and only have 2TB Storage.
Posted
The only issue with RAID 5 on the 1TB drives is the rebuild time if one failed. It does take some time to do on spinning rust and it another drive goes than you have lost the RAID. I would said RAID 10 on them and only have 2TB Storage.

 

RAID10 would be nice, but in our budget range, a bit of a luxury.

 

Plus I really do need to maximise storage space.... I'm at 1.8TB with backups already, 2TB would leave no wiggle room.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...