Recycling a server designed for Exchange

[tinto1970]

Hi all,

a couple of years ago we bought a Dell R720xd with 14 SAS 900 GB disks (+2 140 GB disks). The RAID controller has 1 GB of cache.
We have installed vSphere ESXi on the 140GB virtual disk (2 disks in raid1), then we created 2 virtual disk for the virtual machines, mainly a not as little Exchange 2010 MBX store:

- one v.d. is composed by 6 of the 900gb disks in RAID10 (2.500 GB of usable space)
- one v.d. uses 8 of those disk (3.300 GB)

The machine is performing very well but the company has decided to leave exchange for google apps for business.
So this machine can be used for data protection: i have a nice amount of disk space, the hypervisor already licensed and a Windows 2008 VM licensed too.

I guess i can use this 'material' both to host backup files and to hold some critical VM (replicated from the production vSphere cluster) to be started in case of disaster on the main cluster/storage.
My idea is to add some big .vmdk disks to a VM running on this ESXi (maybe a Linux VM or the Windows 2008 already present) and use them as backup repository and, maybe, install Veeam B&R on the Windows machine too (we have the enterprise edition of B&R).

My doubts are about the RAID configuration to use. I know there is not a "better option" because it depends on many things. I see a few options at the moment

- keep the current config
- create a single virtual disk in raid 6 with 14 disk (around 10 TB of usable space, write penalty 6)
- create two v.d. with 7 disks in raid 5 (around 5 TB per v.d., write penalty 4)
- create 1 v.d. with 6 disk in raid 10 (2.5 TB, write penalty 2) for replicated VMs, and one with 8 disk in raid 5 (6 TB, write penalty 4) for backup files

it's very difficult for me to have specs like "how much space you need" and "how many VM/IO you will be running in case of disaster" because i actually don't have still a clear backup policy to apply, so I understand it not as easy to give suggestions about this. Sure I'd like to use reverse incremental policy, which is impossible to use on the Dell DR4100 we are currently using as backup repo.

However i hope to have some comments, critics and suggestions... thanks in advance

[foggy]

My idea is to add some big .vmdk disks to a VM running on this ESXi (maybe a Linux VM or the Windows 2008 already present) and use them as backup repository and, maybe, install Veeam B&R on the Windows machine too (we have the enterprise edition of B&R).

I would recommend to carve a separate LUN, format it as NTFS, and present through iSCSI to Veeam B&R as backup repository (storing backups on VMFS is not considered as best practice).

My doubts are about the RAID configuration to use.

Here is a couple of topics regarding recommended RAID configuraion, worth reviewing:
Reverse incremental performance on Raid 5 vs Raid 10
RAID For VEEAM Replication

[dellock6]

Alexander, I don't see however an easy way to format a volume on an ESXi host other than NTFS and present it to a local VM. The only possible solution is to use RDM, but it depends if the raid controller supports the publishing of the volume as "remote". Alessandro, this kb should be helpful:
http://kb.vmware.com/selfservice/micros ... Id=1017530

Other than that, I agree on not over-complicate the design with additional layers created by the vmfs+vmdk+ntfs partitiona one on top of the other.

[foggy]

Luca, I missed the fact it is local storage, my bad. There's no iSCSI then, sure, and using VMDK seems to be an easier option.

Btw, here's another one on RDMs.

[tinto1970]

I would recommend to carve a separate LUN, format it as NTFS, and present through iSCSI to Veeam B&R as backup repository (storing backups on VMFS is not considered as best practice).

i'm reading the discussion and thinking about it. Sure the cons are to be taken seriously (as Luca wrote "you are saving VMs inside the technology whose faults you are trying to proctect from").
However it will be a separate, independent machine which can be located in another room. So it should be able to protect VMs from floods, fires, blackouts, human mistakes and so on.
Gostev's point abount the vmdk size limit is no longer existing.

I will probably test the RDM option you suggested, i don't know if finally i will use it or... do "the wrong thing"

Here is a couple of topics regarding recommended RAID configuraion, worth reviewing:

ok, it will be raid 10


Thank you, Alexander and Luca

[foggy]

i'm reading the discussion and thinking about it. Sure the cons are to be taken seriously (as Luca wrote "you are saving VMs inside the technology whose faults you are trying to proctect from").
However it will be a separate, independent machine which can be located in another room. So it should be able to protect VMs from floods, fires, blackouts, human mistakes and so on.
Gostev's point abount the vmdk size limit is no longer existing.

Sure, it is still a viable approach. Our intention was just to warn you about the possible consequences, so that you keep them in mind while planning the deployment.