UNSOLVED
Possible to merge Disk groups on MD3620f?
Hi folks
I am building a DR site for Veeam VBR replication target, idea is to minimize amount of administration required by giving DR site one massive CSV instead of bunch of smaller ones. Small LUN groups means that we have to manually pick target repository every time we modify replication jobs.
Our environment:
3x PowerEdge R620 with Failover Cluster Manager setup
1x MD3620f with 2x MD1200 shelves attached, unfortunately with varying disk sizes as these devices are being reused from previous production environment. Connected to all 3 hosts via redundant HBA fibre cables.
4x different disk groups because of different HDD sizes - 8TB, 13TB, 27TB,11TB, totaling 60TB.
Server 2012R2 datacenter edition, so S2D is out of question, Storage spaces does not support FC.
The problem:
What are my options to create a single storage pool combining all enclosures into one massive disk that is accessible by all 3 hosts and that will support Failover Cluster Manager (Hyper-V)? I have spent almost 2 days on the problem and can't find any solution to this. Would Starwinds VSAN work for this setup? Perhaps there's a better solution from Dell for this? I know Compellent storage would fix this, but we do not have budget for that.
Edit: Perhaps something like Scale-out technology used in Veeam, but for LUNs mapped on Windows hosts?
Thank you
Hi Sam, I am just trying to figure out a way to combine all space together into a single pool that shows up as single massive drive on all nodes.
If I can't do this on MD3xxx system, perhaps there's an official way to do it on host?
Storage spaces works for combining space, but only for a single node, the virtual disk cannot be part of failover cluster.
captain_black
4 Posts
782
0
Posted September 3rd, 2019 03:00
Starwind VSAN is a stable and performing solution for high-availability, but unfortunately it cannot present different underlying physical storage as a one virtual disk.
If your main goal is to create on large CSV and you don’t care much about performance and stability, you can use following “workaround”:
- Deploy clustered Windows server VM on shared storage.
- Add all your physical capacity as fixed vhdx to this VM.
- Inside VM create Storage Spaces.
- Create Starwind virtual disk of whole capacity on Storage Spaces.
- Connect Starwind vdisk to 3 physical servers via iSCSI.
You will have one massive CSV. Disadvantages of such deployment:
- If host server with running Windows guest VM will fail, you’ll need to wait until the VM will boot on one of remaining hosts (approx. 1 – 2 minutes CSV will be inaccessible).
- Performance decrease (it should be tested)
As for me I wouldn’t deploy such mess.
DELL-Sam L
Community Manager
•
8147 Posts
•
34706 Points
866
0
Posted August 29th, 2019 11:00
Hello HWIT,
With an MD3xxx system you can’t merge disk groups. What you will have to do is to back up all your data and then delete the current configuration & then create the new configuration & restore the data. When you are using disks pools all drives in the pool need to have the following requirements found on page 97 of the Administrator guide.
Disk Pool Restrictions
CAUTION: If you downgrade the RAID controller module firmware version of a storage array that is configured with a disk pool to a firmware version that does not support disk pools, the virtual disks are lost and the physical disks are treated as unaffiliated with a disk pool.
Please let us know if you have any other questions.