0

In our VMware infrastructure, we have a number of large Windows file servers, with provisioned disk sizes ranging between 25TB and 35TB.

Their individual disks are rarely larger than 4TB, with many being 1TB, 1.5TB, or 2TB.

We're in the process of migrating from a number of older, aging storage arrays (IBM DS5300, HPE 3Par 7400), to a single, newer one (HDS VSP G600).

For reasons beyond my control, my organization has decided that any volumes created on the new array must be no larger than 4TB. I've tried (many times) to have discussions with them about the reasons behind that policy, but that avenue simply isn't one that leads to anything useful.

While there is an exemption process, and for the VMDKs which are at or above 4TB I've been successful in being granted one, I haven't been as lucky for the disks which are smaller than 4TB.

The storage team is saying that it's our problem; they'll provision a number of 4TB volumes from the array, and we'll have to make it work on our end.

Again, not ideal, but it's the hand I'm dealt.

I see a few options here.

  1. Create a single VMware datastore per VM with multiple extents.
    This seems like the simplest option, but I have lots of reservations about doing it.
    I've read everything I can find on extents, but there's nothing recent on the subject, and what articles I do find seem to suggest that they're to be avoided at all costs.

  2. Create Datastore cluster per VM, and let sDRS manage it
    This option feels better, and more "current" than using extents, but it still leaves a lot to be desired. There's still the headache of trying to sort out which VMDK is on which datastore when more space is requested, but at least there's the comfort in knowing that it's a 1:1 mapping of array volumes to datastore, like the rest of them.

  3. Create individual datastores, and manage it manually
    This is much the same as option #2, without the benefit of having sDRS manage placement of new disks. The only advantage I can see here, is that it's a static mapping of volume/datatore to VMDK, which makes growing disks, or tracking down problems easier.

I'm currently doing option #3 with the existing volumes/arrays, but many of the datastores are larger than 4TB, so it's been pretty easy to manage.

The only other thing I'll mention, because it might affect what the options are, is that our backups use VMware vSphere Storage APIs – Data Protection (aka VADP) with CommVault and NetBackup. I haven't been able to find anywhere that extents aren't supported with either of those products, but it's also doesn't say that it is.

What do you folks do in situations like this?
Is your storage team more reasonable and willing to give you larger volumes, or do you use extents?

I know it's a bit of a broad question, but hopefully not so much that it's off-topic.

1

I'd go with option 2 - sDRS - we use it pretty exclusively and it works a treat. Obviously the manual option is fine too, just more work for you.

What I wouldn't do is use multiple extents unless you absolutely have to, it was always something of a cludge and I don't personally trust it long-term.

What I would strongly urge you to investigate with them is VVols, if I recall correct the HDS VSP G600 supports them and it's a far better solution than LUNs+DS's anyway. See if they'll consider that.

  • 1
    Yeah, I feel the same way about extents as you do, and really want to avoid them. I hadn't thought about VVols, but I'll check to see if they'd be open to a discussion about that. It might greatly simplify everyone's lives. – GregL Nov 19 '19 at 17:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.