• A Great friend to the HardForum with a great kid that he is trying to get a scholorship to continue his schooling. Please give hime a vote! Only 24 hours left! Thanks.
    If you have an VOTE FOR KEENAN!

ESX: one server, 24 drives, go openfiler?

T

Thuleman

Supreme [H]ardness
Joined
Apr 13, 2004
Messages
5,833
I have a stand-alone server, 24 internal drives (750GB each), and I need to run Samba (or SharePoint, down the road) as VM to allow access for about 10 clients to share files among each other. Ideally I'd like to set up VDI for those 10-ish clients, but that will have to wait.

Anyway, would you create ~500GB LUNs on separate arrays at the RAID controller level, or would you just set up say 4 drives in RAID10 for the VMs, set up one VM with openfiler, and then use OpenFiler and iSCSI to present the rest of the storage space to the VMs?
 
You shouldn't need openfiler to give the VMs access to the storage, just outside clients.

One thing to look out for though. Make sure to carve out the luns at the controller level as <2TB. ESX does not like volumes over 2TB.
 
That's the core of the issue. I can create VM-friendly LUNs on the controller level, that would be a whole bunch of 500GB LUNs. I could use RDM, completely bypassing vmfs (not sure why I didn't think of this during my first post), or I could just create two (two controllers) ginormous RAID6 arrays, give all that to OpenFiler, and let it carve it up into smaller VMFS friendly LUNs. Heck, since it's all virtual anyway I could even create multiple iSCSI targets, one for each 500GB LUN, or whatever. I am just making this up as I haven't used OpenFiler yet.

The only reason I even bring OpenFiler up is that I figure there will be more flexibility using it rather than creating the LUNs on the controller level, no?
 
geiger said:
You shouldn't need openfiler to give the VMs access to the storage, just outside clients.

One thing to look out for though. Make sure to carve out the luns at the controller level as <2TB. ESX does not like volumes over 2TB.
Click to expand...

And it's a rare case to make a datastore that size. Unless there is a big requirement we stay at 500GB or less. You want to stay at a max of about 15 VMs per datastore as well.
 
Thuleman said:
That's the core of the issue. I can create VM-friendly LUNs on the controller level, that would be a whole bunch of 500GB LUNs. I could use RDM, completely bypassing vmfs (not sure why I didn't think of this during my first post), or I could just create two (two controllers) ginormous RAID6 arrays, give all that to OpenFiler, and let it carve it up into smaller VMFS friendly LUNs. Heck, since it's all virtual anyway I could even create multiple iSCSI targets, one for each 500GB LUN, or whatever. I am just making this up as I haven't used OpenFiler yet.

The only reason I even bring OpenFiler up is that I figure there will be more flexibility using it rather than creating the LUNs on the controller level, no?
Click to expand...

cannot use RDM on local storage technically ;)

Even if you make the RAID6 arrays, openfiler will still be a VM and have the same limitations as any VM, not to mention that you then have to deal with a virt. layer for your storage.

OpenFiler also does Luns - don't need to do targets past the first :D. It's pretty freaking solid tbh.

Performance will be better with local VMFS, you just have to decide if you can deal with multiple local datastores/etc.
 
I'd personally do 8x 3drive RAID-5's, which would give you 1.5TB usable x 8 = 12TB RAW, formatted VMFS3. Whatever space you need for shares, take some of the RAID-5 sets and give them to the openfiler VM to serve up as NFS storage. Or you could make 2x openfiler VMs, one for NFS and one for iSCSI targets (which could be used by software initiators on other VMs, if you needed block level storage).

I would not recommend mixing/matching on what type of RAID sets you set up, just to keep things simple, I would make them all the same. There's no technical reason behind my thinking, but it would serve to keep at least part of the equation simple.
 
sabregen said:
I'd personally do 8x 3drive RAID-5's, which would give you 1.5TB usable x 8 = 12TB RAW, formatted VMFS3. Whatever space you need for shares, take some of the RAID-5 sets and give them to the openfiler VM to serve up as NFS storage. Or you could make 2x openfiler VMs, one for NFS and one for iSCSI targets (which could be used by software initiators on other VMs, if you needed block level storage).

I would not recommend mixing/matching on what type of RAID sets you set up, just to keep things simple, I would make them all the same. There's no technical reason behind my thinking, but it would serve to keep at least part of the equation simple.
Click to expand...

If you're going to do NFS, do FreeNAS. much better NFS performance :)
 
Are you dead set on an ESX implementation?
I only ask because you say you want to do sharepoint and need lots of storage.

Server08 could work pretty good for ya.
Install on a single drive, and then have a massive R10 array for the vm's.
 
Server08 would work, I am actually running Hyper-V on a different box right now. It's certainly trivial to set up RDM on Hyper-V, I had forgotten about that. I could just create whatever LUNs are needed on the controller level, not be subject to the 500GB VMFS guideline, and be merry. Since it's a stand-alone server I don't need it to be ESX, I was just gonna go that route to keep some consistency.
 
It sounds like Server 08 would give you the most consistency and the easiest storage setup.

Some may say vmware just for the sake of vmware, but I would really use the best tool for the job. Remember though there are licensing costs associated with Server 2008 vs. vmware. However if you are going to install sharepoint in a server 2008 VM it would fall into the same license so no additional costs (1 physical 4 virtual).
 
You must log in or register to reply here.
Back
Top