• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 2623
  • Last Modified:

Storage Spaces, SOFS and Shared SAS JBOD Redundancy

I've spent many hours online reading different articles on Server 2012 R2 storage spaces so please refrain from posting links as I need more real world comments and experiences.

My goal is to have two servers for Scale-Out File Server (SOFS) clustering using storage spaces connecting to two JBOD enclosures with full redundancy.  I understand that you need a JBOD enclosure without any RAID abstract layer for the server and to get full redundancy you would need a shared SAS controller to allow each server a connection to each enclosure.

It seems like you also need to use the enclosure awareness feature to support the failover of an enclosure but for some odd reason it requires three enclosures (yes I've read the MS articles but it still doesn't make 100% sense).  The HCL list is also pretty sparse right now with only DataON, RAIDInc, Fujitsu and a few others officially supported.  http://blogs.technet.com/b/privatecloud/archive/2013/04/05/windows-server-2012-about-clustered-storage-spaces-issue.aspx shows an example of two servers and two enclosures but I don't get how it would be possible to have three enclosures.  All the sparse hardware I've found seem to only support two SAS connections per enclosure.  I'm assuming that the theory is supported by Microsoft but there's no easily obtainable hardware that exists yet.

Would it be possible to have each server connect to only one enclosure and then mirror the storage spaces enclosure on each sever?  If the SOFS is clustered over the two servers, shouldn't it be able to handle an enclosure failure?  Finding a shared SAS controller doesn't seem to be easy and I do not want to use a SAS switch as that is another single point of failure.

Has anyone tried storage spaces with a Tier 1 manufacturer's enclosure, ie Dell, HP and make it work in a JBOD?  I am hesitant to use DataON and RaidInc as I have no experience with these manufacturers and worried about support.
  • 2
  • 2
1 Solution
Cliff GaliherCommented:
"I'm assuming that the theory is supported by Microsoft but there's no easily obtainable hardware that exists yet."

Redundant HBAs, redundant SAS switches, and redundant IO controllers in the storage box. You can certainly get four servers connected to a jbod with redundancy.

"Would it be possible to have each server connect to only one enclosure and then mirror the storage spaces enclosure on each sever?"

No. The only way to guarantee that the mirrored data block is written to another disk in a different enclosure instead of another disk in the same enclosure is to turn on enclosure awareness. And doing so means the cluster will look for three enclosures to meet its minimum requirements. You can't sidestep this safely.

The way to avoid the single point of failure in the SAS switch is just like you would when teaming NICs for full network redundancy. instead of one dual port MIC, you get two NICs. And wire each to a separate switch. Similarly, two HBAs wired to two SAS switches can achieve full IO path redundancy.

Regarding tier 1 manufacturers, I know a few have tried it. It works. But it isn't certified and performance wasn't up to par. They are tuned for their own solutions, not storage spaces. So if you are concerned about single points of failure, making sure it all works as planned, and then choose absolution with limited/no support and uncertified hardware, I find that contradiction odd. But sure, it can be done,
convergintAuthor Commented:
I assume that it is at least possible to have two different storage spaces on two different servers each with their own enclosure?  What about using software to mirror a virtual SAN over the two storage spaces like Starwind?  

I guess I'm asking a bit too much at this point to have no single points of failure using all built-in Server 2012 R2 components with Tier 1 manufacturers to build a SAN with reasonable costs.  I hate how most SAN manufacturers lock you into their products and the high price for what you actually get.  I totally understand how Dell and HP don't really want to support this as it takes away from their propriety SAN solutions.

The only SAS switch I have seen is the LSI SAS6160, have you used that before?  Thanks for your honest input.
Cliff GaliherCommented:
While you could use 3rd party software to mirror data, by doing different storage spaces and mirroring, failover would not be as automatic nor would it fit my definition of "scale out" for purposes of a SOFS build. There are benefits and drawbacks to such a solution.

I've not used that particular SAS switch.
convergintAuthor Commented:
It's definitely not "scale out" and that's fine with me but it should be automatic failover as the Starwind will take care of the mirroring and the Windows Clustering will utilize the Starwind "SAN".  It does add another layer of complexity no doubt.

I know thousands of places use a SAN with Microsoft Clustering but I just hate the fact that the SAN itself is a single point of failure and unfortunately I don't have a budget to allow for two expensive SANs.

Thanks for the input.
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

Join & Write a Comment

Featured Post

Upgrade your Question Security!

Your question, your audience. Choose who sees your identity—and your question—with question security.

  • 2
  • 2
Tackle projects and never again get stuck behind a technical roadblock.
Join Now