expansion and failover

[jchen522]

Connect and Activate the Expansion Shelf

https://support.hpe.com/hpesc/public/docDisplay?docId=sd00006088en_us&page=GUID-C001824F-524C-4903-AB06-42EBA0BF68B0.html

Please see diagram. 
We currently have A, B, C in production and want to connect and add D into the expansion.

I was wondering if someone can explain how does expansion and failover work in this scenario. 

What is the procedure for adding D into production?
How many units/controllers/path can fail and still stays up?
What happens if unit B totally dies?  There doesn't seem to be a connection for SAS D going back to A in a loop.  Or it doesn't work that way?

Thank you.

[Azr_geek]

Hello @jchen522.,

NOTE: The answer is given by my colleague and is positive. I am not an expert myself.

A is the two controller ports at the head unit, B and C are the two middle shelves in each chain, and D is the new expansion shelf you want to add below C. The cables form two separate SAS chains / one chain from the left controller port and one from the right controller port / and both chains pass through each shelf in order.

To add D you plug two SAS cables from the ends of the chains on C into D (one cable for the A‑side path and one for the B‑side path), power D up, and let the head controllers discover the drives.

Because there are two independent paths, the system can survive a single path or single module failure (for example if one side’s module or cable in a shelf fails the other side still provides access).

However, if an entire middle shelf (like B) loses both of its SAS modules or power, it can break both chains at that point and any shelves downstream (C and D) will lose access. In short: connect D to both A‑ and B‑side cables on C, you have two paths for redundancy, you can tolerate one path failing but not both paths through the same shelf.

Regards,
Azr_geek

[jchen522]

Hi Azr_Geek,
Thank you and your colleague so much for the detailed explanation. There is no alternative way of cabling that will solve the entire middle shelves failure issue correct?

[Azr_geek]

@jchen522,

Mostly correct, you cannot completely eliminate the risk that a single middle shelf failure will make the shelves beyond it unreachable unless you provide an alternate physical path that bypasses that middle shelf for both SAS domains. In practice, that means either the vendor‑supported dual‑domain/loop cabling is used correctly, or you add additional supported physical links that let the controllers reach the downstream shelves even if the middle shelf’s backplane/IOMs/power fail. If both domains become broken by the same shelf failure, nothing in software can route around that; you need hardware cabling redundancy or a different physical topology.

Regards,
Azr_geek