I have just realized a fundamental problem, with no obvious solution

[ullbeking]

Hello all,

 

I was setting up a my favorite SBC of all time, the Orange Pi +2E, and I can't believe that I've been so oblivious to this issue for so long, and that there is no easy solution!!!

 

OK, so I'm doing a massive hardware clearout, and I've getting to my OPI's.  Part of this hardware clearout is to finish the project I started (if feasible and still useful).  I have loads of approx. Intel DC 3000 series SATA drives around 80-200 GB, and I thought these would be perfect "user data" SSD's for each of the SSD's.  But then I realized... and my heart sank as I realized this issue is a problem across most OPi boards after consulting the table.

 

The only way to directly connect an SSD to my beloved OPi +2E's is via USB 2!!!  If it were USB 3 then that'd be fine, but USB 2 is going to make the SSD perform slow as molasses, relative to normal operation, isn't it?!

 

HUH!!!  How about that!

 

The only other option for local storage is eMMC, which is a paltry 16 GB.

 

This leaves Ethernet.  OK, cool.  It's common, for example, to access one's own "user data" via NFS v3 or v4.

 

By the way, I think the jury is still out on whether it's worth moving on up to v4 now -- still after all these days, is it?

 

So really, if I have a pile of the same model SBC, and I want to set up an HA cluster, then I can do so.  But storage for "user data" is going to have to come from plugging each cluster node into a switch giving access to an HA NAS.  Is this right?

 

Moreover, there's no effective way that I can use hyperconvergence with one disk associated per node, because there's no way of attaching fast or durable storage to each node.  I know that this is kind of the point (in the case of storage that fails because it's not durable, i.e., eMMC or SD flash card), but the failure rate will be so high that I estimate I'd need in the order of 100 SBC's iust to maintain availability... and performance won't ever be that great.

[Werner]

Moved to Common issues / peer to peer technical support