Planing for Backplane Failure with a ZFS Pool?

[unholythree]

I'm starting to outgrow my current storage a 4 disk stripe of mirrors zpool; so better late than never I'm jumping on the SE3016 bandwagon. I'd eventually like to build a new array of 6 to 8 disks in a raidz2 but I can't decide on the safest arrangement of disks.

My current thinking is this: no matter how I arrange the disks there will always be at least two pairs sharing a backplane. If a backplane with two disks in my raidz2 pool fails I'm now badly degraded; however if I have three or more on that backplane the pool is faulted.

It seems that oddly enough it would be better for the pool to fault-out. I could then export the pool, move disks off the failed backplane, and import the pool intact. Does this like a reasonable plan, or is there something I'm missing?

 

[Silhouette]

Backplane failures can have different effects. They can kill your drives or just disconnect them without any permanent damage.

 

[patrickdk]

It is mostly overkill.

For a home system, the pool going offline isn't the end of the world. Just fix the backplane and reboot the server.

A lot of businesses don't protect against a backplane failure. With sas though, generally you will have two backplanes on each sas disk, limiting the failure to a whole shelf failure.

If you take the SE3016, remove the expander, you would have 4 backplanes sharing a common psu though.

 

[Aesma]

You can have a script or something that will alert you if the pool degrades, or shut down the server.

 

[_Gea]

I have had some mainboard, psu or hba problems in the past.
I have had many, many disk failures.

But I have never had a backplane failure.
- overkill -