Future: n-way mirroring

Features
1

When raid5/6 goes stable and is implemented, a lot of users including me are going to be happy.

However, we want to keep pushing the envelope and thus here is a feature request for the future.

N way mirroring. 
Essentially n = the amount of copies btrfs will spread throughout the array.
Have 7 disks? Want 6 copies? N=6. Of course, this impacts available space heavily.

Also, note this request is not about file level n way mirroring, but subvolume/raid level mirroring.
A recent discussion on the mailing list showed that the calculations for file level n way mirroring are expensive and the math gets incredibly complicated fast!
2

When I first saw this, I thought of Share level n-way mirroring. Have you checked out 2-way mirroring of Shares with the 3.6-4 update? We’ve designed it to support n-way mirroring. We’ll enable that in future. But I digress from your topic.  Perhaps your request is appropriate for a discussion on btrfs mailing list.

Regarding raid5/6. I’ve started using it on 3.19 kernel. I am looking forward to btrfs-progs update and making 3.19 default on Rockstor.

3

Suman, thanks for you reply.


Mirroring of shares is not something that I knew about, that is a very nice feature indeed!

You are correct that this is about core-btrfs development.
There was a discussion about this on the mailing list with a comment by Hugo Mills here.

This comment by Roger Binns sums up nicely what this request is about:
"What people really mean is that they want their data to survive the failure of N drives - exactly how that is done doesn’t matter."

I mentioned it precisely for this reason, so that we may look towards the future with anticipation!
It will be an awesome feature for Rockstor to have when it gets released, which admittedly may take time.