Implement Syndrome-based decoding for EC

[xhernandez]

Reed-Solomon (the algorithm used in EC) supports a way to not only detect errors but to also know where they are and how to recover them. This part of the algorithm has some requirements in the encoding part to be able to use it. Currently EC doesn't support it.

A change in EC would be needed to implement this decoding algorithm.

[mchangir]

would this feature interact with or need to interact with bit-rot by any chance ?
would bit-rot need to be implemented differently for EC volumes using the EC specific error-detection and recovery mechanism ?

[xhernandez]

The recovery information is implicit in the encoded data. It doesn't need extra storage for any checksum, but it needs to have the data, even if it's invalid because most probably only a small portion of it will be invalid and this decoding algorithm can identify and correct it.

So it would be interesting to have a way to force bit-rot to return the read data even if it detects corruption (though with some mark to identify it). This is not strictly necessary but it would be very useful to increase recovery chances when multiple corruptions happen.

Basically the algorithm is able to correctly detect up to R errors (being R the redundancy level of the volume) and revcover up to R/2 errors if it's position is not known (i.e. we don't know which bricks have inconsistent data). If we do know which bricks are bad, we can recover up to R errors. So in this case bit-rot could help to improve recovery capabilities for ec.

[stale]

Thank you for your contributions.
Noticed that this issue is not having any activity in last ~6 months! We are marking this issue as stale because it has not had recent activity.
It will be closed in 2 weeks if no one responds with a comment here.

[stale]

Closing this issue as there was no update since my last update on issue. If this is an issue which is still valid, feel free to open it.