BitRot Protection Failure - Anything To Be Done?

Skywalker wrote:
It's certainly not an established fact that we have a false positive. That's what I'm trying to find out. I'm not exactly sure what you mean by "RAID repair and the bitrot detection collided", but if you mean that the RAID array re-writing blocks to force reallocation corrupts filesystem checksums, well, that doesn't appear to be possible.

I'm glad you are still looking at it. This is new territory for home NAS. It'll be interesting to see how often bitrot is detected and how often it is repaired. I'm hoping posters will continue to report the detection events (and the outcome).

-Jeff said he was able to view the file in a photoviewer, and it looked ok. So it seemed likely that is false, but of course it might have been real. And "false positives" are a bit concerning in their own right. Something triggered the detection after all.

-On "collided" I was thinking that when the processes overlapped there might be race conditions that might create a false positive. I didn't have an exact scenario in mind. But something along the lines of (a) a read error occurs on the checksum test (data or checksum itself), (b) normal RAID repair starts in parallel with bitrot repair.

It seems to me that there are potential issues in scenarios like that. Data that is potentially in memory is being changed on the disk by the competing process (the checksum being likely).

Or something simpler. For instance, maybe bitrot repair starts by confirming that the parity block is inconsistent. The main RAID repair perhaps just made it consistent. Then bitrot repair might think there is nothing it can do (though the problem was just fixed).