On 20/05/2021 12:12, Roger Heflin wrote:
The read failures exist on the platter for the most part. Reading the platters or not does not change the fact that sectors are already bad. raid6 reading more sectors just means you have a higher risk of finding the already corrupted sector.
I think you're wrong there ... glitches in the electronics also seem to be not uncommon. But here, the read normally fails, and if the OS retries, it succeeds without trouble.
But as I look at it, if the manufacturers specify "less than 1 error per X bytes read", even if X in practice is much larger than X in the spec, you should plan to handle what the spec says.
Cheers, Wol
