Re: BTRFS corruption after conversion to block group tree

[Russell Haley <yumpusamongus@gmail.com>]

To be clear, there is no reason to be suspicious of the other disk that
was converted to block group tree and has passed a scrub after
rebooting? It should be safe to mount that one read-write again?

On 12/1/23 22:25, Qu Wenruo wrote:> Just one more question, is there any
hibernation/suspension involved in
> this particular corruption?
> 
> I have seen exactly such unexplainable writes-from-future cases, which a
> lot of them have hibernation/suspension involved.
> (Thus personally speaking, I never go hibernation/suspension on my own
> devices)

Neither of those, but the affected disk *is* set hdparm -B 128 -S 240,
which is Advanced Power Management set to the lowest-power value that
"doesn't permit spin-down", AND a 20 minute spin-down timeout. It's
possible that this contradictory combination causes firmware
misbehavior, but this configuration has been in place since late
December 2021 and there were no problems for almost 2 years.

Unfortunately my notes don't say exactly I combined no-spindown APM with
a spindown timeout, but it does for sure result in the drives spinning
down after the configured duration of complete idle. Judging by the data
I gathered before making that decision, the reason was probably that
APM<128 gave different spindown timeouts for my 4 assorted hard drives,
and there was no way to discover the mapping of APM levels to timeout
durations.

Needless to say, I have reverted that configuration and will just live
with the extra noise and $20/year.

Thanks,

 Russell Haley