Hugo Mills writes: > On Mon, Apr 08, 2019 at 11:48:03AM -0400, Scott E. Blomquist wrote: > > > > Hi All, > > > > The weekend btrfs scrub/balance came back with this following... > > > > [Sun Apr 7 06:57:10 2019] BTRFS warning (device sdb1): checksum error at logical 274820497408 on dev /dev/sda1, sector 536758784, root 271471, inode 109421914, offset 491520, length 4096, links 1 (path: yyyy/yyyyy) > [snip] > > Since there doesn't seem to be anything else wrong (no messages > without a filename, which would imply metadata corruption), this is > most likely a simple case of on-device corruption. > > Delete yyyy/yyyyy and restore it from backups. At least, do so in > the working copy; The snapshots of it can safely remain until they get > rotated out normally. > > Check your SMART statistics and see if anything looks wrong there > on the hardware side. Also check dmesg and earlier kernel logs for > signs of the hardware showing an error on read -- it may have tried > several times to read that location before giving up and/or returning > bad data. > > Hugo. Thanks, Hugo. Very helpful. Turns out event log from MegaCli is showing some unexpected sense in the eventlog. Cheers, sb. Scott Blomquist
