I recently got some checksum errors on my btrfs filesystem that seem
"interesting". This is on a F18 kernel (3.6.0-3.fc18.x86_64) with the
filesystem on a single INTEL SSDSA2CW16 disk (ssd). Its my workstation,
so i do normal development on it, i.e. its not a server. Recently I have
been doing some virt work, so I''ve been running qemu a lot. I also for
some time had a near full filesystem (98% full, which caused some long
100% cpu use stalls in flush-btrfs-4) but have now freed up some space.
Today when i was copying a VM image i got an I/O error. Turns out that
two of my VM images had checksum errors in them. A full scrub run showed
that these were the only errors on the entire disk. Logs attached.
I was able to copy the win7 image from the log with dd_rescue, but it
had (as can be seen from the log) 8 checksum failures of 4k blocks at
different places in the file.
I''ve now got a working system again as no other files were affected.
However, this doesn''t seem like a hardware error to me, so I''m
worried
that there is some underlying issue in btrfs that caused these checksum
errors.