On Mar 31, 2014, at 2:23 PM, Eric Jaw <naisa...@gmail.com> wrote: > Doing a scrub is just obliterating my pool.
Is it? I don’t think so: > scan: scrub in progress since Mon Mar 31 10:14:52 2014 > 1.83T scanned out of 2.43T at 75.2M/s, 2h17m to go > 0 repaired, 75.55% done Note the “0 repaired.” > I'm also running ZFS on FreeBSD 10.0 (RELEASE) in VirtualBox on Windows 7 > Ultimate. Are the disks that the VM sees file-backed or passed-through raw disks? > Things seem to be pointing to non-ECC RAM causing checksum errors. It looks > like I'll have to swap out my memory to ECC RAM if I want to continue this > project, otherwise the data is pretty much hosed right now. Did you actually run a memory tester (e.g., memcheck86), or is this just based on gut feeling? Lots of things can manifest as checksum errors. If you import the pool read-only, do successive scrubs find errors in different files (use “zpool status -v”) every time, or are they always in the same files? The former would indeed point to some kind of memory corruption issue, while in the latter case it’s much more likely that your on-disk data somehow got corrupted.
Description: S/MIME cryptographic signature