ZFS data errors introduced when reading dataset, scrub increases data errors

[naisanza]

ZFS checksum errors grow increasingly when the datasets are scrubbed, leading to unrecoverable data errors. ZFS checksum errors are introduced when copying files out of a dataset, leading to unrecoverable data errors (e.g. rsync over ssh srcZFS dstHOST)

Windows HDD Write-caching policy was set to: [x] Enable write caching on device [ ] Turn off Windows write-cache buffer flushing on device

Guest settings attached Guest log attached

zpool status can be found, here: ​https://forums.virtualbox.org/viewtopic.php?f=6&t=60975

Specifications: + i7 970 + OCZ Reaper HPC DDR3 PC3-12800 (ocz3rpr1600lv2g) 3x2GB + Corsair Vengeance (cmz12gx3m3a1600c9) 3x4GB + Gigabyte ex58-ud5 + HX850W PSU + GTX780 + 2x Intel SSD (RAID0) + 6x Toshiba 3TB HHD

+ Windows 7 x64 Ultimate + Avira AV + VirtualBox 4.3.10 r93012 (upgraded from previous version last night)

[naisanza]

ZFS VirtualBox Guest Settings

[naisanza]

ZFS-2014-04-01-14-11-09.log

[naisanza]

The vmdk's used by the Guest OS were created with the command:

vboxmanage internalcommands createrawvmdk -filename "C:\PhysicalDrive.vmdk" -rawdisk "\\.\PhysicalDrive"

[naisanza]

I have turned off Windows Write-caching policy, and am testing for checksum and data errors by scrubbing the pool

[off] Enable write caching on device 

[off] Turn off Windows write-cache buffer flushing on device 

Edit: checksum errors still occur

Proceeding to enable Cache Flushing in VirtualBox

VBoxManage setextradata "VM name" "VBoxInternal/Devices/ahci/0/LUN#[x]/Config/IgnoreFlush" 0

[aeichner]

Please reopen if still relevant with a recent VirtualBox release, there were a lot of I/O related fixes since.