Fork of https://github.com/hiliev/py-zfs-rescue with the following improvements:
- lz4 and gzip decompression
- fletcher4 cksum
- first level child datasets
- blkptr with embedded data
- improved block server protocol
- bigger than 2TB disk support
- support SystemAttributes, bonus type 0x2c
- variable sectorsize
- bugfixed
- fuse (llfuse) interface for recovery
- raid0
This make it possible to read datapools created via i.e.
zpool create datapool -f -o ashift=12 -O atime=off -O canmount=off -O compression=lz4 -O normalization=formD raidz /dev/loop0 /dev/loop1 /dev/loop2
zfs create datapool/datadir
Rescue is done via the following steps:
- dump partitions of pool via dd
- use
zdb -l -u <disk>and determine the uberblock to use and edit variable TXG in zfs_rescue.py - For partitins greater than 2TB split partition into 1TB chunks and suffix with index
<diskname>.0, <diskname>.1, .... use block_proxy json init syntax<vdev> : [ "<diskname>.0", "<diskname>.1",... ] - edit BLK_PROXY_ADDR, INITIALDISKS, BLK_INITIAL_DISK
- run zfs_rescue.py to determine the dataset to use and edit list variable DS_TO_ARCHIVE
- edit MOUNTPOINT for fuse filesystem directory via which to access dataset and
rsync -avPX <MOUNTPOINT>/. <MOUNTPOINT>.bck/.to recover - hopefully... :-)