Conversion from Ext3/4 and ReiserFS

From btrfs Wiki
Revision as of 04:54, 16 February 2020 by Muxator (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Conversion from other filesystems (ext2, ext3, ext4, reiserfs) to btrfs

Btrfs has very few pieces of metadata that live in fixed locations, making it relatively easy to implement an in place conversion utility from other filesystems. Copy on write algorithms allow Btrfs to preserve an unmodified copy of the original FS, and allow administrator to undo the conversion, even after making changes in the resulting Btrfs filesystem.

The conversion program btrfs-convert uses libe2fs to read the ext3/4 metadata, and uses the free blocks in the ext3 filesystem to hold the new Btrfs filesystem. The basic conversion algorithm works like this:

  • Duplicate the first 1MB of the device
  • Duplicate directories and inodes, creating copies in Btrfs
  • Take references on ext3 file data blocks from the Btrfs files

This does create copies of all the ext3 metadata, but the Btrfs files simply point to the same blocks used by the ext3 files. This shares the bulk of the blocks in use between the two filesystems. Because Btrfs uses copy on write for all file modifications, the original ext3 version of the file data blocks is preserved.

The first 1MB of the device is copied to an alternate location so that Btrfs metadata may be written there. Undoing the conversion simply involves restoring these blocks.

The result of the conversion looks something like this:

The conversion program creates a snapshot in Btrfs that references all of the blocks used by ext3, allowing the administrator to control how long the original ext3 filesystem is preserved. If the admin wishes to recover the space used by ext3, the snapshot can be deleted and the conversion will be permanent.

Until the snapshot is deleted, only blocks that were marked as free in ext3 are used to hold new Btrfs modifications. This means the conversion can be undone at any time, restoring the original ext3 FS. The converter creates a sparse file to hold all of the references to the original ext3 filesystem, and the offsets in this file correspond to offsets on the block device. This allows the admin to mount the image file via readonly loopback and see the original ext3 filesystem and the converted Btrfs filesystem at the same time.

The manual page contains further information how to use the conversion utility.

Example usage

# Always run fsck first
fsck.ext3 -f /dev/xxx

# Convert from Ext3/4->Btrfs
btrfs-convert /dev/xxx

# Mount the resulting Btrfs filesystem
mount -t btrfs /dev/xxx /btrfs

# Mount the ext3/4 snapshot
mount -t btrfs -o subvol=ext2_saved /dev/xxx /ext2_saved

# Loopback mount the image file
mount -t ext3 -o loop,ro /ext2_saved/image /ext3

In the example above, the files in /ext3 and in /btrfs will be identical. Any modifications made in /btrfs will not show up in /ext3. If you wish to roll back the conversion:

# Completely unmount the Btrfs filesystem
umount /ext3
umount /ext2_saved
umount /btrfs

# roll back the conversion
btrfs-convert -r /dev/xxx

# use the original filesystem
mount -t ext3 /dev/xxx /ext3

Or, if you wish to recover the space used by the original Ext3/4 FS and make the conversion permanent, simply delete /ext2_saved/image. Once the image is deleted, the disk will look something like this:

Before first use

Suppose you're satisfied now and do not intend to rollback. Delete the saved subvolume:

# on a mounted filesystem
btrfs subvol delete ext2_saved

Wait until all it's fully deleted, this happens on the background. You can check whether there's a DELETED subvolume in the listing of btrfs subvol list -d /mnt. Now any logical relation to the original filesystem is removed.

The new filesystem inherits the block placement and file data fragmentation. Due to this it is highly recommended to do full defragmentation and full rebalance before "production" use. It is not required for general operation but will impact performance. A notable caveat is that a balance can fail with "ENOSPC" if the defragment is skipped. This is usually due to large extents on ext being larger than the maximum size btrfs normally operates with (1 GB). A defrag of all large files will avoid this:

btrfs fi defrag -r /mnt
btrfs balance start /mnt

Alternative to defragment only files larger than 1 GB:

find /mnt -type f -size +1G -print -exec btrfs fi defrag {} \;
btrfs balance start /mnt

Please note, that the balance process can take very long, you may want to do it in smaller steps, eg. metadata only, then data. See Balance Filters.

Personal tools