Linux – Identify files that would be deleted when removing an rsync –link-dest directory

backuphardlinklinuxrsync

I have a Time Machine-style backup system set up on my NAS (running BusyBox v1.16.1), using rsync --link-dest as described here: http://blog.interlinked.org/tutorials/rsync_time_machine.html

Is there an easy way to find out which files will be deleted if I delete an old backup directory, and which will just be unlinked? I assume one way would be to find which inodes in the old directory have only one link to them, but I don't know a way to do that. rm --dry-run anyone?

Bonus Points: Find a way to do the above, and also calculate the total disk space freed if the old backup is deleted.

Best Answer

A simple way would be to just use find with the -links option.

Specifically you would want to do something like find path -type f -links 1.

I haven't tested, but I believe a command like this would delete all the multi-linked files.

# you might not need to escape the !, depends on your shell
# should find all the files that do NOT have a link count of 1 and delete them
find path -type f \! links 1 -print -delete

Related Solutions

Using rsync with link-dest from HFS to NTFS

-a implies that rights that may not translate correctly to NTFS are to be copied and matched. I use only -rltD. -a implies -rlptgoD.

Note: I use rsync from linux with EXT4, and I don't know how HFS compares to NTFS.

Here is a complete script I use to backup some of my folders to removable USB-disc. This works fine in Ubuntu 10.4

#!/bin/bash
# Rotated backup from EXT4 to removable NTFS-disc using rsync.
# Four generations are saved and automatically purged each run.
# Generations: current, backups/old, backups/older, backups/oldest.
# This script is stored on and run from the root of the removable disc
# where the backups are stored. Destination paths in the rsync commands
# are relative to current working directory below.

# Purge oldest backup
rm -rf backups/oldest

# Prepare recieving folder.
mkdir inprogress

# Grab new contents. Use rsync to create hard links to files already backed up on media. 
# Note: --link-dest is set relative to dest.
# Note: Since we copy from EXT4 to NTFS we can't use -a. Rights are different in NTFS.
#       If we tried then rsync would copy every file, since rights don't match. 
#       I use -rltD instead of -a. Care must be taken when restoring files!

echo "Backup of Musik is updated with changes since last backup"
rsync -rltD --verbose --modify-window=1 --delete \
      --link-dest=../../current/Musik \
      /home/anders/Musik/ \
      inprogress/Musik

echo "Backup of tv is updated with changes since last backup"
rsync -rltD --verbose --modify-window=1 --delete \
      --link-dest=../../current/tv \
      /home/anders/Video/tv/ \
      inprogress/tv

echo "Backup of Calibre Library is updated with changes since last backup"
rsync -rltD --verbose --modify-window=1 --delete \
      --link-dest="../../current/Calibre Library" \
      "/home/anders/Calibre Library/" \
      "inprogress/Calibre Library"

# Rotate the backups
# mkdir backups (only needed first run)
mv backups/older backups/oldest
mv backups/old backups/older
mv current backups/old
mv inprogress current

echo Done!

Sample output from a run:

anders@anders-desktop:/media/Samsung S2$ ./refresh.sh 
Backup of Musik is updated with changes since last backup
sending incremental file list
./
Artists/
Various Artists/

sent 1787165 bytes  received 3256 bytes  102309.77 bytes/sec
total size is 230838013393  speedup is 128929.46
Backup of tv is updated with changes since last backup
sending incremental file list
./

sent 7558 bytes  received 35 bytes  5062.00 bytes/sec
total size is 64808873338  speedup is 8535344.84
Backup of Calibre Library is updated with changes since last backup
sending incremental file list
./

sent 227427 bytes  received 1883 bytes  91724.00 bytes/sec
total size is 825094709  speedup is 3598.16
Done!

Linux – hard link files with rsync instead of copying them

The answer has turned out to be yes, with rsync's --link-dest option. It's just not obvious because it's not a simple on/off flag like cp -l, because it takes a path argument.

So, cp -l a/ b/ can also be done like this:

rsync -a --link-dest=a a/ b/