Skip to content

Interactive, file-level Time Machine-like tool for ZFS/btrfs/nilfs2 (and even Time Machine and Restic backups!)

License

Notifications You must be signed in to change notification settings

kimono-koans/httm

Repository files navigation

asciicast

httm

The dream of a CLI Time Machine is still alive with httm.

httm prints the size, date and corresponding locations of available unique versions (deduplicated by modify time and size) of files residing on snapshots, but can also be used interactively to select and restore files, even snapshot mounts by file! httm might change the way you use snapshots (because ZFS/BTRFS/NILFS2 aren't designed to find unique file versions) or the Time Machine concept (because httm is very fast!).

httm boasts an array of seductive features, like:

  • Search for versions of multiple files on distinct datasets simultaneously
  • Search for and recursively list deleted files. Even browse files hidden behind deleted directories.
  • List file snapshots from all local pools (detect local snapshot versions as well as locally replicated snapshot versions)!
  • List file snapshots from remote backup pools (even overlay replicated remote snapshot directories over live directories).
  • For use with rsync-ed non-ZFS/BTRFS/NILFS2/APFS local datasets (like ext4, XFS, or NTFS), not just ZFS/BTRFS/NILFS2 snapshots and TM backups.
  • Optionally preview snapshot file versions with a custom command (default is a diff compare to the live version)
  • List or even snapshot the mounts for a file directly
  • Roll forward to a previous snapshots, instead of rolling back (avoids destroying interstitial snapshots)
  • Guard any restore actions with precautionary snapshots
  • List snapshot names, even prune snapshots, which include a file name
  • Shortcut features: only display last snapshot, omit duplicates of the live file, etc.
  • Verification level: Like rsync, httm can determine whether file is unique based solely on metadata, or use checksums
  • 4 native interactive modes: browse, select, prune and restore
  • ANSI ls colors from your environment
  • Detect and display categories of unique file versions available (multiple, single, single-with-snap,..., etc.)
  • Select from several formatting styles (newline, null, tab delimited, JSON, CSV, etc.). Parseable ... or not ... oh my!
  • Packaged scripts which help you, and show you how to, use httm: ounce, bowie, equine, and nicotine
  • Supports ZFS/BTRFS/NILFS2 snapshots, and Restic and Time Machine backups!

Use in combination with you favorite shell's hot keys for even more fun.

Inspired by the findoid script, fzf, skim and many zsh key bindings.

Install via Native Packages

For Debian-based distributions (like Ubuntu), I maintain an (unsupported) personal package archive, or PPA. See the linked repository.

For Debian-based and Redhat-based Linux distributions (like Ubuntu or Fedora, etc.), also check the tagged releases for native packages for your distribution. For Redhat-based Linux distributions, you may need to use the --replacefiles option when installing via rpm -i, see the linked issue.

You may also create and install your own native package from the latest sources, for Debian-based and Redhat-based Linux distributions, like so:

#!/bin/bash
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
cargo install cargo-deb 
git clone https://github.com/kimono-koans/httm.git
cd ./httm/; cargo deb
# to install on a Debian/Ubuntu-based system
dpkg -i ./target/debian/httm_*.deb
# or convert to RPM 
alien -r ./target/debian/httm_*.deb
# and install on a Redhat-based system
rpm -i --replacefiles ./httm_*.rpm

For Arch-based Linux distributions, you can create and install your own native package from the latest sources, like so:

#!/bin/bash
# warning: this is only an example PKGBUILD
# you may need to edit the example, as needed, to conform to the latest release
wget https://raw.githubusercontent.com/kimono-koans/httm/master/packaging/arch/example/PKGBUILD
makepkg -si

For MacOS users, a user from the community (very exciting) has posted a formulae for Homebrew.

Install via Source

Installation via a package is suggested for support purposes, but, those wild childs, who perhaps can't stop (or won't stop), they can install by piping a script to bash:

bash -c "$(curl -fsSL https://raw.githubusercontent.com/kimono-koans/httm/master/packaging/install-from-source.bash)"

For those who simply wish to build httm incrementally themselves, the project contains only a few components:

  1. The httm executable. To build and install:

    #!/bin/bash
    curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
    latest="(wget -nv -O - "https://api.github.com/repos/kimono-koans/httm/releases/latest" 2>/dev/null | grep tag_name | cut -d: -f2 | cut -d'"' -f2)"
    # cargo-install places a release binary in your $HOME/.cargo/bin, however, 
    # you may wish to: cp $HOME/.cargo/bin/httm /usr/local/bin/httm; cargo uninstall httm
    cargo install --locked --git https://github.com/kimono-koans/httm.git --tag "$latest"
  2. The optional zsh hot-key bindings: Use ESC+s to select snapshots filenames to be dropped to your command line (for instance after the cat command), or use ESC+m to browse for all of a file's snapshots. After you install the httm binary, to copy the hot key script to your home directory, and source that script within your .zshrc:

    ➜ httm --install-zsh-hot-keys
  3. The optional man page: cargo has no native facilities for man page installation (though it may in the future!). You can use manpath to see the various directories your system uses and decide which directory works best for you. To install, just copy it to a directory in your man path, like so:

    ➜ cp ./httm/httm.1 /usr/local/share/man/man1/
  4. The optional scripts. See script usage below, in the Example Usage section, or follow the links (ounce, bowie, equine, and nicotine), for more information. To install, just copy it to a directory in your path, like so:

    #!/bin/bash
    cp ./httm/scripts/ounce.bash /usr/local/bin/ounce
    # bowie is "required" for the default/best "--preview" behavior
    cp ./httm/scripts/bowie.bash /usr/local/bin/bowie
    # equine is "required" for Time Machine support on MacOS
    cp ./httm/scripts/equine.bash /usr/local/bin/equine
    cp ./httm/scripts/nicotine.bash /usr/local/bin/nicotine
    chmod +x /usr/local/bin/bowie /usr/local/bin/ounce /usr/local/bin/equine /usr/local/bin/nicotine

Caveats

Right now, you will probably need to use a Unix-ish-y Rust-supported platform to build and install (that is: only Linux, FreeBSD, and MacOS are known to work). Note, your platform does not need to support ZFS/BTRFS/NILFS2 snapshots or TM backups to use httm. And there is no fundamental reason a non-interactive Windows version of httm could not be built, as it once did build, but Windows platform support is not a priority for me right now. Contributions from users are, of course, very welcome.

On FreeBSD, after a fresh minimal install, the interactive modes may not render properly, see the linked issue for the fix.

On some Linux distributions, which include old versions of libc, cargo may require building with musl instead, see the linked issue.

Example Usage

Note: Users may need to use sudo (or equivalent) to view versions on BTRFS or NILFS2 datasets, or Restic repositories, as BTRFS or NILFS2 snapshots or Restic repositories may require root permissions in order to be visible. Restic and Time Machine backups also require an additional flag, see further discussion of Restic --alt-store in the below.

Like other UNIX utilities (such as cat, uniq, sort), if you include no path/s as arguments, then httm will pause waiting for input on stdin:

# Press CTRL+C to send a SIGINT and quit the program
➜ httm 
# Pipe output of find command to httm
➜ find . -maxdepth 1 | httm

Print all unique versions of your history file:

➜ httm ~/.histfile

Print all unique versions of your history file, as formatted JSON:

➜ httm --json ~/.histfile

Browse all files in your home directory, recursively, and view unique versions on local snapshots:

➜ httm -b -R ~

Print all files on snapshots deleted from your home directory, recursive:

➜ httm -d -R ~

Browse all files deleted from your home directory, recursively, and view unique versions on all local and alternative replicated dataset snapshots:

➜ httm -d=only -b -a -R ~

Print all files on snapshots deleted from your home directory, recursive, newline delimited, piped to a text file:

# pseudo live file versions
➜ httm -d -n -R --no-snap ~ > pseudo-live-versions.txt
# unique snapshot versions
➜ httm -d -n -R --no-live ~ > deleted-unique-versions.txt

Browse all files in your home directory, recursively, and view unique versions on local snapshots, to select and ultimately restore to its original location, in overwrite mode:

➜ httm -r=overwrite -R ~

Browse all files in your home directory, recursively, and view unique versions on local snapshots, to select and ultimately restore to your working directory, in "guard" mode, which overwrites and guards any restore action with a pre-recovery ZFS snapshot:

printf "
# export a default httm restore mode
export HTTM_RESTORE_MODE=\"guard\"" >> ~/.zshenv
➜ httm -r -R ~

View unique versions of a file for recovery (shortcut, no need to browse a directory):

➜ httm -r /var/log/samba/log.smbd

httm is also a good Unix citizen, which means -- you should use the other Unix utilities to organize your queries how you like them. find and awk are especially useful here:

# search for the text "pattern" among snapshots of httm manpage
➜ httm -n --omit-ditto /usr/share/man/man1/httm.1.gz | xargs rg "pattern" -z

# similarly, print the directory sizes of each unique snapshot
➜ httm -n --omit-ditto /srv/downloads | xargs -I{} du -sh "{}"

# print all unique versions of your `/var/log/syslog` file, 
# newline delimited piped to `find` to print only versions 
# with modify times of less than 1 day from right now.
➜ httm -n --omit-ditto /var/log/syslog | xargs -I{} find '{}' -mtime -1

# httm usually sorts snapshot versions in chronological order, 
# oldest to newest, but since these are just paths/strings 
# you may choose to sort them differently.
#
# here, print all unique versions of your `/var/log/syslog` file, 
# then print each snapshot version's size in bytes first, 
# then reverse sort by its size, then remove the number of bytes, 
# leaving only the paths in their new sorted order
➜ httm -n --omit-ditto /var/log/syslog | xargs -I{} find '{}' -printf '%s\t%p\n' | \
sort -rn | awk 'BEGIN {FS="\t"}; {print $2}'

View bowie-formatted diff of each unique snapshot of ~/.zshrc against the live file version (requires the bowie script be installed):

➜ httm --preview -s ~/.zshrc

View cat output of each unique snapshot of ~/.zshrc:

➜ httm --preview="cat {snap_file}" -s ~/.zshrc

Recover the last-in-time unique file version (shortcut, no need to browse a directory or select from among other unique versions):

➜ httm -l -r /var/log/samba/log.smbd

Snapshot the dataset upon which /etc/samba/smb.conf is located:

➜ sudo httm -S /etc/samba/smb.conf

Browse all files, recursively, in a folder backed up via rsync to a remote share, and view unique versions on remote snapshots directly (only available for BTRFS Snapper and ZFS datasets). (Note: Remember to make ZFS snapshots visible in your smb.conf with zfsacl:expose_snapdir=True).

# mount the share
➜ open smb://<your name>@<your remote share>.local/Home
# execute httm
➜ httm -b -R /Volumes/Home

Browse all files, recursively, in your MacOS home directory backed up via rsync to a ZFS or BTRFS Snapper remote share, shared via smbd, and view unique versions on remote snapshots. Note: The difference from above is, here, you're browsing files from a "live" directory:

# mount the share
➜ open smb://<your name>@<your remote share>.local/Home
# execute httm
➜ httm -b -R --map-aliases /Users/<your name>:/Volumes/Home ~

Print all unique versions of your .zshrc file. Note: The difference from above is, httm, now, even supports your Time Machine backups directly. After using equine to mount my personal Time Machine ZFS network share:

➜ sudo equine --mount-local
Discovering backup locations (this can take a few seconds)...
Mounting snapshots...
...
➜ sudo equine --mount-remote
Connecting to remote Time Machine: smb://timemachine@montrose._smb._tcp.local./TM%20Volume ...
Mounting sparse bundle (this may include an fsck): Backups of kiev ...
/dev/disk4
/dev/disk5           EF57347C-0000-11AA-AA11-0030654
/dev/disk5s1         41504653-0000-11AA-AA11-0030654 /Volumes/Backups of kiev
Discovering backup locations (this can take a few seconds)...
Mounting snapshots...
...
# restic users can do something similar by:
# 1. mounting their repositories: restic -r /path/to/repo mount /path/to/mountpoint
# 2. invoking httm with --alt-store: httm --alt-store=restic .zshrc
➜ httm --alt-store=timemachine .zshrc
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Tue May 09 22:57:09 2023  6.7 KiB  "/Volumes/.timemachine/842A693F-CB54-4C5A-9AB1-C73681D4DFCD/2023-11-08-212757.backup/2023-11-08-212757.backup/Data/Users/kimono/.zshrc"
Sun Nov 12 20:29:57 2023  6.7 KiB  "/Volumes/.timemachine/842A693F-CB54-4C5A-9AB1-C73681D4DFCD/2023-11-18-011056.backup/2023-11-18-011056.backup/Data/Users/kimono/.zshrc"
Sun Nov 26 02:14:56 2023  6.7 KiB  "/Volumes/.timemachine/842A693F-CB54-4C5A-9AB1-C73681D4DFCD/2023-12-13-054342.backup/2023-12-13-054342.backup/Data/Users/kimono/.zshrc"
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Sun Nov 26 02:14:56 2023  6.7 KiB  "/Users/kimono/.zshrc"
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

View the differences between each unique snapshot version of the httm man page and each previous version (this simple script is the basis for bowie):

#!/bin/bash
filename="./httm/httm.1"
# previous version is unset
previous_version=""
for current_version in $(httm -n --omit-ditto $filename); do
    # check if initial "previous_version" needs to be set
    if [[ -z "$previous_version" ]]; then
        previous_version="$current_version"
        continue
    fi

    # print that current version and previous version differ
    diff -q "$previous_version" "$current_version"
    # print the difference between that current version and previous_version
    diff "$previous_version" "$current_version"

    # set current_version to previous_version
    previous_version="$current_version"
done

Create a simple tar archive of all unique versions of your /var/log/syslog:

➜ httm -n --omit-ditto /var/log/syslog | tar -zcvf all-versions-syslog.tar.gz -T -

Create a kinda fancy tar archive of all unique versions of your /var/log/syslog:

#!/bin/bash
file="/var/log/syslog"
dir_name="${$(dirname $file)/\//}"
base_dir="$(basename $file)_all_versions"
# squash extra directories by "transforming" them to simply snapshot names 
httm -n --omit-ditto "$file" | \
tar \
--transform="flags=r;s|$dir_name|$base_dir|" \
--transform="flags=r;s|.zfs/snapshot/||" \
--show-transformed-names \
-zcvf "all-versions-$(basename $file).tar.gz" -T  -

Create a super fancy git archive of all unique versions of /var/log/syslog (this simple script is the basis for nicotine):

#!/bin/bash
# create variable for file name
file="/var/log/syslog"
# create git repo
mkdir ./archive-git; cd ./archive-git; git init
# copy each version to repo and commit after each copy
for version in $(httm -n --omit-ditto $file); do
    cp "$version" ./
    git add "./$(basename $version)"
    # modify commit date to match snapshot modify date-time
    git commit -m "httm commit from ZFS snapshot" \
    --date "$(date -d "$(stat -c %y $version)")"
done
# create git tar.gz archive
tar -zcvf "../all-versions-$(basename $file).tar.gz" "./"
# and to view
git log --stat

Use ounce (codename: "dimebag"), a wrapper script for httm, for no mental overhead, non-periodic dynamic snapshots:

# request ZFS snapshot privileges
➜ ounce --give-priv
# here you create a "dummyfile", ounce will add a snapshot of "dummyfile" 
# before you remove it, and httm will allow you to view the snapshot created
➜ touch ~/dummyfile; ounce rm ~/dummyfile; httm ~/dummyfile
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Wed Feb 15 12:59:42 2023  0 bytes  "/home/kimono/.zfs/snapshot/snap_2023-02-15-12:59:42_ounceSnapFileMount/dummyfile"
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
                                   "/home/kimono/dummyfile"
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
# use as an alias around programs which modify files/dirsprintf "
# ounce aliases 
alias vim=\"ounce --trace vim\"
alias emacs=\"ounce --trace emacs\"
alias nano=\"ounce --trace nano\"
alias rm=\"ounce rm\"" >> ~/.zsh_aliases

Use bowie, a wrapper script for httm, to display the difference between unique snapshot versions and the live file:

➜ bowie ~/.zshrc
/home/kimono/.zshrc
__
Files /home/kimono/.zfs/snapshot/snap_2023-02-14-13:42:11_ounceSnapFileMount/.zshrc and /home/kimono/.zshrc differ
1c1
< ### If you come from bash you might have to change your $PATH.
---
> # If you come from bash you might have to change your $PATH.

Use nicotine, a wrapper script for httm, to convert unique snapshot file versions to git archives:

➜ nicotine .zshrc
nicotine git archive created successfully: /home/kimono/zshrc-git.tar.gz

Roll forward to a previous ZFS snapshot, instead of rolling back (this avoids destroying interstitial snapshots):

➜ sudo httm --roll-forward=rpool/scratch@snap_2023-04-01-15:26:06_httmSnapFileMount
[sudo] password for kimono:
httm took a pre-execution snapshot named: rpool/scratch@snap_pre_2023-04-01-15:27:38_httmSnapRollForward
...
httm roll forward completed successfully.
httm took a post-execution snapshot named: rpool/scratch@snap_post_2023-04-01-15:28:40_:snap_2023-04-01-15:26:06_httmSnapFileMount:_httmSnapRollForward

Yo, @kimono-koans, where do your snapshots come from?

If you'd like to read more about how someone else/I personally create snapshots, you might try my A Somewhat Opinionated Guide to Effective ZFS Snapshots.

I know what you're thinking, but slow your roll

To be clear, httm is not...

License

httm is licensed under the MPL 2.0 License - see the LICENSE file for more details.