• increment backup of home dir

    From tim wade@21:1/5 to All on Fri Mar 14 07:10:01 2025
    Hello

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment
    backup?

    Thank you.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Nicolas George@21:1/5 to All on Fri Mar 14 09:10:01 2025
    tim wade (HE12025-03-14):
    besides rsync, do you know any other software/service for increment backup?

    Ahem. rsync is not a backup tool.

    rsync can be *part of* a backup tool, for example in rsnapshot, or
    manually with filesystem snapshots, but rsync alone does not allow you
    to recover a file you damaged, which is part of the job description of a
    backup tool.

    Regards,

    --
    Nicolas George

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Tim Woodall@21:1/5 to Jeffrey Walton on Fri Mar 14 08:50:01 2025
    On Fri, 14 Mar 2025, Jeffrey Walton wrote:

    On Fri, Mar 14, 2025 at 2:08?AM tim wade <[email protected]> wrote:

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment
    backup?

    I use Duplicity to backup a webserver and MySQL database. The
    directories are different, but the operations are the same.

    Jeff



    If it's ext4 then dump and restore. Although ideally you want to use the version from trixie.

    I've been using it for 20+ years and am now the upstream maintainer.

    Tim.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Gareth Evans@21:1/5 to tim wade on Fri Mar 14 08:50:01 2025
    Hi Tim,

    Borg is perhaps worth a mention.

    https://borgbackup.readthedocs.io/en/stable/

    It's not truly "incremental", though combines compression, deduplication and optional encryption, which may improve on that.


    $ sudo borg info <user>@<user>.rsync.net:home::d14
    Archive name: d14
    ...
    Time (start): Fri, 2025-03-14 00:02:20
    Time (end): Fri, 2025-03-14 00:09:23
    Duration: 7 minutes 3.39 seconds
    Number of files: 476356
    ...
    Utilization of maximum supported archive size: 0% ------------------------------------------------------------------------------
    Original size Compressed size Deduplicated size This archive: 165.52 GB 152.74 GB 81.86 kB All archives: 9.37 TB 8.64 TB 127.69 GB

    That repo stores (in ~128GB) deduplicated archives corresponding to backups every
    - hour for 24 hours
    - day, for 31 days
    - month, for a year

    There is a high if somewhat fuzzy limit to archive size, which seems likely to remain theoretical in many cases, and if not there is a simple workaround:
    https://borgbackup.readthedocs.io/en/stable/faq.html#are-there-other-known-limitations

    You can run borg locally, but it needs to run in server mode on the backup target, so requires to be installed there or to have a binary at --remote-path if specified. This effectively limits remote provider options in comparison to eg. restic which
    might also be worth looking at.

    https://restic.net

    A now slightly dated but interesting comparison: https://www.reddit.com/r/BorgBackup/comments/v3bwfg/why_should_i_switch_from_restic_to_borg/

    I use borg on rsync.net for remote backup. The only remote providers I know of are:

    https://rsync.net
    https://www.borgbase.com

    rsync.net is a FreeBSD-based platform so also offers snapshots (on standard or custom schedule) of your filesystem, in which borg files reside. Encryption is not offered in the filesystem so must be done via borg if desired.

    The version of borg in the Bookworm apt repo is not the latest, but it works well in my experience.

    HTH
    Gareth

    On Fri 14/03/2025 at 06:07, tim wade <[email protected]> wrote:
    Hello

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment
    backup?

    Thank you.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Loris Bennett@21:1/5 to tim wade on Fri Mar 14 09:20:01 2025
    Hi Tim,

    tim wade <[email protected]> writes:

    Hello

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment
    backup?

    Thank you.

    I have a Raspberry Pi running Nextcloud and so have my important data
    both on the Raspberry Pi and on two computers which sync with it. The Nextcloud instance also does an rsync to a drive connected to my router
    and also creates BTRFS snapshots on a further external drive.

    Just for backup up it is probably overkill and was a bit of a faff to
    set up. However, it works well for me and the backup aspect was
    actually my main motivation for doing in.

    Cheers,

    Loris

    --
    This signature is currently under constuction.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Henrik Ahlgren@21:1/5 to Gareth Evans on Fri Mar 14 09:20:01 2025
    "Gareth Evans" <[email protected]> writes:

    It's not truly "incremental", though combines compression,
    deduplication and optional encryption, which may improve on that.

    Not "incremental" in the traditional tape-backup sense, which requires
    periodic full backups. A closer term might be "incremental forever"
    (like IBM TSM), meaning you never need to resend the same data. For use
    cases such as typical workstations (where data churn is minimal), hourly backups are practical and cost-effective and is much better RPO
    (recovery point objetive) than daily backups.

    Borg has a few pain points (e.g., locking, which prevents simultaneous
    data restoration and new backups), but overall, it is awesome.

    I use borg on rsync.net for remote backup. The only remote providers I know of are:

    https://rsync.net
    https://www.borgbase.com

    Hetzner supports Borg and offers reasonably priced plans.

    https://docs.hetzner.com/storage/storage-box/access/access-ssh-rsync-borg/

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Chris Green@21:1/5 to Nicolas George on Fri Mar 14 10:50:02 2025
    Nicolas George <[email protected]> wrote:
    tim wade (HE12025-03-14):
    besides rsync, do you know any other software/service for increment backup?

    Ahem. rsync is not a backup tool.

    rsync can be *part of* a backup tool, for example in rsnapshot, or
    manually with filesystem snapshots, but rsync alone does not allow you
    to recover a file you damaged, which is part of the job description of a backup tool.

    Ah, yes, that's the one I was trying to remember when I suggested
    rsyncbackup. rsnapshot is the one I used to use before I wrote my
    own.

    --
    Chris Green
    ·

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Chris Green@21:1/5 to tim wade on Fri Mar 14 10:50:02 2025
    tim wade <[email protected]> wrote:
    Hello

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment
    backup?

    Most of the replies so far seem to have ignored 'increment[al]'. While
    complete backups are sometimes useful I find incremental ones much
    better most of the time and they allow one to keep really old backups
    without running out of space.

    There are a couple of programs that use rsync to do incremental
    backups, the rsync resources page (https://rsync.samba.org/resources.html) provides some possibilities.

    I used to use rsyncbackup which worked OK but I decided I wanted more
    ways to configure things so I wrote my own system which is simply a
    bash script that uses rsync to do the actual file copying and making
    hard links.

    I have hourly backups of /home which are kept until they're five weeks
    old, these often save me from myself when I delete or change files by
    mistake or whatever. These are kept on the same machine but on a
    separate drive.

    Then I have long term backups which are made daily to a remote
    machine, every day's backup is kept for a month and then they are
    selectively thinned out so I have monthly backups for the past year
    and then yearly back ups 'for ever'. They go back around twwnty years
    or more!

    I did actually dig around one of the very old backups (well over ten
    years old) for some photos my daughter had lost - and I found them!


    One of the big advantages of using rsync's hard-link for doing this is
    that the backups all look like a complete /home directory and you can
    dig around in them without any effort at all, no unpacking, no
    decoding, nothing, it's all just the files as you know them.


    I do use .rsync-filter to filter out things that either don't need
    backup or that change so frequently that they make backups pointless,
    things like .bash_history and .xsession-errors for example.


    --
    Chris Green
    ·

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Jerome BENOIT@21:1/5 to tim wade on Fri Mar 14 11:30:02 2025
    Hi,

    backup2l is simple and has been reliable for me for years.

    Cheers,
    Jerome

    On 14/03/2025 07:07, tim wade wrote:
    Hello

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment backup?

    Thank you.


    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From [email protected]@21:1/5 to Harri Suutari on Fri Mar 14 13:40:01 2025
    Harri Suutari <[email protected]> wrote:
    On 14/03/2025 08.07, tim wade wrote:

    besides rsync, do you know any other software/service for increment backup?

    Dirvish - "Disk based virtual image network backup system."

        Dirvish can create user browseable (daily) backup directories.

    I use dirvish and have for many years. I create daily incremental
    backups. It uses rsync under the hood. There's an article at https://news.ycombinator.com/item?id=16085001 that summarises. I think
    the mailing list is still going, just about. I know perl so it suits me.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Hans@21:1/5 to All on Fri Mar 14 13:50:01 2025
    Depends on, what you prefer: in console, or with GUIn as conjob or whatever

    There is "borgbackup2" in my mind, as well as "backintime-qt" (for wide ways
    of settings), and for the easiest way, I suggest "deja-dup".

    Hans

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From gene heskett@21:1/5 to Chris Green on Fri Mar 14 14:00:02 2025
    On 3/14/25 05:40, Chris Green wrote:
    tim wade <[email protected]> wrote:
    Hello

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment
    backup?

    Most of the replies so far seem to have ignored 'increment[al]'. While complete backups are sometimes useful I find incremental ones much
    better most of the time and they allow one to keep really old backups
    without running out of space.

    There are a couple of programs that use rsync to do incremental
    backups, the rsync resources page (https://rsync.samba.org/resources.html) provides some possibilities.

    I used to use rsyncbackup which worked OK but I decided I wanted more
    ways to configure things so I wrote my own system which is simply a
    bash script that uses rsync to do the actual file copying and making
    hard links.

    I have hourly backups of /home which are kept until they're five weeks
    old, these often save me from myself when I delete or change files by
    mistake or whatever. These are kept on the same machine but on a
    separate drive.

    Then I have long term backups which are made daily to a remote
    machine, every day's backup is kept for a month and then they are
    selectively thinned out so I have monthly backups for the past year
    and then yearly back ups 'for ever'. They go back around twwnty years
    or more!

    I did actually dig around one of the very old backups (well over ten
    years old) for some photos my daughter had lost - and I found them!


    One of the big advantages of using rsync's hard-link for doing this is
    that the backups all look like a complete /home directory and you can
    dig around in them without any effort at all, no unpacking, no
    decoding, nothing, it's all just the files as you know them.


    I do use .rsync-filter to filter out things that either don't need
    backup or that change so frequently that they make backups pointless,
    things like .bash_history and .xsession-errors for example.

    That sounds most interesting Chris. Something that can work as you work
    w/o interfering with your work.

    Can you write up some docs, pkg and publish it?

    Cheers, Gene Heskett, CET.

    --
    "There are four boxes to be used in defense of liberty:
    soap, ballot, jury, and ammo. Please use in that order."
    -Ed Howdershelt (Author, 1940)
    If we desire respect for the law, we must first make the law respectable.
    - Louis D. Brandeis

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Loris Bennett@21:1/5 to Chris Green on Fri Mar 14 14:00:02 2025
    Chris Green <[email protected]> writes:

    tim wade <[email protected]> wrote:
    Hello

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment
    backup?

    Most of the replies so far seem to have ignored 'increment[al]'. While complete backups are sometimes useful I find incremental ones much
    better most of the time and they allow one to keep really old backups
    without running out of space.

    Indeed, I should have explicitly mentioned that the Btrfs snapshots are incremental.

    [snip (33 lines)]

    --
    This signature is currently under constuction.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Nicolas George@21:1/5 to All on Fri Mar 14 14:10:01 2025
    Hans (HE12025-03-14):
    Depends on, what you prefer: in console, or with GUIn as conjob or whatever

    A tool that cannot be automated is not a backup tool.

    Corollary: a GUI-only tool is not a backup tool.

    Regards,

    --
    Nicolas George

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Chris Green@21:1/5 to gene heskett on Fri Mar 14 15:30:02 2025
    gene heskett <[email protected]> wrote:

    That sounds most interesting Chris. Something that can work as you work
    w/o interfering with your work.

    Exactly! For example one can sometimes get in a real mess editing a
    file and you want to get back to where you were an hour, or a few
    hours ago. Look in the hourly incremental backups and copy the file
    back from there. (I do use mercurial configuration management as well
    but that only works if I have done a commit of the version I want back).


    Can you write up some docs, pkg and publish it?

    Er, um, possibly. :-) Where would one publish something like this? I
    don't really want to go to the extent of putting it on github or
    somewhere like that but I'd be quite happy to put the source code as
    it stands somewhere so that people can copy, modify, whatever as they
    want.

    It's fairly well documented by comments.

    As I said there are two similar but separately coded versions. The
    hourly backup one is a bash script and two python scripts (prebak.py
    and postbak.py). The daily backup one uses an rsync server on the
    remote machine where the backups are stored but is otherwise pretty
    similar to the hourly backup one.

    --
    Chris Green
    ·

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Charles Curley@21:1/5 to tim wade on Fri Mar 14 15:40:01 2025
    On Fri, 14 Mar 2025 14:07:53 +0800
    tim wade <[email protected]> wrote:

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment
    backup?

    I use and recommend amanda and rsnapshot, both already mentioned. If you
    like Apple's Time Machine, take a look at backintime.

    --
    Does anybody read signatures any more?

    https://charlescurley.com
    https://charlescurley.com/blog/

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Hans@21:1/5 to All on Fri Mar 14 15:40:02 2025
    A tool that cannot be automated is not a backup tool.

    Corollary: a GUI-only tool is not a backup tool.

    Regards,

    You are making it too easy for you. A tool, than can be configured with a GUI can of course run automated. Ask Microsoft or Apple!

    But there are also tools, which are only running cli and can (or even must be) configured with the help of cron - rsync, for example.

    But this discussion belongs not here and does not help the op.

    Hans

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Stefan Monnier@21:1/5 to All on Fri Mar 14 22:20:01 2025
    besides rsync, do you know any other software/service for increment backup?
    I use borg. It stores files in its own archive format with
    deduplication and compression. 4 backups of 32G /+/home of my old
    netbook created every month stored in ~11GB backup directory.
    Slower than rsync, eat more cpu, need borg binary at remote storage
    (or mount storage local). Chosen backup archive (see borg list
    $backup-path) may be r/o mounted for restoring some files if needed.

    I use Bup, which provides a fairly similar featureset to Borg (tho
    doesn't support encryption yet). AFAIK the main difference is that
    instead of its own archive format, Bup uses the Git repository format.

    I haven't actually tried Borg, but I'm surprised by "slower than rsync,
    eat more cpu" because I'd expect it to perform similarly to Bup, and
    (for my use case at least) Bup performs incremental backups
    significantly faster and using less CPU (especially on the server to
    which I send my backups) than the rsync system I used before.

    OTOH, Bup is much slower than Rsync when it comes to browsing the
    resulting backup (done via a read-only `fuse` mount).


    Stefan

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From David Christensen@21:1/5 to tim wade on Fri Mar 14 22:20:02 2025
    On 3/13/25 23:07, tim wade wrote:
    Hello

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment backup?

    Thank you.


    "Incremental backup" implies "full backup" -- e.g. make full backups on
    Sunday nights and make incremental backups Monday through Saturday
    nights, etc..


    Conventionally, you use the same tool for both full and incremental
    backups, and specify the desired operating mode when running it.


    What tool are you using for backups?


    Please consider describing how you use currently the backup tool and/or
    how you would like to use the backup tool -- e.g. manual or scheduled/ automated operation, what computer and user account runs the tool, what computers, directories, and files are included, what computers,
    directories, and files are excluded, where the backups are stored,
    strategies for rotating backup media, if, how, and where the backups are duplicated, the scale and procedure for restoring computers,
    directories, and files, etc.. Such details would facilitate better replies.


    David

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Nicolas George@21:1/5 to All on Fri Mar 14 22:40:02 2025
    David Christensen (HE12025-03-14):
    "Incremental backup" implies "full backup" -- e.g. make full backups on Sunday nights and make incremental backups Monday through Saturday nights, etc..

    Or a full backup in 2025 and the next one in 3025, which is functionally equivalent to no full backup after the very first one. And the very
    first one can be considered an incremental backup over a full one of no
    data at all.

    Regards,

    --
    Nicolas George

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Jonathan Dowland@21:1/5 to Stefan Monnier on Fri Mar 14 23:10:01 2025
    On Fri Mar 14, 2025 at 9:14 PM GMT, Stefan Monnier wrote:
    I use Bup, which provides a fairly similar featureset to Borg (tho
    doesn't support encryption yet). AFAIK the main difference is that
    instead of its own archive format, Bup uses the Git repository format.

    Can bup purge old backups?

    --
    Please do not CC me for listmail.

    👱🏻 Jonathan Dowland
    [email protected]
    🔗 https://jmtd.net

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Russell L. Harris@21:1/5 to tim wade on Fri Mar 14 23:20:02 2025
    On Fri, Mar 14, 2025 at 02:07:53PM +0800, tim wade wrote:
    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    I use git. I keep terminal open running a ssh connection open to the
    backup system. Whenever I wish to save the state of the system, I
    switch to the terminal and execute git commit. To check the previous
    state of a file, Emacs provides git-timemachine.

    Simple and miminal hassle.

    RLH

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Stefan Monnier@21:1/5 to All on Fri Mar 14 23:20:02 2025
    Jonathan Dowland [2025-03-14 22:03:21] wrote:
    On Fri Mar 14, 2025 at 9:14 PM GMT, Stefan Monnier wrote:
    I use Bup, which provides a fairly similar featureset to Borg (tho
    doesn't support encryption yet). AFAIK the main difference is that
    instead of its own archive format, Bup uses the Git repository format.
    Can bup purge old backups?

    Yes, it has `bup prune-older`.
    I must admit I don't use it, but I'd expect it to be fairly
    costly/slow.


    Stefan

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Henrik Ahlgren@21:1/5 to Stefan Monnier on Fri Mar 14 23:20:02 2025
    Stefan Monnier <[email protected]> writes:

    I haven't actually tried Borg, but I'm surprised by "slower than rsync,
    eat more cpu" because I'd expect it to perform similarly to Bup, and
    (for my use case at least) Bup performs incremental backups
    significantly faster and using less CPU (especially on the server to
    which I send my backups) than the rsync system I used before.

    Borg is implemented in Python, which may not be the fastest language.
    Borg is fast enough for most common use cases, though it may struggle to
    scale with very large datasets. Backup is a surprisingly complex
    problem, and every software solution involves trade-offs.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Stefan Monnier@21:1/5 to All on Fri Mar 14 23:30:01 2025
    Yes, it has `bup prune-older`.
    I must admit I don't use it, but I'd expect it to be fairly
    costly/slow.

    The thing I do use and like very much is `bup get` which transfers
    backups between two repositories. I use it for off-site backups, and
    compared to my previous Rsync based backups (where I used a plain
    `rsync` to update the off-site backup), that's blazingly fast and cheap
    (my Bup/Rsync repositories live in resource constrained machines
    (BananaPi)). I used to update the off-site backup once a week at
    a carefully chosen time to minimize disruption, whereas now I do it
    daily at random times and don't notice it happening.


    Stefan

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Jerome BENOIT@21:1/5 to Greg on Fri Mar 14 23:40:02 2025
    Hello,

    On 14/03/2025 14:39, Greg wrote:
    On 2025-03-14, Jerome BENOIT <[email protected]> wrote:
    Hi,

    backup2l is simple and has been reliable for me for years.

    Is that a Debian package?

    Yep, here is its tracker page:

    https://tracker.debian.org/pkg/backup2l

    Cheers,
    Jerome



    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Russell L. Harris@21:1/5 to Default User on Sat Mar 15 04:20:01 2025
    On Fri, Mar 14, 2025 at 07:53:09PM -0400, Default User wrote:
    On 2025-03-14 [FR], Russell Harris wrote:

    I use git. I keep terminal open running a ssh connection open to the
    backup system. Whenever I wish to save the state of the system, I
    switch to the terminal and execute git commit. To check the previous
    state of a file, Emacs provides git-timemachine.

    Simple and miminal hassle.

    Would you mind sharing with us some details about how you do this (and
    how you decided to do so), or at least give some pointers to more
    information about this from other sources?

    = I write for a living, but with age and arthritis, a simple
    typographic error can be disastrous, particularly with Emacs. Also, I
    may rewrite a paragraph or several pages, and then decide that the
    former version was better. Accordingly, I vary the backup frequency
    partly upon finger dexterity (which can vary from day to day) and
    partly upon the complexity of the material on which I am working.
    When proofreading, I may backup with each change, no matter how minor.

    = I back up only the repository of the articles I write, together with supporting documents such as LaTeX document class files and scripts.
    Everything I wish to backup, I keep in a single directory, with the
    associated ".git" file. Of course, you can have multiple git
    repositories on a single drive.

    = With git, I can do a hundred backups in a morning or afternoon
    session without worry about consumption of disk space. I am not
    concerned with an orderly and uncluttered commit record; my concern is
    the ability to recover.

    = The time required for each backup is negligible:

    (1) switch to a virtual terminal
    (2) type: git commit -a -m journal #050
    (3) switch back to Emacs

    = I have in the LAN a machine which I keep running almost constantly.
    The machine runs getmail to pull messages from the mail servers of one
    or more ISPs. It seemed logical to me to utilize this machine as a
    git repository, and also as an approx server for the LAN. I suppose I
    ought also to have it generate web pages from the data stream of my
    backyard weather station and upload them to a website.

    = At the beginning of a session, I use the bash control-r command to
    pull up commonly-used git commands:

    git status
    git add -A
    git commit -a -m document #13

    = For recovery, in Emacs I type:

    alt-x git-timemachine

    which allows me to use "p" and "n" (previous and next) to scroll
    through saved states.

    = If I need to work outside of the office, I pull out the laptop and
    clone the repository:

    git clone 192.168.1.40:/git.repository.git /home/rlh/git.clone-2025.03.15

    RLH

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Henrik Ahlgren@21:1/5 to Russell L. Harris on Sat Mar 15 10:30:01 2025
    "Russell L. Harris" <[email protected]> writes:

    = With git, I can do a hundred backups in a morning or afternoon
    session without worry about consumption of disk space. I am not
    concerned with an orderly and uncluttered commit record; my concern is
    the ability to recover.

    Git is an awesome tool that can be utilized with a wide variety of
    files, not just software source code or text. However, it faces
    scalability challenges with larger files, such as videos. While
    extensions like git-annex can assist in managing these larger files,
    they can struggle when dealing with tens of thousands or more of files.

    Additionally, Git's core feature of maintaining immutable history is not well-suited for backing up large, dynamic datasets, where retaining the
    entire history indefinitely may be undesirable. There may also be
    privacy and legal concerns related to maintaining permanent records of
    all changes, since establishing a retention policy is difficult.

    So, most people cant't have their whole $HOME in a Git repository.

    Moreover, storing Git repositories on the same storage device as your
    original files only safeguards your data against your own mistakes. This
    is analogous to the (opposite) misconception that RAID systems serve as
    a backup solution; they only protect against hardware failures of one or
    more disks, and happily mirror all your mistakes for you. Proper backup
    system need to protect your data from most (ideally all) data loss
    scenarios.

    Fortunately, with Git, setting up remote copies on another disk or
    machine is straightforward, so it is an excellent backup tool for data
    sets that are well suited to version control, such as your writing
    projects.

    = The time required for each backup is negligible:

    (1) switch to a virtual terminal
    (2) type: git commit -a -m journal #050
    (3) switch back to Emacs

    I recommend checking out the Magit package for Emacs. Once people try
    it, few return to using the command line for git operations.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Tim Woodall@21:1/5 to Russell L. Harris on Sat Mar 15 21:00:02 2025
    On Fri, 14 Mar 2025, Russell L. Harris wrote:

    On Fri, Mar 14, 2025 at 02:07:53PM +0800, tim wade wrote:
    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    I use git. I keep terminal open running a ssh connection open to the
    backup system. Whenever I wish to save the state of the system, I
    switch to the terminal and execute git commit. To check the previous
    state of a file, Emacs provides git-timemachine.

    Simple and miminal hassle.

    git fails to preserve ownership, permissions or timestamps. While this
    may not be relevant to your usecase, for example backing up /etc would
    be catastrophic (which is why we have etckeeper)

    For a homedir, ownership may not be a problem but permissions might be.
    IIRC, for example, ssh (and sshd) will not work properly if the
    permissions on ~/.ssh are wrong.

    git will also not backup (or even see) extended attributes.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From David Christensen@21:1/5 to Tim Woodall on Sat Mar 15 21:20:03 2025
    On 3/15/25 12:50, Tim Woodall wrote:
    git fails to preserve ownership, permissions or timestamps. While this
    may not be relevant to your usecase, for example backing up /etc would
    be catastrophic (which is why we have etckeeper)


    I am curious -- if I make my /etc directory tree into a version control
    system working directory (Git or otherwise), please explain how this
    would be catastrophic.


    David

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Andy Smith@21:1/5 to David Christensen on Sat Mar 15 21:40:01 2025
    Hi,

    On Sat, Mar 15, 2025 at 01:18:45PM -0700, David Christensen wrote:
    On 3/15/25 12:50, Tim Woodall wrote:
    git fails to preserve ownership, permissions or timestamps. While this
    may not be relevant to your usecase, for example backing up /etc would
    be catastrophic (which is why we have etckeeper)

    I am curious -- if I make my /etc directory tree into a version control system working directory (Git or otherwise), please explain how this would
    be catastrophic.

    /etc has things in it which require specific ownership and permissions.
    These aren't stored in git or in many other version control systems. It
    will be fine until you actually restore anything, at which point the
    files will get ownership/group of your user and mode 0644 or 0755
    depending upon whether they were executable or not when committed.

    Thanks,
    Andy

    --
    https://bitfolk.com/ -- No-nonsense VPS hosting

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From David Christensen@21:1/5 to Andy Smith on Sun Mar 16 02:20:01 2025
    On 3/15/25 13:36, Andy Smith wrote:
    On Sat, Mar 15, 2025 at 01:18:45PM -0700, David Christensen wrote:
    I am curious -- if I make my /etc directory tree into a version control
    system working directory (Git or otherwise), please explain how this would >> be catastrophic.

    /etc has things in it which require specific ownership and permissions.
    These aren't stored in git or in many other version control systems. It
    will be fine until you actually restore anything, at which point the
    files will get ownership/group of your user and mode 0644 or 0755
    depending upon whether they were executable or not when committed.

    Thanks,
    Andy


    Thank you for the clarification.


    David

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Ryan Nowakowski@21:1/5 to tim wade on Sun Mar 16 05:30:01 2025
    ------05ARS48RL5XN4T0DYBBF8YWLCOJ9SJ
    Content-Type: text/plain;
    charset=utf-8
    Content-Transfer-Encoding: quoted-printable

    I use rdiff-backup. It does incremental backups.

    On March 14, 2025 1:07:53 AM CDT, tim wade <[email protected]> wrote:
    Hello

    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    besides rsync, do you know any other software/service for increment backup?

    Thank you.


    ------05ARS48RL5XN4T0DYBBF8YWLCOJ9SJ
    Content-Type: text/html;
    charset=utf-8
    Content-Transfer-Encoding: quoted-printable

    <html><head></head><body><div dir="auto">I use rdiff-backup. It does incremental backups.</div><br><br><div class="gmail_quote"><div dir="auto">On March 14, 2025 1:07:53 AM CDT, tim wade &lt;[email protected]&gt; wrote:</div><blockquote class="gmail_
    quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
    <pre class="k9mail"><div dir="auto">Hello<br><br>I plan to make increment backup for my home dir.<br>It's currently in the size of 1xx GB.<br><br>besides rsync, do you know any other software/service for increment backup?<br><br>Thank you.<br><br></div></
    </blockquote></div></body></html>
    ------05ARS48RL5XN4T0DYBBF8YWLCOJ9SJ--

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From [email protected]@21:1/5 to Andy Smith on Sun Mar 16 07:10:01 2025
    On Sat, Mar 15, 2025 at 08:36:48PM +0000, Andy Smith wrote:
    Hi,

    On Sat, Mar 15, 2025 at 01:18:45PM -0700, David Christensen wrote:
    On 3/15/25 12:50, Tim Woodall wrote:
    git fails to preserve ownership, permissions or timestamps. While this may not be relevant to your usecase, for example backing up /etc would
    be catastrophic (which is why we have etckeeper)

    I am curious -- if I make my /etc directory tree into a version control system working directory (Git or otherwise), please explain how this would be catastrophic.

    /etc has things in it which require specific ownership and permissions.

    Git *does* preserve permissions [1]. For the ownerships (and more accurate mtime, atime and ctime) cf. etckeeper.

    Cheers

    [1] Unless you have /that/ colleague cloning into a stupid
    Windows and then pushing. Don't ask me how I know.

    --
    t

    -----BEGIN PGP SIGNATURE-----

    iF0EABECAB0WIQRp53liolZD6iXhAoIFyCz1etHaRgUCZ9Zq+AAKCRAFyCz1etHa RvTxAJwPxWmFx92kLrfTssELFKdqFju51gCfSYtS2hvHl8ZrJLf8uQFNuMNmv20=
    =XoUr
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From [email protected]@21:1/5 to Tim Woodall on Sun Mar 16 07:10:01 2025
    On Sat, Mar 15, 2025 at 07:50:27PM +0000, Tim Woodall wrote:
    On Fri, 14 Mar 2025, Russell L. Harris wrote:

    On Fri, Mar 14, 2025 at 02:07:53PM +0800, tim wade wrote:
    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    I use git. I keep terminal open running a ssh connection open to the backup system. Whenever I wish to save the state of the system, I
    switch to the terminal and execute git commit. To check the previous
    state of a file, Emacs provides git-timemachine.

    Simple and miminal hassle.

    git fails to preserve ownership, permissions or timestamps. While this may
    [...]

    Half-wrong. Git doesn't preserve ownerships, but it does preserve
    permissions (the POSIX things). As for timestamps, it records the
    last commit timestamp, which may be a "good enough" approximation
    of the mtime or not.

    If you want more, have a look at etckeeper [1], which was made on top
    of git to address ownership et al.

    Cheers
    --
    tomás

    -----BEGIN PGP SIGNATURE-----

    iF0EABECAB0WIQRp53liolZD6iXhAoIFyCz1etHaRgUCZ9ZqRwAKCRAFyCz1etHa RjWcAJ95OAWxQfmQvvUFg1Mr/hJVcdxsggCePzN2VfacoVNKpKucomi87wPQP+o=
    =R5tv
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Tim Woodall@21:1/5 to [email protected] on Sun Mar 16 09:00:01 2025
    On Sun, 16 Mar 2025, [email protected] wrote:

    Git *does* preserve permissions [1]. For the ownerships (and more accurate mtime, atime and ctime) cf. etckeeper.

    Git only tracks the execute bit. And because it always writes a new file
    rather than truncate then write by most editors, the permissions get
    reset very easily:

    $ touch test
    $ chmod 700 test
    $ git add test
    $ echo >>test
    $ ls -l test
    -rwx------ 1 tim tim 1 Mar 16 07:45 test
    $ git checkout test
    Updated 1 path from the index
    $ ls -l test
    -rwxrwxr-x 1 tim tim 0 Mar 16 07:46 test

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From [email protected]@21:1/5 to Tim Woodall on Sun Mar 16 09:50:01 2025
    On Sun, Mar 16, 2025 at 07:51:24AM +0000, Tim Woodall wrote:
    On Sun, 16 Mar 2025, [email protected] wrote:

    Git *does* preserve permissions [1]. For the ownerships (and more accurate mtime, atime and ctime) cf. etckeeper.

    Git only tracks the execute bit. And because it always writes a new file rather than truncate then write by most editors, the permissions get reset very easily:

    I stand corrected. More precisely, it only tracks the first "x" (i.e. the user's). Sorry for the confusion.

    Cheers
    --
    t

    -----BEGIN PGP SIGNATURE-----

    iF0EABECAB0WIQRp53liolZD6iXhAoIFyCz1etHaRgUCZ9aQMAAKCRAFyCz1etHa RhfzAJsH13kM3y8TXvaD+XiDVOknjg5YfACdHY3IqhKLjoO6efvEG4IiReIzam8=
    =ed5/
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Todd Zullinger@21:1/5 to [email protected] on Sun Mar 16 15:20:01 2025
    [email protected] wrote:
    On Sat, Mar 15, 2025 at 07:50:27PM +0000, Tim Woodall wrote:
    git fails to preserve ownership, permissions or timestamps. While this may
    [...]

    Half-wrong. Git doesn't preserve ownerships, but it does preserve
    permissions (the POSIX things).

    Still quarter(?)-wrong. ;)

    It's misleading (at best) to say git preserves permissions.
    It simply tracks whether a file is executable or not. That
    isn't the same as tracking permissions.

    If you set a file to mode 2755, git doesn't record any
    change and will not restore the file with the setgid bit.

    $ git init /tmp/test && cd /tmp/test
    Initialized empty Git repository in /tmp/test/.git/
    $ >foo
    $ git add foo && git commit -m 'add foo'
    [main (root-commit) 0925be1] add foo
    1 file changed, 0 insertions(+), 0 deletions(-)
    create mode 100644 foo
    $ chmod -c g+s foo
    mode of 'foo' changed from 0644 (rw-r--r--) to 2644 (rw-r-Sr--)
    $ ls -l foo
    -rw-r-Sr--. 1 todd todd 0 Mar 16 09:49 foo
    $ git status
    On branch main
    nothing to commit, working tree clean
    $ rm -f foo ; git restore foo
    $ ls -l foo
    -rw-r--r--. 1 todd todd 0 Mar 16 09:50 foo

    As for timestamps, it records the last commit timestamp,
    which may be a "good enough" approximation of the mtime or
    not.

    Notably, for use in a backup system, git doesn't set the
    timestamp to the last change time when it restores or checks
    out a file, so you'd have to do that manually if you're
    using it as backup system and you care about the timestamps.

    Of course, if you're using git as the primary backup tool,
    you might not care deeply about the timestamps, which is
    fine.

    --
    Todd

    -----BEGIN PGP SIGNATURE-----

    iHUEARYIAB0WIQSvlwC4tRNlCF6x+moHOcdGE+n45gUCZ9ba7QAKCRAHOcdGE+n4 5pnLAP9vho/wnpywtbaiBgtUyZDwlU1V7Bpl5S4WOn7yf5bIPAEAjZwA2rgXlZDv gT5UJTuJ+UK5uKjf53TfSPC7te7y8w8=
    =2SLV
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From songbird@21:1/5 to Jerome BENOIT on Sat Mar 22 14:10:01 2025
    Jerome BENOIT wrote:
    Hi,

    backup2l is simple and has been reliable for me for years.

    Cheers,
    Jerome

    interesting, at first glance it might help me out, but
    i don't know for sure. i'm a bit worried though that the
    debian package doesn't look like it is actively maintained.

    i have old backups in tar format. will backup2l allow
    those to be used as a source in such a way that each one
    is not duplicated without having to unpack them first?

    ages ago i used bup and it was ok, but i really did not
    like the presentation for going back and finding things.
    this was a long time ago though and now perhaps it isn't
    so bad. doesn't matter now as i have no old bup anyplace.

    ultimately i really need a way to do backups that will
    deduplicate and must be 100% bulletproof and stable.


    songbird

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From songbird@21:1/5 to Henrik Ahlgren on Sat Mar 22 14:30:01 2025
    Henrik Ahlgren wrote:
    ...
    Git is an awesome tool that can be utilized with a wide variety of
    files, not just software source code or text. However, it faces
    scalability challenges with larger files, such as videos. While
    extensions like git-annex can assist in managing these larger files,
    they can struggle when dealing with tens of thousands or more of files.

    Additionally, Git's core feature of maintaining immutable history is not well-suited for backing up large, dynamic datasets, where retaining the entire history indefinitely may be undesirable. There may also be
    privacy and legal concerns related to maintaining permanent records of
    all changes, since establishing a retention policy is difficult.

    So, most people cant't have their whole $HOME in a Git repository.

    Moreover, storing Git repositories on the same storage device as your original files only safeguards your data against your own mistakes. This
    is analogous to the (opposite) misconception that RAID systems serve as
    a backup solution; they only protect against hardware failures of one or
    more disks, and happily mirror all your mistakes for you. Proper backup system need to protect your data from most (ideally all) data loss
    scenarios.

    ...

    unfortunately git has a fatal flaw as far as i'm concerned.
    it does not keep proper track of file attributes like last changed
    time and date and restore them properly (and yes, i know about
    make and issues with make not working right but to me that's
    not something git should have been set up wrong about to begin
    with, etc.).

    yes, there are patched ways of keeping track and restoring
    that information but until there is some way to have what i
    would consider normal archive data kept as a core feature i can't
    use git how i'd like.


    songbird

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From songbird@21:1/5 to Tim Woodall on Sat Mar 22 14:40:01 2025
    Tim Woodall wrote:
    On Fri, 14 Mar 2025, Russell L. Harris wrote:

    On Fri, Mar 14, 2025 at 02:07:53PM +0800, tim wade wrote:
    I plan to make increment backup for my home dir.
    It's currently in the size of 1xx GB.

    I use git. I keep terminal open running a ssh connection open to the
    backup system. Whenever I wish to save the state of the system, I
    switch to the terminal and execute git commit. To check the previous
    state of a file, Emacs provides git-timemachine.

    Simple and miminal hassle.

    git fails to preserve ownership, permissions or timestamps. While this
    may not be relevant to your usecase, for example backing up /etc would
    be catastrophic (which is why we have etckeeper)

    yes, i just wrote another post about this flaw.


    For a homedir, ownership may not be a problem but permissions might be.
    IIRC, for example, ssh (and sshd) will not work properly if the
    permissions on ~/.ssh are wrong.

    git will also not backup (or even see) extended attributes.

    not without patching somehow. :(


    songbird

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Jerome BENOIT@21:1/5 to songbird on Sat Mar 22 15:00:01 2025
    Hi,

    On 22/03/2025 13:50, songbird wrote:
    Jerome BENOIT wrote:
    Hi,

    backup2l is simple and has been reliable for me for years.

    Cheers,
    Jerome

    interesting, at first glance it might help me out, but
    i don't know for sure. i'm a bit worried though that the
    debian package doesn't look like it is actively maintained.

    It does not look like either that it needs a major fix.
    Note that the upstream material is as fresh as the package.

    I see this as a stability factor: this is what you ask for
    a backup tool.


    i have old backups in tar format. will backup2l allow
    those to be used as a source in such a way that each one
    is not duplicated without having to unpack them first?

    No.

    But somehow the reverse is true:
    you can access the material produced by backup2l without it.
    If backup2l is no more maintained or get broken for any reason
    you can still get the saved material by hand (I will not say that
    it would be easy). Anyway, backup2l is written in an easy to
    read bash.


    ages ago i used bup and it was ok, but i really did not
    like the presentation for going back and finding things.
    this was a long time ago though and now perhaps it isn't
    so bad. doesn't matter now as i have no old bup anyplace.

    ultimately i really need a way to do backups that will
    deduplicate

    I do not see what you mean.

    and must be 100% bulletproof and stable.

    backup2l is stable (it uses stable tools (as tar) and it is written in an easy-to-read bash).
    For the bulletproof part, the choice of the compressor might be crucial: backup2l allows to
    choose your favorite one.



    songbird

    Cheers,
    Jerome



    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From songbird@21:1/5 to Jerome BENOIT on Sat Mar 22 15:30:01 2025
    Jerome BENOIT wrote:
    songbird wrote:
    ...
    ultimately i really need a way to do backups that will
    deduplicate

    I do not see what you mean.

    my old backups are not incremental so they will contain a
    lot of files that would be identical copies to other backups.


    and must be 100% bulletproof and stable.

    backup2l is stable (it uses stable tools (as tar) and it is written in an easy-to-read bash).
    For the bulletproof part, the choice of the compressor might be crucial: backup2l allows to
    choose your favorite one.

    sounds ok to me, but i may also decide for some things to
    not compress at all. very large archives of materials
    already compressed would be one i'd like to skip messing
    with at all.


    songbird

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Jerome BENOIT@21:1/5 to songbird on Sat Mar 22 18:20:01 2025
    Hi Again,

    On 22/03/2025 15:14, songbird wrote:
    Jerome BENOIT wrote:
    songbird wrote:
    ...
    ultimately i really need a way to do backups that will
    deduplicate

    I do not see what you mean.

    my old backups are not incremental so they will contain a
    lot of files that would be identical copies to other backups.

    backup2l is incremental



    and must be 100% bulletproof and stable.

    backup2l is stable (it uses stable tools (as tar) and it is written in an easy-to-read bash).
    For the bulletproof part, the choice of the compressor might be crucial: backup2l allows to
    choose your favorite one.

    sounds ok to me, but i may also decide for some things to
    not compress at all. very large archives of materials
    already compressed would be one i'd like to skip messing
    with at all.

    Not compressing the tarballs is possible.






    songbird


    hth,
    Jerome

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From songbird@21:1/5 to Jerome BENOIT on Sat Mar 22 22:30:01 2025
    Jerome BENOIT wrote:
    ...
    Not compressing the tarballs is possible.

    i don't think it is really deduplicating but that is ok
    for now. what i've been playing with this afternoon seems
    to be going ok.

    i'm not sure i have a daily run that is working since i
    don't always leave my machine on, but that too is also ok
    as there's a good chance i'll just create a script that i
    can run manually on my own as i shutdown at night.


    songbird

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Jerome BENOIT@21:1/5 to songbird on Sat Mar 22 23:10:01 2025
    Ho,

    On 22/03/2025 21:54, songbird wrote:
    Jerome BENOIT wrote:
    ...
    Not compressing the tarballs is possible.

    i don't think it is really deduplicating but that is ok
    for now. what i've been playing with this afternoon seems
    to be going ok.

    i'm not sure i have a daily run that is working since i
    don't always leave my machine on, but that too is also ok
    as there's a good chance i'll just create a script that i
    can run manually on my own as i shutdown at night.

    You can play with the script /etc/cron.daily/zz-backup2l

    On my box, anacron does the job .

    Best wishes,
    Jerome



    songbird


    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)