Re: rsync --delete
On 2020-10-16 15:09, Mike McClain wrote:
I've been using rsync to backup to a flash drive but it's not
performing exactly as I expected.
The man page says:
--delete delete extraneous files from dest dirs
A section of the backup script is so:
Params=(-a --inplace --delete);
I try to use all lower case letters for variable names and all upper
case letters for constants.
'$Params' -- Variable, function, media, source, destination, etc., names
are critical to understanding. I don't like 'Params' -- it's too
generic. I would prefer 'rsync_opts'.
I use braces whenever evaluating a variable -- '${Params}'. I forget
the details why, but I do recall that this practice is important.
I don't use Bourne arrays, and I barely understand how the shell
interpolates lists and preserves items containing whitespace. When I
can't figure it out, I switch to Perl.
-a -- I use this option for backups.
--inplace -- If your backup media cannot hold the last backup with
enough room for the next backup, get bigger backup media.
--delete -- I use this option for backups.
Additional options I use in my rsync(1) backup script:
--one-file-system -- I use this option for backups. The bottom-level
backup script is driven by a higher level script that backs up multiple
filesystems on multiple hosts via configuration files.
-e ${SSH} -- where:
SSH=/usr/bin/ssh
(To be pendantic, I should put single parentheses around the RHS?)
When doing rsync(1) backups over SSH, I seem to recall that rsync(1)
will use whatever login shell is specified for that account. Some of my
machines are FreeBSD, and the root login shell is tcsh(1). Strange
things were happening before I discovered the '-e' / '--rsh' option.
Flash=/sda/rpi4b
Mixed case variable name -- as above.
Flash -- I would prefer 'backup_media'.
Is /sda the mount point for your backup media? If so, that is confusing
-- 'sda' implies '/dev/sda', which should be your system drive (e.g.
root). I would label the backup filesystem 'backup-rpi4b' and mount it
at '/mnt/backup-rpi4b' or '/media/backup-rpi4b' (your desktop might be
able to do this for you).
cd /home/mike
I prefer scripts that I can run from anywhere. The script should know
where to do its work, either from built-in variables, environment
variables, resource script, configuration file, command-line arguments/
options, etc.. Sophisticated scripts can draw this information from
multiple resources, with some hierarchy of precedence. (I would switch
to Perl if I wanted to get fancy with options.)
If the script must change the working directory, I would display that --
'set -x', 'cd ...', and 'set +x'.
[ ! -d $Flash/mike ] && mkdir $Flash/mike;
I would set a variable for the destination directory. Something like:
dst="${Flash}/mike"
I would do an old-school 'if' block and display that a directory is
being created -- 'set -x', 'mkdir ...', 'set +x'.
# exclude compressed files and the contents of most of the .* directories
/mc/bin/mk_rsync_exclude.sh
What is /mc?
mk_rsync_exclude.sh creates '.rsync_exclude' in the current working
directory?
A problem with dynamically generated code/ arguments/ options is: how do
you know what code was actually run in the past?. Now you need a log
file. I prefer to use a static configuration files and static scripts,
and check everything into a version control system. My previous backup
scripts generated log files, but my current rewrite does not have that
feature (yet).
I prefer to keep my backup scripts, include files, etc., outside of the
backup source and destination paths (except when backing up the
filesystem that contains the backup stuff). This reduces confusion, and
might prevent a chicken-and-egg situation.
echo /usr/bin/rsync $Params --exclude-from=/home/mike/.rsync_exclude . $Flash/mike
/usr/bin/rsync $Params --exclude-from=/home/mike/.rsync_exclude . $Flash/mike ||
echo rsync $Params --exclude-from=/home/mike/.rsync_exclude . $Flash/mike Failed $? ;
You cut and pasted the following code three times:
/usr/bin/rsync $Params --exclude-from=/home/mike/.rsync_exclude .
$Flash/mike
DRY:
https://en.wikipedia.org/wiki/Don%27t_repeat_yourself
I prefer 'set -x', 'command ...', and 'set +x' when I want to see what
the shell is actually doing (which might not be the same output as 'echo
...').
I use 'set -e' at the top of my scripts so that the shell will stop and
display an error message if a script command fails.
/usr/bin/rsync -- I also use absolute paths for tools. But, I put them
into upper-case variables at the top of my script.
--exclude-from -- It is too easy to screw up exclude specifications and
exclude a file you need. Therefore, I backup entire filesystems.
When invoking rsync(1), I make sure that SRC and DEST are directories,
that their paths are absolute, and that their paths end with '/'. This
prevents confusion and works as I expect.
If I delete a file from my home directory then backup over last
week's copy the deleted file stays in the backup directory and these
build up over time.
That is another good reason not to use 'exclude' options.
Am I misusing rsync or am I just not understanding how it works?
rsync(1) is a very flexible tool. I use only a small subset of its
features, and I do so in a consistent manner. Thus, I can obtain the
results I need and I have confidence that they are correct.
David
Reply to: