Link to home
Start Free TrialLog in
Avatar of Dan Craciun
Dan CraciunFlag for Romania

asked on

Shell script to delete old backups

Hi,

I need a script to delete old backups created by Cobian Backup on a NAS (for some reason the program does not delete the old backups, gave up trying to find out why).

The NAS is running Synology 5, with ssh access enabled. The built-in shell is ash.

The backups are basically one full, followed by 0 or more incremental backups, then the next full one etc.
The backups are stored, for each user, in /volume1/homes/username/. See the structure below.

What I need is to delete, for each backed up folder, all the backups apart from the most recent full backup and its incremental backups. After the script runs, I should have, for each user and each folder, a single full backup and 0 or more incremental backups with a more recent date than the full backup.

How can I do that?

Thank you.

The file structure is as follows:
volume1/homes/user1/folder1 2014-12-34 56;78;90 (Full)
volume1/homes/user1/folder1 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user1/folder1 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user1/folder1 2014-12-34 56;78;90 (Full)
volume1/homes/user1/folder1 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user1/folder1 2014-xx-xx xx;xx;xx (Incremental)
volume1/homes/user1/folder1 2014-12-34 56;78;90 (Full)
volume1/homes/user1/folder1 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user1/folder2 2014-12-34 56;78;90 (Full)
volume1/homes/user1/folder2 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user1/folder2 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user1/folder2 2014-12-34 56;78;90 (Full)
volume1/homes/user1/folder2 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user1/folder2 2014-xx-xx xx;xx;xx (Incremental)
volume1/homes/user1/folder2 2014-12-34 56;78;90 (Full)
volume1/homes/user1/folder2 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user2/folder1 2014-12-34 56;78;90 (Full)
volume1/homes/user2/folder1 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user2/folder1 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user2/folder1 2014-12-34 56;78;90 (Full)
volume1/homes/user2/folder1 2014-xx-xx xx;xx;xx (Incremental)
...
volume1/homes/user2/folder1 2014-xx-xx xx;xx;xx (Incremental)
volume1/homes/user2/folder1 2014-12-34 56;78;90 (Full)
volume1/homes/user2/folder1 2014-xx-xx xx;xx;xx (Incremental)
...

Open in new window

Avatar of omarfarid
omarfarid
Flag of United Arab Emirates image

How you distinguish between full and incremental backups and are these file names or folder names?
Avatar of Dan Craciun

ASKER

The full backups end in (Full), the incremental ones in (Incremental). For ex, see below a real listing:

ServerBackup> ls -l
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-01 09;30;17 (Full)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-09 09;30;28 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-10 09;30;24 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-13 09;30;19 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-14 09;30;28 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-15 09;30;16 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-16 09;30;24 (Full)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-17 09;30;30 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-20 09;30;28 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-21 09;30;24 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-22 09;30;27 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-23 09;30;19 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-24 09;30;24 (Full)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-27 09;30;26 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-28 09;30;33 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-29 09;30;28 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-30 09;30;30 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-10-31 09;30;20 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-03 09;30;23 (Full)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-04 09;30;27 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-05 09;30;17 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-06 09;30;24 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-07 09;30;27 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-11 09;30;20 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-12 09;30;30 (Full)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-13 09;30;24 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-14 09;30;20 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-17 09;30;26 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-18 09;30;26 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-19 09;30;17 (Incremental)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-11-20 09;30;17 (Full)
drwxrwxrwx    3 user1   users         4096 Nov 25  2009 Profiles 2014-12-19 09;30;27 (Incremental)

Open in new window

should
volume1/homes/user1/folder1 2014-xx-xx xx;xx;xx (Incremental)
be deleted because it is older than
volume1/homes/user1/folder2 2014-12-34 56;78;90 (Full)
or should it be kept because it is more recent than
volume1/homes/user1/folder1 2014-12-34 56;78;90 (Full)
?

i.e. is
volume1/homes/user1/folder2 2014-12-34 56;78;90 (Full)
part of the same backup set as
volume1/homes/user1/folder1 2014-12-34 56;78;90 (Full)
or is it part of a separate backup set that should be pruned separately?
It's a separate backup. Each user has 1 or more folders that they chose to backup, i.e. Desktop, Documents, etc, so for ex folder1 is Desktop, folder 2 is Documents, folder 3 is Pictures etc. And each can have it's own schedule.
ASKER CERTIFIED SOLUTION
Avatar of ozo
ozo
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
At the end of the day, what you should prune more often will be the full backups, since they will be the biggest hogs.

So why not just do a find to delete fulls older than x days (in this example older than 30 days):

find /volume1/homes/user1 -name *Full* -mtime +30 | xargs rm

Open in new window


For incrementals, you could do similar:

find /volume1/homes/user1 -name *Incremental* -mtime +30 | xargs rm

Open in new window


It's not exactly what you asked, but it uses KISS methodology (Keep It Short and Simple).
@Carlos: if you notice, all the folder have the same date: Nov 25  2009. That's probably the reason why Cobian does not delete old backups (it's searching for timestamps, not folder names).

@ozo: thank you. I'll try to understand the script first, will post back with what I could not figure out by myself. I'm more of a powershell guy, so Perl is a little greek to me.
Here's what I got so far:
perl -e 'for( @ARGV ){  # create a short script (-e). 
#Run the block for each of the arguments array (basically for each folder in /volume1/homes/). 
#The current folder name (/volume1/homes/user1) will be stored in $_
  my %b; # declare a hash b. Does it have an initial value?
  m#(.*?)\s\d{4}-\d\d-\d\d\s\d\d;\d\d;\d\d\s\((Full|Incremental)\)# && unshift @{$b{$1}},$_ for grep -d, <"$_"/*>;
#then it becomes fuzzy: for each file in the current path ($_/*) search for names like folder2 2014-12-34 56;78;90 (Full)
#and then prepend the current path (/volume1/homes/user1 for ex) to what array??? $b{$1} contains what? 
# How is the hash constructed?
  for( values %b ){      #for each value of the hash b
     my($f,$i);               #declare 2 local variables, $f and $i. Are they the (key,value) pair for b?
     system "rm","-f",$_ for grep {$i|=$f,$f|=/\(Full\)$/,$i}@$_;
#again, fuzzy: remove (run rm -f) for each file that matches the grep. But I cannot figure out the grep. It reverses the 
#file order and preserves only those after the first folder with "(Full)" in the name?
  }
}' /volume1/homes/*

Open in new window


I tested the script (replacing system "rm", "-f" with system "echo") and it seams to work. But I can't still figure out why :)

Can you enlighten me, ozo, or shall I post this as a new question?
# declare a hash b. Does it have an initial value?
It is initially empty
# $b{$1} contains what?
for names like folder2 2014-12-34 56;78;90 (Full)
$1 would contains "folder2"
@{$b{$1}} would contain a list associated with folder2

# $f and $i. Are they the (key,value) pair for b?
$_ is the value for b
@{$_} is the list of names in that value
$f is set the first [or last in date order] time we see a (full) backup
$i is set on the next name we see
Thank you Ozo. I tested the script and it works, with minor tweaks (it's /volume1 and "-rf").

The delete part takes a long time, so I would like to print the name of the folder being deleted as a "proof of life", so I know the script is still running.
Would this work?
perl -e 'for( @ARGV ){
  my %b;
  m#(.*?)\s\d{4}-\d\d-\d\d\s\d\d;\d\d;\d\d\s\((Full|Incremental)\)# && unshift @{$b{$1}},$_ for grep -d, <"$_"/*>;
  for( values %b ){
     my($f,$i);
     for (grep {$i|=$f,$f|=/\(Full\)$/,$i}@$_) {
         system "echo", "Deleting $_";
         system "rm","-rf",$_ ;
     }
  }
}' /volume1/homes/*

Open in new window

Yes, although
print "Deleting $_\n";
may be easier.
It probably takes a while deleting since it does multiple "rm file" commands.
It would be a lot faster if you issue a "xargs | rm" command.
Thank you.