Automatic File Removal script

[Joseph Bague]

I have a daily routine to delete a five day's old file (Oracle archive log). Can you give a me sample script that will delete a five day's old file.

Many thank's

[H.Merijn Brand (procura]

using find


find . \! -mtime 5 -exec rm -f {} \;

test first!

[Stefan Farrelly]

We always use;
cd
find . -mtime + -exec rm {} \;

The + is usually +7 or +14 for one or two weeks or +30 for a month. Im not sure what the earlier reply using find . \! -mtime does!?

[H.Merijn Brand (procura]

\! -mtime 7
=
-mtime +7

but older (read very old) find versions do not support + notation

exclamation mark is 'not' or 'reverse' and is escaped for (t)csh users

HTH

BTW GNU find has many more nifty features

[V. V. Ravi Kumar_1]

hi,
find -type f -mtime +5 -exec rm {} \;

here if u say 5 it deletes files which are modified exactly 5 days back, if us say +5 it deletes files modified 5 or more days back, and -5 deletes files which are modified during last 5 days. here mtime modification time, u can also specify atime (accessed time) there.
type f indicates it is a file.

regds

[James Beamish-White]

You need to be very careful with these, to make sure that your don't start using variables which may be empty, and which may effectively rm -R / (trust me - I've done it ;-)

I would use something like:


find /your/log/path -name "*log" -mtime 5 | xargs rm

The xargs is a bit better on performance than find/exec. The use of a full path means you won't remove anything from another path, and the -name part again limits what you will be deleteing.

Cheers!
James

[George Petrides_1 · Accepted solution]

Just because I saw people before making mistakes in the find path... you might want to specify a name pattern for the archive logs... e.g.
find /ORACLE_ARCH_DIR -mtime +5 -name "ora_arch*" -type f -exec rm -f {} \;
In UNIX you have to make sure you protect yourself from making mistakes else the results can be catastrophic...
George

[Jean-Luc Oudart]

Here's is a scripts we run on a monthly basis.
1st it prints the list of file we're about to delete, then run the deletion.
We delete files with different patterns.
See attachment

Jean-Luc

[Vytautas Vysniauskas]

Hi,

Some of proposed solutions are simple but realy dangerous. Removal of old files sometimes can be very tricky. There are 2 MUST rules for any script or program performing cleanup of filesystem

1. Do not follow symlinks in the directories (even if it is the top level directory for cleaning).

2. Remove only regular files and EMPTY directories.


Personally I prefer to use a 'tmpwatch' program which comes with many versions of RedHat Linux. Source code is available on Redhat web site.
The source is tested to compile and work correctly on
HP-UX as well.


Vytas.

[Joseph Bague]

Thank's to every one :)