Can I delete everything in /var/log? Or should I only delete files (recursively) in /var/log but leave folders?

Does anyone have a good rm command line? (My admin skills leave me nervous.)

Note: I am using Debian. I am not sure what version.

  • 3
    Deleting log files is a bad idea (you'll also need to find every running process that has it's own log file and "kill -HUP" it, a soft restart that will result in the program recreating any necessary log files). I would strongly advise against deleting log files, rely on utilities like logrotate to manage the contents of /var/log for you automatically (it does stuff like HUP the processes) If I may I'd like to tackle this from a different angle. What problem are you trying to resolve that's led you to consider this? – Twirrim Sep 29 '10 at 17:34
up vote 21 down vote accepted

Instead of deleting the files you should rotate them, e. g. using logrotate.

You never know when you'll actually need the logs from some time ago, so it's better to archive them (up to a reasonable age, e. g. 3 months).

logrotate can compress your old log files so they don't occupy a lot of disk space.

  • 3
    logrotate can also delete the oldest files. – Kevin M Sep 28 '10 at 12:13
  • 8
    Well, IMHO deleting all logs can make perfect sense in some cases. For example I want to build a Virtial Machine image to be used for new deployments. Needless to say I would like it to be a really clean system without any logs, histories, caches etc. saved. – Ivan Oct 15 '12 at 18:54
  • 1
    Sorry, but looking at three months old log files is archeology. If you collect logs to identify problems, then evaluate them quickly. – countermode Aug 14 '14 at 14:40
  • 4
    @countermode You are never in the mood for nostalgia? Like looking at the 3 month old log files thinking about good ol' times? – Broco Aug 14 '14 at 14:51
  • OK, I see the command. How to use it? man logrotate says use it in cron. I suppose with the -f option? – SDsolar Aug 13 '17 at 23:30

If you delete everything in /var/log, you will most likely end up with tons of error messages in very little time, since there are folders in there which are expected to exist (e.g. exim4, apache2, apt, cups, mysql, samba and more). Plus: there are some services or applications that will not create their log files, if they don't exist. They expect at least an empty file to be present. So the direct answer to your question actually is "Do not do this!!!".

As joschi has pointed out, there is no reason to do this. I have debian servers running that haven't had a single log file deleted in years.

  • I didnt realize that. good to know. +1 + changed my accept. – user274 Sep 28 '10 at 21:31
  • I have just done this. Wish! I had read this answer earlier – VarunAgw May 30 '15 at 18:41
  • There are valid reasons to remove log files, IMHO. For instance, you are exporting a virtual machine for use by others, but you don't want the virtual machine image to contain details of everything that has happened before exporting. – a3nm Nov 16 '17 at 20:43

Delete all files:

find /var/log -type f -delete

Delete all .gz and rotated file

find /var/log -type f -regex ".*\.gz$"
find /var/log -type f -regex ".*\.[0-9]$"

Try run command without "-delete", to test it.

  • I found this useful to clear up a Vagrant box's log files before packaging. – Rudolf Vavruch May 1 '17 at 15:01

I'm cloning virtual machines from a master. It makes perfect sense to clear the log on the master so that when you boot the clones you won't get the master's log. I did in tcsh:

cd /var/log
foreach ii ( `find . -type f` )
foreach? cp /dev/null $ii
foreach? end

which clears the logs but keeps the files.

  • This should be restricted to a use-case like you describe. – Sven Nov 9 '12 at 10:41
  • 4
    In bash: find /var/log/ -type f -exec cp /dev/null {} \; – gerard Jun 1 '16 at 20:28

Cleaning all logs on a Linux system without deleting the files:

for CLEAN in $(find /var/log/ -type f)
do
    cp /dev/null  $CLEAN
done

Samba (/var/www/samba) creates log file-names with ip addresses, you may want to delete them:

for CLEAN in $(find /var/log/samba -type f)
do
    rm -rf $CLEAN
done
  • 2
    Useful script . – Anmol Singh Jaggi Jun 12 '16 at 12:50
  • You could swap cp /dev/null $CLEAN by > $CLEAN. – ThoriumBR Oct 5 '17 at 20:34

You can use the option ctime to find old files... for example:

find -ctime +30

As bindbn explain, first try the find fetch files and after use the option delete :D

I've implemented a simple cleaner here:

https://github.com/Lin-Buo-Ren/Coward-Unix-Log-Cleaner

It simply:

  • Deletes file names with the following logrotated filename patterns under /var/log
    • ^.*/.+\.[[:digit:]]+(\.[[:alpha:]]+)?$
    • ^.*/.+\.old$ (case-insensitive)
  • Truncate/Empty files with filenames with the following log filename patterns under /var/log
    • ^.*/.+\.log$ (case-insensitive)

/var/log often has permissions of drwxrwxr-x, so is not user writable unless the user is root or belongs to a privileged group. That means new log files cannot be created by non-privileged users.

Applications that expect to log to a point within /var/log will often touch a file into existence somewhere in the /var/log hierarchy during install time (which often occurs with elevated privileges), and will chmod and possibly chown it at that time to permissions appropriate for the unprivileged users who will be using the application.

Apache logs, for example, are usually written to by nobody, who is a user with as few privileges as possible for Apache to get its job done without putting the system at undue risk. But even a more run-of-the-mill application often expects to be able to write to a logfile in /var/log.

So what happens if the logfile, and the path to the logfile don't exist? That's entirely up to the application. Some applications will quietly skip logging. Others will create a lot of warnings. And others will simply bail out. There's no hard-fast rule; it's up to the vigilance of the application developer, as well as how critical the developer considers its ability to log. At best the application will attempt to either write to, or possibly create and then write to a log file at a destination within /var/log, and will find itself unable to do so because it's being run by a user who doesn't have privileges to write into that part of the filesystem.

So the short answer is no, don't delete everything in /var/log -- it breaks the contract users with sufficient privileges to do such things have with the applications that run on their system, and will cause some noise, some silent failure to log, and some all-out breakage.

The appropriate action to take is to set up logrotate with appropriate config files. Typically rotation will be associated with a cron job. Rotation can be interval based, or size based, or both. It's even possible to set up rules that avoid interval based rotation if the logfile is still empty when the interval expires. Rotation can include mailing of logfiles, compression, deletion, shredding, and so on.

The average user wouldn't need to be too concerned about log rotation. Developers would probably want to ensure that logs they use have rotation rules established. In fact, it is likely good manners on the part of developers to set up log rotation at install time for any software-specific logs that software will be creating and writing.

function goodbyelogs {
find /var/log -type f
}

for i in return $(goodbyelogs);
do sudo cat /dev/null > $i;
echo "Log $i has been cleared";
done

make an executable script and try run as root if sudo isnt working for you

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.