Deleting millions of files in a linux directory

Today I found myself with a directory containing about 6 million files which shouldn’t have been there in the first place. Any attempt to ls or rm failed.

Thank god for the “strace” command that I discovered a few weeks ago I was able to delete the files I wanted.

Here are the steps :

cd directory-with-millions-of-files
ls -l &  
#[1] 7453
strace -p 7453 2>&1 | grep lstat | cut -d ‘”‘ -f 2 > /tmp/to_delete.list

of course 7453 needs to be changed to the proper pid wait about a minute and then ctrl-c to stop strace and kill -9 7453 to stop the ls

for i in `cat /tmp/to_delete.list`; do rm ./$i; done

What we did is basically:

• launched the ls -l as a background process which outputs the process id.
• strace this pid which shows us all that’s going on internally in ls -l.
• strace echoes to much stuff on the std and err console. we apply some refirecting, grepping and cutting to get the filename which we write into a tmp file
• after about a minute on my system, the file already has a million lines, we kill strace and ls
• last but not least, we loop over the lines in out tmp file and start deleting the files.
• You could change your grep to only target specific files/extensions

Have fun