undelete linux

D.P. · 25 Jul 2007 at 12:33

Probably too late to do anything now but perhaps for future reference this would be useful.

I made the horrible mistake of making a bash script which batch deletes some files, and accidentally put the the wrong directory name in. 3 days of simulation time wasted, and on monday I have to hand in a report with my result. Should just about be able to get the work done on time.

But, what is the best way to recover files? I only have a single HD so unmounting the drive would be a pain.

The files were text files, containing 100 lines of text which conisted of more or less random real valued numbers (0.264,32.8972,.etc)
Thus I do not know the starting string, just the size, old location and filename.

Grep'ing the filesystem proved to be unproductive.

I found this script which can provide partion info:

Code:

echo "DEVICE DIRECTORY FS-TYPE" > tmp; mount | cut -d" " -f1,3,5 | sort >> tmp; cat tmp | column -t | sed -e "1s/.*/`tput smso`&`tput rmso`/"

tntcoder · 25 Jul 2007 at 13:04

I think you can use debugfs to access the raw file system data to recover files, it probably wouldn't be easy though, ive never tried it but i heard thats the tool for the job. Maybe theres some 3rd party commercial apps for it as well.

Dj_Jestar · 25 Jul 2007 at 13:16

I far prefer to bare in mind that "Once it's gone, it's gone!" Rather than hack together a fallback.

D.P. · 25 Jul 2007 at 13:31

Dj_Jestar said:
I far prefer to bare in mind that "Once it's gone, it's gone!" Rather than hack together a fallback.

yep, I have decided the data is gone and I have started my simulations again.

Off-topic:

Luckily I had the the Mean value and standard deviation of the data saved. This can allow me to do some testing, e.g. with a t-test, but I know the data was non-parametric so ideally I needed the raw data to perform a rank based test.