Where does the -i option in sed store the lines matching a regex?

Question

I am trying to release space in my server's hd so I'm matching old dates in a log file with sed and it is decreasing but my root partition is now using 98% percent of it's space instead of 91% it was using before I invoke sed. I did this sed -i '/^Jan/ d' file.log and my log file decrease from 6G to 400M. I'm using find / -size +1G trying to search a temporal file or something like that but it's not matching new files. Edit: I did echo -n > file.log to avoid a service crash but it's temporal because the issue remains

Answer 1

The --in-place option to sed cannot really edit a file in place, so it actually does the equivalent of:

if sed ... input > /tmp/output$$ ; then
    mv /tmp/output$$ input
else
    rm /tmp/output$$
fi

which means prior to the mv command you can double (or worse) the size of the input file on disk. When file-systems reach nearly full, they become dreadfully less efficient.

It sounds like the larger problem is that you are just plain out of disk-space or you have /var/log and /tmp mounted on the same partition as / and if / is on a small partition, you will overfill it now and next week and the week after that, etc.

Look at mount and df /tmp and df /var/log to see how your disk is set up. You either need to:

1. fix the disk mounting
2. get a larger disk
3. look into logrotate
4. stop logging so much
5. or some combination of the above

Answer 2

Running strace on the sed -i should make it pretty obvious where it's putting the temporary files. But the easiest thing to do is not use -i. Redirect the output to a file on a partition where you have space.