Find Command

[John Carver]

I am writting debug logs to the /tmp directory for an application fix. These logs fill up the /tmp directory very quickly and I only need logs for the last 2 hours at any given moment in time. I want to include the find command in a script that continuously runs every 15 minutes and finds files with a specific name older than 2 hours and deletes them. I've been running the example below for awhile. It deletes files older than 1 hour and runs every hour so I have files anywhere between 1 and 2 hours old. Now I need to run every 15 minutes but delete files older than 2 hours. In other words, I need to retain roughly 2 hours of logs at all times.

I can't see if there's a way to do this with find or if I have to resort to something else.

#!/usr/bin/sh

while true
do
find /tmp -name "udapiserver*" ! -newer /tmp/findref -exec rm {} \;
touch /tmp/findref
sleep 3600
done

exit

[James R. Ferguson]

Hi:

One way is to use a small Perl snippet :

...
WHEN=$(perl -MPOSIX -e 'print strftime "%Y%m%d%H%M\n",localtime(time-(2*60*60))')
touch -amt ${WHEN} /tmp/findref
...

Now, your 'findref' file has a modification timestamp two hours ago (2*60*60 seconds ago) and you can awaken an compare to it as you need.

Regards!

...JRF...

[Hein van den Heuvel]

Since you are not asking anything tricky from find (and even if you did), you may want to consider to have a shell or perl script do a glob and look at the date attribute to decide whether a file is a keeper or not.


for example:

$ perl -e 'for () { unlink if -M $_ > 2/24 }'

The -M function returns the age of the file as a floating point value with unit days.

So for 2 hours you want to compare using 2/24


hth,
Hein

[James R. Ferguson · Accepted solution]

Hi (again) :

By the way, if you want to avoid poor 'find' performance, use the '+' terminator to the '-exec' argument like:

# find /tmp -type f -name "udapiserver*" ! -newer /tmp/findref -exec rm {} +

This causes multiple arguments to be collected for every 'rm' process spawned instead of forking one 'rm' for every file to be removed.

Regards!

...JRF...

[John Carver]

Thanks James!

I've incorporated the Perl snippet and the ref file has a timestamp exactly two hours old. I've also taken note of the "+" terminator, I did not know this.

Hein,

I did not try your suggested syntax, but will when I get the chance. Thanks for the response.

[Dennis Handly]

>I only need logs for the last 2 hours at any given moment in time. I want to include the find command in a script that runs every 15 minutes

If you don't want to use the perl solution to create the reference files, you'll need to have a sequence of 8 reference files and use and touch the appropriate one each 15 minutes.

[Hein van den Heuvel]

>> I did not try your suggested syntax, but will when I get the chance.


Try a non-destructive version first.

Just after your purge runs, rerun the find, and compare with output from:

perl -e 'for () {$m = -M; print qq($m $_\n) if $m > 2/24 }'

That would just be to get familiar with this method.

for () {...} # walk list, setting $_, executing code block each time
# glob list of filenames matching specification
$m = -M; # store file modification date in varariable $m

print qq($m $_\n) if $m > 2/24 # real work.

hth,

Hein