On the qmail list [EMAIL PROTECTED] wrote:
>On Mon, Feb 08, 1999 at 11:16:35AM +0100, Harald Hanche-Olsen wrote:
>
>> You may have to change some path names. Also, note that this job
>> keeps the last 10 log files uncompressed, for ease in checking the
>> lateste logs. I run cyclog with the arguments -s304000 -n30; you may
>> have to adjust these parameters.
>
>I've actually been wondering about this for some time now: With syslog and
>nightly log rotation, we always have at least 7 days (or whatever is the
>rotational cycle) worth of logs to look at. Often this is useful if we want
>to track messages a few days old. With cyclog, if there is a lot of
>activity suddenly, we may end up with logs a couple of days old. Has anyone
I am not familiar with cyclog; Debian has a script called
savelog, which I had supposed does about the same thing.
^^^ (I'm not so sure anymore)
What I do is that in my cron script, run once a day (or once a
week, whatever, replace below), I test if the size of the log
file over a certain limit (x kB), and if it is I run the script
(for n cycles).
That means that the minimum days of logs is n, since logs are
not rotated faster than that, and the maximum size of logs is
((x+m)*2) kB + gzip((x+m)*(n-2) kB)), where m is the (unknown)
maximum size of one day of logs.
Good enough for me, but of course YMMV.
--
#include <std_disclaim.h> Lorens Kockum
