Harald Hanche-Olsen <[EMAIL PROTECTED]> writes:
> - Russ Allbery <[EMAIL PROTECTED]>:
> | There are cases where I'd rather fill the disk than lose logs. And
> | predictable log rotation is actually something I'd rather like to
> | have for a lot of things. I've wanted cyclog to be able to rotate
> | on a time interval rather than on a size limit for a while, but I do
> | realize that it's unlikely to happen since it's rather far away from
> | the intended purpose of cyclog.
> Easy: Give cyclog such a large limit on the file size and the number
> of files that it is sure to fill the disk before hitting its own
> limit. Then run a cron job to rotate the logs.
Er... how exactly is that going to work? I want to rotate the logs on a
daily (or weekly, perhaps) basis at precisely the same time each day so
that I don't have to spend a lot of time or thought on getting consistent
time intervals for statistics. How do I tell cyclog to stop writing to
one log and start writing to a new one?
As far as I can tell from the cyclog documentation, it doesn't support
being sent a HUP to rebind to a new file the way syslogd does, and I can't
just kill and restart it because it's part of a pipe.
> Just make sure the file size limit is not too large, since there must be
> some finished log files for the log rotater to rotate.
Or is your suggestion to set cyclog's file size limit low enough that it
will rotate logs at least once a day and for the log rotator to patch
together daily log files from cyclog's old logs? I suppose that would
work, come to think of it....
Hm. Maybe I'll do that. Unfortunately, on some hosts, that's a really
small log size limit.
--
Russ Allbery ([EMAIL PROTECTED]) <URL:http://www.eyrie.org/~eagle/>
