logrotate(8) and copytruncate as default

Lennart Poettering mzerqung at 0pointer.de
Sat Jun 29 00:42:11 UTC 2013

On Fri, 28.06.13 16:33, Matthew Miller (mattdm at fedoraproject.org) wrote:

> On Fri, Jun 28, 2013 at 09:48:58PM +0200, Zbigniew Jędrzejewski-Szmek wrote:
> > 'uid' as default doesn't make sense, at least with the current way of accesing
> > logs. It is really nice to be able to view messages about a service
> > interleaved from various sources. Now when you say 'journalctl -u httpd',
> > you get logs from the processes in the httpd.service, but also messages
> > from systemd about this service, and possibly information about coredumps
> > and hopefully, soon, messages from setroubleshoot. With 'uid', and looking
> > only at one file, you'd get only the first kind.
> But many cases for accesssing HTTP logs don't need that kind of mixed
> information. The service is running fine, but people or programs may want to
> look for information about visitors, hits, referrers, etc., etc.
> The sysadmin wants the unified view, but the web admin just needs http logs.
> Never mind the basic principle of least necessary privilege -- all that
> other stuff is just "noise".

That's why we are so strong on filtering the dataset when you look at
it. May I recommend watching this video?


if you want apache's logs, then run "journalctl -u httpd"... 


Lennart Poettering - Red Hat, Inc.

More information about the devel mailing list