User config file management. Include is good. Static is problematic.
Naoki
naoki at valuecommerce.com
Wed Sep 20 08:01:22 UTC 2006
Hi all,
In Fedora various applications/tools/utils (like yum, Apache, cron,
xinetd, logrotate, pam, etc) gracefully handle user configuration files
by providing a "*.d" directory in which user config files can be placed
(that do not conflict with the base package's configuration files) and
are then include'ed into the config on host/application startup. This
allows user configurations to be installed on systems via RPM without
conflicts and other clunkiness.
Squid is one package that is the exception with a squid.conf file owned
by the RPM and no way for user definitions to be included ( that I know
of ). I'm just using squid as an example, not blaming it for anything by
the way as I know many other packages operate this way like postfix etc.
Now normally in small environments this isn't a problem as squid updates
would drop in a squid.conf.rpmnew file and all will be ok with the
world. But this means that home grown configs cannot be deployed via
RPM due to conflict issues.
So if you have a cluster of 10,15,20 servers running an application like
this what does the Fedora community recommend? Use CFengine, rpm
--force, roll our own RPMs, scp scripts to copy over files? Right now I
prefer to have configs in RPM as it allows an easy deploy via
yum/kickstart and of course easy version control, verification and
rollback.
I've always been a fan of the most simple solution so I'd be interested
to hear what others are doing. I'm also considering either requesting a
change from various package maintainers or requesting this as a fedora
standard if anybody else thinks that would be useful.
Cheers,
More information about the fedora-list
mailing list