On Fri, Sep 17, 2010 at 11:14 AM, Paul Heinlein heinlein@madboa.com wrote:
I know the OP asked for "cool" things to do, but I'll add my vote to those who suggested highlighting configuration management. I'm not sure how much puppet or cfengine you teach in a half-day, but I'm fairly confident you could cover:
Yes! If there's anything I wish were taught to new system administrators, it's that your configuration is your code.
1. considering configuration files to be code -- it needs to be in a repository
2. setting up a Subversion or git repository and some possible ways of laying out a configuration repository (per host, per service, etc)
3. committing changes, recovering older configs when newer ones introduce regressions
My general method is to keep a CVS committed directory somewhere on the root filesystem with all configurations. Then I symlink the tracked files back to that repository. For example:
/etc/hosts --> /configs/HOSTNAME/etc/hosts /etc/syslog.conf --> /configs/HOSTNAME/etc/syslog.conf
Restoring a machine's "identity" is just a simple matter of checking out that host's configuration directory then running a script to create the symlink.s
Personally, I like Subversion for configuration repositories because (imo) sysadmins usually like having an authoritative repo rather than a widely branched one -- but git is on the rise and is certainly worth considering.
Other, slightly related, suggestions include setting up a documentation wiki and/or a ticketing system. Trac can do both, but there are plenty of worthwhile alternatives.
-- Paul Heinlein <> heinlein@madboa.com <> http://www.madboa.com/ _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos