We use puppet to manage all of our configs and systems, and keep our puppet
configs in git.  I'm a little depressed to hear you don't know anyone
personally who does that.

I actually laid down the law and mandated that nothing go into production
that isn't built by puppet (we FAI a minimal base system from "bare metal"
then kick off puppet), and I wouldn't have it any other way.  It's so much
easier to manage systems that way.  It's maybe 20% more work up front and
80% less down the road.  :)

Nicholas

--
Sent via mobile.
On Oct 3, 2012 11:22 PM, "Howard Bampton" <[email protected]> wrote:

> On Wed, Oct 3, 2012 at 7:00 PM, Jesse Becker <[email protected]> wrote:
> > On Wed, Oct 3, 2012 at 6:46 PM, Yves Dorfsman <[email protected]> wrote:
> >> On 2012-10-03 16:16, Howard Bampton wrote:
> >>> All my NIS maps were historically under RCS (periodically purged by
> >>> hand), with a secondary repository of daily snaps kept in a second
> >>> spot for disaster recovery/"how long has this defect been out there"
> >>> purposes. I'll probably do something similar to the second style for
> >>> my LDAP stuff some day (hard to tell when a field changed/got
> >>> added/whatever).
> >>>
> >>
> >> The sad part is that you consider this unorthodox.
> >> RCS, mercurial, git, etc... are so cheap to use that there is
> absolutely no
> >> reason not to use them for NIS, DNS, ldif files etc... I mean unless you
> >> hire those type of admins who never ever make even a single mistake.
> >>
> >> Personally I think any machine that needs any tweaking after a rebuild
> >> should have /etc under VC.
> >>
> >
> > To take it a step further:  any machine should have its configuration
> > management  tool rules/promises/recipes under VC.  Nothing on the
> > machine should be modified directly.
>
> I've worked with people that did one or more of the following:
> * hadn't goofed up badly enough that the "pain" of using some form of
> backups of such files was less painful than redoing something from
> memory/seat of the pants
> * had goofed up, but used comments in various files to track changes
> (nearly impossible to find the live content after a while and knowing
> when the changes were made is problematic)
> * used brute force and saved entire copies by wrapping "vi" (horribly
> wasteful of space at a time when disks were in the <10 GB range and
> large files change dozens of times a day)- history files got wiped
> after a month or two as a result, but part of the lesson was learned.
> * manually made backup copies with random naming schemes (for the same
> file on the same host- hosts.YYYYMMDD, hosts.save, hosts.old,
> hosts.older, holds.$username.<something>, ...)
> * knew I did something on the sly for select things, and came to me
> when things broke
> * hadn't done so...yet.
>
> I can't say any of the people I have worked with or know personally
> have ever done something to manage OS configs on an ongoing basis via
> chef/puppet/other-free-alternative/$$$$ome-commercial-product/something
> homebrewed. The closest I've seen was a solid Jumpstart setup and a
> nearly as good Kickstart one.
>
> Clearly such shops exist, just not at the handful of larger employers
> (or subsets of their infrastructure) I know about first or second
> hand.
> _______________________________________________
> Discuss mailing list
> [email protected]
> https://lists.lopsa.org/cgi-bin/mailman/listinfo/discuss
> This list provided by the League of Professional System Administrators
>  http://lopsa.org/
>
_______________________________________________
Discuss mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/discuss
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to