[linux-elitists] Show Us The Code

James Sparenberg james@linuxrebel.org
Thu Mar 1 14:24:34 PST 2007


On Wednesday 28 February 2007 22:39:20 Don Marti wrote:
> begin Greg Folkert quotation of Wed, Feb 28, 2007 at 08:47:55PM -0500:
> > On Wed, 2007-02-28 at 17:20 -0800, Rick Moen wrote:
> > > Quoting Teh Entar-Nick (nick@teh.entar.net):
> > > > Proof that that sign is working:
> > > > https://www.redhat.com/archives/fedora-devel-list/2007-February/threa
> > > >d.html#01006
> > >
> > > I remember, back around 1995, recursively chmod'ing some poorly
> > > specified glob while (inattentively) wielding root authority, in /lib.
> > > It was embarrassing and a chore to recover from.  Blaming the
> > > distribution never occurred to me.  Lack of imagination, no doubt.
> >
> > I have to admit, I did a similar fubar on an HP-UX machine ~1996 but
> > in /usr. I did it in the middle of the day.
> >
> > Printing stopped, the the DB failed to execute queries, then people
> > tried to reboot the windows machines they used and re-log in... none
> > could.
> >
> > It is the only time I restored from the previous Friday's bootable tape
> > the middle of a day, glad I made them that way and didn't rely on SAM to
> > do it.
>
> This is where I ask about RPM rollback.  Is anyone
> on the list who administers RPM-based systems turning
> this feature on?  Anyone used it for real?  It looks
> great (disk space is cheap compared to admin time or
> worse, downtime), but what's the catch?  (Other than
> Red Hat apparently not liking to turn it on by default
> in RHEL.)

Don,

   I haven't used it in the last 2 years, mainly due to the situation I'm in 
now were we do images and they are constantly being updated for every 
project.  However in a past life with another company I did some work with 
it. I found the folllowing caveats.

1.  As much as a 10 fold normally a tripling of the time it takes to update 
each rpm
2.  The roll backs are not well designed when it comes to managing over time. 
Meaning that I got the feeling they expected that one would wipe and 
re-install every 6 months (new distro time) 
3.  Documentation sucks on this feature.
4.  The role back rpms seemed to work the way one expected them too.  Boxes 
all came up and ran OK after the roll back.
5.  You can't pick and chose roll back.  Roll back is done by total system 
state per day/time not by app.    So you may roll back something you didn't 
expect.  
6.  I've not confirmed or used this with kernels or kernel modules.

James


-- 
  (o_
  //\
  V_/_
Linuxrebel



More information about the linux-elitists mailing list