Re: tlug: Redhat 6.1: (Bad) First Impressions

["Frank Bennett (=?iso-2022-jp?B?GyRCJVUlaSVzJS8kWSVNJUMlSBsoQg==?= )" <bennett@example.com>]

On Fri, Oct 08, 1999 at 10:51:46PM +0900, Chiew Farn Chung wrote:

> >I decided to give Redhat 6.1 a try yesterday.  It looks like they rushed the
> >release (again). [1]
>
> Me too (relative to the day you wrote it). I also installed it on 
> the 6th of Oct.

[snip]

> This is a little long winded. But anyone noticed this problem? Jim's right.
> RH might have rushed it a little.

I didn't have the problems mentioned, but we certainly were burned as well
by 6.1 when we tried to install it. 

We upgraded our fileserver here last Thursday.  We have just three
partitions, one for user data, one for /usr/local, and one for the system. 
We had been running a mildly hacked version of TurboLinux on this box, and
we decided to try a clean install of Debian stable.  Their installation
kernel hung attempting to access the SCSI chain, so we dropped that and
shifted to RedHat 6.1.  This started off the install (I chose text, not
graphic mode, thinking to play it safe), and all went according to plan,
right down to the network fetch of packages.  This ran until 90% complete
and then THE BL***Y SCRIPT CRASHED.  Error messages from the depths of
Python.  We tried again, selecting only a minimal set of packages.  An
hour later, we returned to find the same error message.

At this point, the server system had been scrubbed, of course, and the
machine, which is critical to user service here, was no longer in running
condition. I spent the entire night and, fighting through lots of stupid
mistakes (brought on by fatigue), succeeded (just) in getting a Debian
boot disk together that put enough of their system onto the disk for me to
finish the job by hand. 

This routine upgrade turned into a nightmare because of failed automation
(an over-ambitious script in the case of RedHat, an over-ambitious kernel
in the case of Debian).  I chose to fight things out with the Debian
distro instead of RedHat because doing so would not require lots of
distro-specific knowledge (learning Python, learning their script,
learning the format of text streams fed to the script, etc.). This has
taught me, for my own work, that system-critical automation should be well
documented and easily deconstructable so that it doesn't leave the admin
completely mystified in an emergency.  Debian at least satisfied the
second half of that.  RedHat satisfied neither. 

Cheers,
-- 
-x80
Frank G Bennett, Jr         @@
Faculty of Law, Nagoya Univ () email: bennett@example.com
Tel: +81[(0)52]789-2239     () WWW:   http://rumple.soas.ac.uk/~bennett/

-------------------------------------------------------------------
Next Technical Meeting: October 9 (Sat), 13:30   place: Temple Univ.
* Linux Internationalisation Initiative (Li18nux) speaker: Akio Kido
* Japanese TrueType Fonts                     speaker: Adrian Havill
Next Technical Meeting: November 13 (Sat), 13:30 place: Temple Univ.
* Network Security                               speaker: Steve Baur
Next Nomikai:  December 17 (Fri), 19:00 Tengu TokyoEkiMae 03-3275-3691
-------------------------------------------------------------------
more info: http://www.tlug.gr.jp        Sponsor: Global Online Japan